WO2026002004A1 - 数据传输方法、装置和系统 - Google Patents

数据传输方法、装置和系统

Info

Publication number
WO2026002004A1
WO2026002004A1 PCT/CN2025/103299 CN2025103299W WO2026002004A1 WO 2026002004 A1 WO2026002004 A1 WO 2026002004A1 CN 2025103299 W CN2025103299 W CN 2025103299W WO 2026002004 A1 WO2026002004 A1 WO 2026002004A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
recombination
mode based
data type
mode
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2025/103299
Other languages
English (en)
French (fr)
Inventor
李佳徽
马梦瑶
唐子涵
杜雨晴
谢俊文
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Original Assignee
Huawei Technologies Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co Ltd filed Critical Huawei Technologies Co Ltd
Publication of WO2026002004A1 publication Critical patent/WO2026002004A1/zh
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00—Packet switching elements
    • H04L49/90—Buffering arrangements
    • H04L49/9057—Arrangements for supporting packet reassembly or resequencing

Definitions

  • This application relates to the field of communication technology, and more specifically, to a data transmission method, apparatus, and system.
  • sensing/imaging data channel data
  • AI artificial intelligence
  • These data have characteristics such as large data volume, high redundancy, and correlation in the time domain, frequency domain, or spatial domain. These characteristics can be used to compress the data to be transmitted in order to reduce transmission overhead.
  • CSI channel state information
  • sensor data compression e.g., sensor data compression
  • AI data such as training data, model parameters, etc.
  • operations such as data reassembly, data filtering, data transformation, data selection, quantization, entropy coding, or channel mapping can be performed to obtain the bitstream to be transmitted. Then, after processing such as channel coding and modulation, it is transmitted over the wireless interface.
  • CSI channel state information
  • AI data such as training data, model parameters, etc.
  • This application provides a data transmission method, apparatus, and system to adapt to different task or business needs and ensure transmission performance.
  • a data transmission method is provided. This method can be executed by a first device (or encoding device).
  • the "first device” in this application can refer to the communication device itself (e.g., a terminal device or a network device), a component in the communication device (e.g., a communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.
  • the method includes: determining a first recombination pattern, the first recombination pattern being used to indicate the grouping method for data of a first data type, the first recombination pattern belonging to P recombination patterns, the P recombination patterns corresponding to the first data type, where P is an integer greater than 1; grouping the data of the first data type according to the first recombination pattern to obtain a first data group; and outputting the first data, the first data being obtained by processing the first data group.
  • the first device can determine a first reassembly mode and then use this first reassembly mode to group data of the first data type.
  • This implementation designs multiple reassembly modes (e.g., P reassembly modes) for the same data type (e.g., the first data type), allowing the first device to flexibly select the first reassembly mode that meets the transmission requirements according to different business or task transmission needs, thus supporting diverse data transmission requirements and ensuring data transmission performance.
  • the first device can also switch reassembly modes based on specific compression and transmission requirements, thereby improving end-to-end performance.
  • the first device can process the grouped data, such as data selection, compression, or quantization, supporting dynamic matching of transmission resource constraints. This allows limited transmission resources to be used to send more important data packets, achieving better signal drop effects and supporting flexible and efficient data compression and transmission at the physical layer, ensuring data transmission performance.
  • the first data is obtained by processing the first data group. It can be understood that: after the first type of data is grouped by the first recombination mode, it can be further processed by at least one of the following operations: data filtering, data transformation, data selection (or data discarding), quantization, entropy coding, channel mapping, channel coding, or modulation to obtain the first data.
  • the first data type includes one of the following: environmental reflection points, environmental patches, environmental reconstruction maps, radio frequency maps, AI training data, AI inference data, AI model parameters, AI gradient parameters, AI intermediate feature data, AI performance data, channel matrix, channel state information (CSI) matrix, or channel indication information.
  • the recombination pattern includes one or more of the following: a recombination pattern based on sequential splitting; a recombination pattern based on bit position; a recombination pattern based on spatial position; a recombination pattern based on semantic information; or a recombination pattern based on importance ranking.
  • the sequential partitioning-based reorganization mode refers to dividing the data into different groups according to the order of data storage or recording. This mode has low complexity and can support multiple data types.
  • the bit position-based reorganization mode divides the data into different groups based on the high and low values of bits, facilitating integration with bit layering or incremental transmission designs.
  • the spatial location-based reorganization mode divides the data into different groups based on its location information, suitable for perceptual data with geographic location information, and can be combined with geographic location information for joint data processing.
  • the semantic information-based reorganization mode divides the data into different groups based on its corresponding semantic category, suitable for data with semantic features (such as semantically segmented environmental reflection points, environmental patches, and reconstructed environmental maps, whose semantic information can indicate different objects, foreground/background, etc.), and can be combined with the data's semantic features for grouping and processing.
  • the importance-based reassembly mode means that relatively more important bitstreams use channel coding or modulation methods with higher reliability (lower channel code rate and modulation order); conversely, relatively less important bitstreams use channel coding or modulation methods with lower reliability (higher channel code rate or modulation order). It can obtain data bitstreams of different importance according to the specific application scenario of the data, and support unequal error protection (UEP) transmission design.
  • UDP unequal error protection
  • the first reconstruction mode when the first data type is perceptual data, if the first data type is environmental reflection points, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information.
  • the corresponding first reconstruction mode can be determined according to the specific application scenario of the environmental reflection points.
  • a reconstruction mode based on spatial position or semantic information can be selected to achieve better task accuracy; as another example, when environmental reflection point data is used for environmental reconstruction, a reconstruction mode based on sequential splitting can be selected to achieve better reconstruction performance; and as yet another example, when environmental reflection point data is used for incremental transmission, a reconstruction mode based on bit position can be selected to flexibly match different transmission resources.
  • Incremental transmission refers to obtaining multiple bit-layer information; the terminal device can send one or more bit layers at a time, and when the scheduled transmission resources change, higher-order bit-layer (higher importance) information can be sent first.
  • the first reconstruction mode when the first data type is perceptual data, if the first data type is an environment patch, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, or a reconstruction mode based on semantic information.
  • the first reconstruction mode can be determined based on the specific application scenario of the environment patch. For example, when environment patch data is used for target/object recognition tasks, a reconstruction mode based on semantic information can be selected to achieve better task accuracy; as another example, when environment patch data is used for environment reconstruction, a reconstruction mode based on sequential splitting can be selected to achieve better reconstruction performance.
  • the first reconstruction mode when the first data type is perceptual data, if the first data type is an environment reconstruction map, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information.
  • the first reconstruction mode can be determined according to the specific application scenario of the environment reconstruction map.
  • a reconstruction mode based on spatial position or semantic information can be selected to achieve better task accuracy; as another example, when the environment reconstruction map data is used for environment reconstruction, a reconstruction mode based on sequential splitting can be selected to achieve better reconstruction performance; when the environment reconstruction map data is used for incremental transmission, a reconstruction mode based on bit position can be selected to flexibly match different transmission resources.
  • the first reassembly mode when the first data type is sensor data, if the first data type is an RF map, then the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit location, a reassembly mode based on spatial location, or a reassembly mode based on importance ranking.
  • the first reassembly mode can be determined based on the specific application scenario of the RF map.
  • a reassembly mode based on spatial location or semantic information can be selected to achieve better task accuracy
  • a reassembly mode based on sequential splitting can be selected to achieve better reconstruction performance
  • a reassembly mode based on bit location can be selected to flexibly match different transmission resources.
  • the first recombination mode when the first data type is AI data, if the first data type is AI training data or AI inference data, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, a recombination mode based on semantic information, or a recombination mode based on importance ranking.
  • the first recombination mode can be determined based on the specific application scenario of the AI training data or AI inference data.
  • a recombination mode based on semantic information can be selected to achieve better task accuracy
  • a recombination mode based on sequential splitting can be selected to achieve better reconstruction performance
  • a recombination mode based on bit position can be selected to flexibly match different transmission resources.
  • the first reassembly mode when the first data type is AI data, if the first data type is AI model parameters or AI gradient parameters, then the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on importance ranking.
  • the first reassembly mode can be determined based on the specific application scenario of the AI model parameters or AI gradient parameters. For example, in a UEP transmission scenario, when it is necessary to obtain data streams of different importance for AI model parameters or AI gradient parameters, a reassembly mode based on importance ranking can be selected to achieve better transmission performance.
  • a reassembly mode based on sequential splitting can be selected to achieve lower complexity.
  • a reassembly mode based on bit position can be selected to flexibly match different transmission resources.
  • the first reassembly mode when the first data type is AI data, if the first data type is AI intermediate feature data, then the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on semantic information.
  • the first reassembly mode can be determined based on the specific application scenario of the AI intermediate feature data.
  • a reassembly mode based on semantic information can be selected to achieve better task accuracy
  • a reassembly mode based on sequential splitting can be selected to achieve better reconstruction performance
  • a reassembly mode based on bit position can be selected to flexibly match different transmission resources.
  • the first reorganization mode when the data of the first data type is AI data, if the first data type is AI performance data, then the first reorganization mode includes a reorganization mode based on sequential splitting.
  • the first reassembly mode when the first data type is channel data, if the first data type is a channel matrix or a CSI matrix, then the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on importance ranking.
  • the first reassembly mode can be determined based on the specific application scenario of the channel matrix or CSI matrix. For example, in a UEP transmission scenario, when it is necessary to obtain data streams of different importance in the channel matrix or CSI matrix, a reassembly mode based on importance ranking can be selected to achieve better transmission performance.
  • a reassembly mode based on sequential splitting can be selected to achieve lower complexity.
  • a reassembly mode based on bit position can be selected to flexibly match different transmission resources.
  • the first reassembly mode when the data of the first data type is channel data, if the first data type is channel indication information, then the first reassembly mode includes a reassembly mode based on sequential splitting.
  • first indication information is acquired or output.
  • the first indication information is used to indicate a first mapping relationship.
  • the first mapping relationship is used to indicate the correspondence between K recombination patterns and L data types.
  • the K recombination patterns include the first recombination pattern
  • the L data types include the first data type
  • the K recombination patterns contain P recombination patterns, where P is an integer less than or equal to K and L is an integer greater than or equal to 1.
  • the first mapping relationship can be predefined or preconfigured.
  • the first device and the second device can determine the correspondence between K recombination modes and L data types, and then determine one or more recombination modes corresponding to the first data type. They can then select the appropriate first recombination mode to group the data, realize the transmission requirements of the service, and meet different service transmission scenarios.
  • determining the first recombination pattern includes: determining the first recombination pattern based on a first data type and a first mapping relationship.
  • the first recombination mode can be predefined or preconfigured.
  • the first device can determine the first reassembly mode according to the first data type and the first mapping relationship, and then group the data according to the first reassembly mode to meet the transmission requirements of the service and satisfy different service transmission scenarios. It is understandable that both the first device and the second device simultaneously support this first reassembly mode, which helps to ensure data transmission performance.
  • the method includes: acquiring or outputting second indication information, the second indication information being used to indicate a first recombination mode.
  • the first device and the second device can determine the first reassembly mode by transmitting the second instruction information. Then, the first device can group the data of the first data type according to the first reassembly mode to obtain the first data group and output the first data to adapt to the service transmission requirements and ensure transmission performance.
  • the method before obtaining the second indication information, further includes: sending third indication information, the third indication information being used to indicate capability information of the first device, or one or more reconfiguration modes supported by the first device, wherein the one or more reconfiguration modes include the first reconfiguration mode.
  • the third indication information determines one or more reassembly modes supported by the first device, which in turn helps the first or second device to determine the first reassembly mode. For example, by comprehensively considering one or more reassembly modes supported by the first device, one or more reassembly modes supported by the second device, and one or more reassembly modes corresponding to the first data type, a suitable first reassembly mode can be finally determined, which helps to ensure transmission performance and improve user experience.
  • the method further includes: obtaining fourth indication information, the fourth indication information being used to indicate a second recombination mode, the second recombination mode being used to indicate a grouping method for data of the first data type; grouping the data of the first data type according to the second recombination mode to obtain a second data group, the second data group being different from the first data group; and outputting second data, the second data being obtained by processing the second data group.
  • the first data type corresponds to P reassembly modes, it can adapt to different task or service requirements. That is, the first data type supports flexible switching between different reassembly modes, thereby ensuring transmission performance. For example, for the same data type, if the service type on the network side changes, such as the initial feedback of the first data being used for data fusion, and the subsequent feedback of the second data being used to perform a specific detection task, then the fourth indication information can be sent to instruct the use of different reassembly modes to group the data of the first data type, so as to meet different task or service requirements and improve end-to-end performance.
  • the method before obtaining the fourth indication information, further includes: obtaining fifth indication information, which is used to indicate a second mapping relationship, and the second mapping relationship is used to indicate the correspondence between K’ recombination patterns and L’ data types, where K’ recombination patterns include the second recombination pattern, L’ data types include the first data type, K’ is an integer greater than 1, and L’ is an integer greater than or equal to 1.
  • the second device can update the correspondence between data types and reassembly modes to the first device by sending a fifth indication message. For example, it can indicate an update from a first mapping relationship to a second mapping relationship, or it can indicate an update from a third mapping relationship to a second mapping relationship, where the third mapping relationship can be obtained by updating the first mapping relationship.
  • the first device determines the second reassembly mode by obtaining the fourth indication message and the updated second mapping relationship, supporting the selection and switching of multiple reassembly modes for the same data type, in order to adapt to different task or service requirements and ensure transmission performance.
  • the method before obtaining the fourth indication information, the method further includes obtaining a sixth indication information, which indicates Q recombination patterns corresponding to the first data type, including the second recombination pattern, where Q is an integer greater than or equal to 2.
  • the sixth indication information may not be sent.
  • the second device can send a sixth indication message to the first device to indicate Q reassembly modes corresponding to the first data type. Furthermore, the first device can determine the updated second reassembly mode by obtaining the fourth indication message. This supports the selection and switching of multiple reassembly modes for the same data type, aiming to adapt to different task or service requirements and ensure transmission performance.
  • the Q reassembly modes can exist in the form of a set.
  • the first device acquires or outputs reassembly indication information, which indicates the number M of the first data packets and/or the size (or length) of each of the M first data packets.
  • the second device can determine the number M of the first data packets and/or the size (or length) of each of the M first data packets, which facilitates subsequent data packet reconstruction and ensures end-to-end data transmission performance.
  • a data transmission method is provided. This method can be executed by a second device (or decoding device).
  • the "second device” in this application can refer to the communication device itself (e.g., network device or terminal device), a component in the communication device (e.g., communication module, processor, circuit, chip, or chip system), or a logic module or software that can implement all or part of the functions of the communication device.
  • the method includes: acquiring first data; processing the first data to obtain a first data group; restoring the first data group according to a first recombination mode to obtain reconstructed data for the first data type; wherein, the first recombination mode is used to indicate the grouping method of the data for the first data type, the first recombination mode belongs to P recombination modes, the P recombination modes correspond to the first data type, and P is an integer greater than 1.
  • first indication information is acquired or output.
  • the first indication information is used to indicate a first mapping relationship.
  • the first mapping relationship is used to indicate the correspondence between K recombination patterns and L data types.
  • the K recombination patterns include the first recombination pattern
  • the L data types include the first data type
  • the K recombination patterns contain P recombination patterns, where P is an integer less than or equal to K and L is an integer greater than or equal to 1.
  • the first recombination pattern is determined based on the first data type and the first mapping relationship.
  • the method further includes: acquiring or outputting second indication information, the second indication information being used to indicate the first recombination mode.
  • the method before obtaining the second indication information, further includes: receiving third indication information, the third indication information being used to indicate capability information of the first device, or one or more recombination modes supported by the first device, wherein the one or more recombination modes include the first recombination mode.
  • the recombination pattern includes one or more of the following: a recombination pattern based on sequence splitting; a recombination pattern based on bit position; a recombination pattern based on spatial position; a recombination pattern based on semantic information; or a recombination pattern based on importance ranking.
  • the first data type includes one of the following: environmental reflection points, environmental patches, environmental reconstruction maps, radio frequency maps, artificial intelligence (AI) training data, AI inference data, AI model parameters, AI gradient parameters, AI intermediate feature data, AI performance data, channel matrix, channel state information (CSI) matrix, or channel indication information.
  • AI artificial intelligence
  • the first reconstruction mode when the data of the first data type is sensory data, if the first data type is an environmental reflection point, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information; if the first data type is an environmental patch, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, or a reconstruction mode based on semantic information; if the first data type is an environmental reconstruction map, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information; if the first data type is a radio frequency map, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on importance ranking.
  • the first recombination mode when the data of the first data type is AI data, if the first data type is AI training data or AI inference data, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, a recombination mode based on semantic information, or a recombination mode based on importance ranking; if the first data type is AI model parameters or AI gradient parameters, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, or a recombination mode based on importance ranking; if the first data type is AI intermediate feature data, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, or a recombination mode based on semantic information; if the
  • the first reassembly mode when the data of the first data type is channel data, if the first data type is a channel matrix or a CSI matrix, then the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on importance order; if the first data type is channel indication information, then the first reassembly mode includes a reassembly mode based on sequential splitting.
  • the method further includes: acquiring second data; processing the second data to obtain a second data group; and performing data recovery on the second data group according to a second recombination mode to obtain reconstructed data for the first data type; wherein the second recombination mode is used to indicate the grouping method for the data of the first data type.
  • the method before acquiring the second data, further includes: sending a fifth indication message, which indicates a second mapping relationship, indicating the correspondence between K’ recombination patterns and L’ data types, wherein the K’ recombination patterns include the second recombination pattern, and the L’ data types include the first data type, where K’ is an integer greater than 1 and L’ is an integer greater than or equal to 1.
  • the method before acquiring the second data, the method further includes: sending a sixth indication message, which indicates Q recombination patterns corresponding to the first data type, including the second recombination pattern, where Q is an integer greater than or equal to 2.
  • the second device acquires or outputs reassembly indication information indicating the number M of the first data packets and/or the size (or length) of each of the M first data packets.
  • the first device can determine the number M of the first data packets and/or the size (or length) of each of the M first data packets, which facilitates subsequent data packets and ensures end-to-end data transmission performance.
  • a communication device which has the functions of the first aspect above.
  • the communication device includes modules, units or means corresponding to the operations involved in the first aspect above.
  • the modules, units or means can be implemented by software, or by hardware, or by a combination of software and hardware.
  • the communication device may be the first device described above, such as a module or unit (e.g., a chip, a chip system, or a circuit) that corresponds to the method, operation, step, or action described in the first aspect above.
  • a module or unit e.g., a chip, a chip system, or a circuit
  • the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.
  • the processing unit is used to determine a first recombination pattern, which indicates the grouping method for data of a first data type.
  • the first recombination pattern belongs to P recombination patterns, which correspond to the first data type, where P is an integer greater than 1.
  • the processing unit is also used to group the data of the first data type according to the first recombination pattern to obtain a first data group.
  • the transceiver unit is used to output the first data, which is obtained by processing the first data group.
  • the transceiver unit is also used to acquire or output first indication information, which is used to indicate a first mapping relationship.
  • the first mapping relationship is used to indicate the correspondence between K recombination patterns and L data types.
  • the K recombination patterns include the first recombination pattern
  • the L data types include the first data type
  • the K recombination patterns contain P recombination patterns, where P is an integer less than or equal to K and L is an integer greater than or equal to 1.
  • the processing unit is also used to determine a first recombination mode based on a first data type and a first mapping relationship.
  • the transceiver unit is also used to acquire or output second indication information, which is used to indicate the first reassembly mode.
  • the transceiver unit is also configured to transmit third indication information, which indicates capability information of the first device, or one or more reconfiguration modes supported by the first device, wherein the one or more reconfiguration modes include the first reconfiguration mode.
  • the recombination pattern includes one or more of the following: a recombination pattern based on sequential splitting; a recombination pattern based on bit position; a recombination pattern based on spatial position; a recombination pattern based on semantic information; or a recombination pattern based on importance ranking.
  • the first data type includes one of the following: environmental reflection points, environmental patches, environmental reconstruction maps, radio frequency maps, artificial intelligence (AI) training data, AI inference data, AI model parameters, AI gradient parameters, AI intermediate feature data, AI performance data, channel matrix, channel state information (CSI) matrix, or channel indication information.
  • AI artificial intelligence
  • the first reconstruction mode when the data of the first data type is sensor data, if the first data type is an environmental reflection point, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information; if the first data type is an environmental patch, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, or a reconstruction mode based on semantic information; if the first data type is an environmental reconstruction map, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information; if the first data type is a radio frequency map, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on importance ranking.
  • the first recombination mode when the data of the first data type is AI data, if the first data type is AI training data or AI inference data, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, a recombination mode based on semantic information, or a recombination mode based on importance ranking; if the first data type is AI model parameters or AI gradient parameters, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, or a recombination mode based on importance ranking; if the first data type is AI intermediate feature data, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, or a recombination mode based on semantic information; if the
  • the first reassembly mode when the data of the first data type is channel data, if the first data type is a channel matrix or a CSI matrix, then the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on importance order; if the first data type is channel indication information, then the first reassembly mode includes a reassembly mode based on sequential splitting.
  • the transceiver unit is further configured to acquire fourth indication information, which indicates a second recombination mode and a grouping method for data of the first data type; the processing unit is further configured to group the data of the first data type according to the second recombination mode to obtain a second data group, which is different from the first data group; the transceiver unit is further configured to output the second data, which is obtained by processing the second data group.
  • the transceiver unit is also used to acquire fifth indication information, which is used to indicate a second mapping relationship.
  • the second mapping relationship is used to indicate the correspondence between K’ recombination patterns and L’ data types.
  • K’ recombination patterns include the second recombination pattern
  • L’ data types include the first data type.
  • K’ is an integer greater than 1
  • L’ is an integer greater than or equal to 1.
  • the transceiver unit is also used to acquire a sixth indication information, which indicates Q recombination patterns corresponding to the first data type, including a second recombination pattern, where Q is an integer greater than or equal to 2.
  • a communication device which has the functions of the second aspect above.
  • the communication device includes modules, units or means corresponding to the operations involved in the second aspect above.
  • the modules, units or means can be implemented by software, or by hardware, or by a combination of software and hardware.
  • the communication device may be the second device described above, such as a module or unit (e.g., a chip, a chip system, or a circuit) that corresponds to one-to-one execution of the methods, operations, steps, or actions described in the second aspect above.
  • a module or unit e.g., a chip, a chip system, or a circuit
  • the communication device includes a transceiver unit (or communication module) and a processing unit (or processing module) connected to the transceiver unit.
  • the processing unit is used to acquire first data; the processing unit is also used to process the first data to obtain a first data group; the processing unit is also used to perform data recovery on the first data group according to a first recombination mode to obtain reconstructed data for the first data type; wherein, the first recombination mode is used to indicate the grouping method of data for the first data type, the first recombination mode belongs to P recombination modes, the P recombination modes correspond to the first data type, and P is an integer greater than 1.
  • the transceiver unit is also used to acquire or output first indication information, which is used to indicate a first mapping relationship.
  • the first mapping relationship is used to indicate the correspondence between K recombination patterns and L data types.
  • the K recombination patterns include the first recombination pattern
  • the L data types include the first data type
  • the K recombination patterns contain P recombination patterns, where P is an integer less than or equal to K and L is an integer greater than or equal to 1.
  • the processing unit is also used to determine a first recombination mode based on a first data type and a first mapping relationship.
  • the transceiver unit is also used to acquire or output second indication information, which is used to indicate the first reassembly mode.
  • the transceiver unit is also configured to receive third indication information, which indicates capability information of the first device, or one or more reconfiguration modes supported by the first device, wherein the one or more reconfiguration modes include the first reconfiguration mode.
  • the recombination pattern includes one or more of the following: a recombination pattern based on sequential splitting; a recombination pattern based on bit position; a recombination pattern based on spatial position; a recombination pattern based on semantic information; or a recombination pattern based on importance ranking.
  • the transceiver unit is also used for a first data type including one of the following: environmental reflection points, environmental patches, environmental reconstruction maps, radio frequency maps, artificial intelligence (AI) training data, AI inference data, AI model parameters, AI gradient parameters, AI intermediate feature data, AI performance data, channel matrix, channel state information (CSI) matrix, or channel indication information.
  • AI artificial intelligence
  • the first reconstruction mode when the data of the first data type is sensor data, if the first data type is an environmental reflection point, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information; if the first data type is an environmental patch, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, or a reconstruction mode based on semantic information; if the first data type is an environmental reconstruction map, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information; if the first data type is a radio frequency map, then the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on importance ranking.
  • the first recombination mode when the data of the first data type is AI data, if the first data type is AI training data or AI inference data, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, a recombination mode based on semantic information, or a recombination mode based on importance ranking; if the first data type is AI model parameters or AI gradient parameters, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, or a recombination mode based on importance ranking; if the first data type is AI intermediate feature data, then the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, or a recombination mode based on semantic information; if the
  • the first reassembly mode when the data of the first data type is channel data, if the first data type is a channel matrix or a CSI matrix, then the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on importance order; if the first data type is channel indication information, then the first reassembly mode includes a reassembly mode based on sequential splitting.
  • the transceiver unit is further configured to acquire second data; the processing unit is further configured to process the second data to obtain a second data packet; the processing unit is further configured to perform data recovery on the second data packet according to a second reassembly mode to obtain reconstructed data for the first data type; wherein, the second reassembly mode is used to indicate the grouping method for the data of the first data type.
  • the transceiver unit is also used to send a fifth indication information, which indicates a second mapping relationship.
  • the second mapping relationship indicates the correspondence between K’ recombination patterns and L’ data types.
  • the K’ recombination patterns include the second recombination pattern, and the L’ data types include the first data type.
  • K’ is an integer greater than 1
  • L’ is an integer greater than or equal to 1.
  • the transceiver unit is also used to send a sixth indication message, which indicates Q recombination patterns corresponding to the first data type, including a second recombination pattern, where Q is an integer greater than or equal to 2.
  • a communication device is provided.
  • This communication device may be either the first or second device described above.
  • the communication device includes a transceiver, a processor, and a memory.
  • the processor controls the transceiver to transmit and receive signals, the memory stores a computer program, and the processor retrieves and runs the computer program from the memory, causing the communication device to perform the method in any possible implementation of either the first or second aspect described above.
  • processors there may be one or more processors and one or more memories.
  • the memory can be integrated with the processor, or the memory can be set up separately from the processor.
  • the communication device may also include a transmitter and a receiver.
  • a communication device including one or more processors, the one or more processors being configured to execute a computer program or instructions, which, when executed, cause the communication device to implement the methods in any possible design or implementation of the first or second aspect described above.
  • the communication device further includes a memory for storing part or all of the computer program or instructions implementing the functions involved in the first or second aspect described above.
  • the communication device may further include an interface circuit, wherein the processor is used to communicate with other devices or components through the interface circuit.
  • the aforementioned communication device may be a terminal, or a communication module in a terminal, or a chip in a terminal that is responsible for communication functions, such as a modem chip (also known as a baseband chip), or a system-on-chip (SoC) chip or system-in-a-package (SIP) chip that includes a modem module.
  • a modem chip also known as a baseband chip
  • SoC system-on-chip
  • SIP system-in-a-package
  • the aforementioned communication device may be a network device, or a communication module in a network device, or a circuit or chip in a network device responsible for communication functions, or a functional module in a network device capable of calling and executing programs.
  • a communication system in a seventh aspect, includes a first device and/or a second device, wherein the first device is configured to perform the method in any possible implementation of the first aspect, and the second device is configured to perform the method in any possible implementation of the second aspect.
  • the first or second device may be a terminal device, or a chip or circuit in the terminal device, or a functional module in the terminal device capable of calling and executing a program; or, the first or second device may be a network device, or a chip or circuit in the network device, or a central unit (CU) or distributed unit (DU) in the network device, or a functional module in the network device capable of calling and executing a program.
  • the first or second device may be a terminal device, or a chip or circuit in the terminal device, or a functional module in the terminal device capable of calling and executing a program
  • the first or second device may be a network device, or a chip or circuit in the network device, or a central unit (CU) or distributed unit (DU) in the network device, or a functional module in the network device capable of calling and executing a program.
  • CU central unit
  • DU distributed unit
  • a computer-readable storage medium stores computer program code or instructions to cause the method in any possible implementation of the first or second aspect to be executed, for example, when a computer reads and executes the computer program code or instructions, causing the method in any possible implementation of the first or second aspect to be implemented.
  • a ninth aspect provides a computer program product.
  • the computer program product includes computer program code or instructions to cause the method in any possible implementation of the first or second aspect to be implemented. For example, when a computer reads and executes the computer program product, the method in any possible implementation of the first or second aspect is implemented.
  • a computer program is provided. When the computer program is run, it causes the method in any of the possible implementations of the first or second aspect to be implemented.
  • Figure 1 is a schematic diagram of a communication system applicable to an embodiment of this application.
  • Figure 2 is a schematic diagram of a native data compression and transmission processing flow
  • FIG. 3 is a flowchart illustrating the communication method provided in an embodiment of this application.
  • Figure 4 is a schematic diagram of the data grouping structure based on the sequential splitting recombination mode provided in the embodiments of this application;
  • Figure 5 is a schematic diagram of the data grouping structure based on bit position recombination mode provided in an embodiment of this application;
  • Figure 6 is a schematic diagram of the data grouping structure based on spatial location reorganization pattern provided in an embodiment of this application;
  • Figure 7 is a schematic diagram of the structure for grouping data based on semantic information recombination patterns provided in an embodiment of this application;
  • Figures 8 and 9 are schematic diagrams of the data grouping structure based on the importance ranking reorganization pattern provided in the embodiments of this application;
  • Figure 10 is a schematic diagram of the grouping and filtering scheme provided in an embodiment of this application.
  • FIG 11 is a schematic diagram of the structure of the scalar quantization (SQ) and vector quantization (VQ) codebooks provided in the embodiments of this application;
  • Figure 12 is a schematic diagram of a grouping and filtering scheme under fixed resource constraints and quantized bit length provided in an embodiment of this application;
  • Figure 13 is a schematic diagram of a grouping and quantification scheme under fixed resource constraints provided in an embodiment of this application;
  • Figure 14 is a schematic diagram of the processing flow of the encoding unit provided in the embodiment of this application.
  • Figure 15 is a schematic diagram of the processing flow of the decoding unit provided in the embodiment of this application.
  • Figure 16 is a schematic block diagram of a communication device provided in an embodiment of this application.
  • Figure 17 is a schematic block diagram of another communication device provided in an embodiment of this application.
  • Figure 18 is a schematic block diagram of a chip system provided in an embodiment of this application.
  • Figure 19 is a schematic block diagram of another chip system provided in an embodiment of this application.
  • At least one of a, b, and c can mean: a, or, b, or, c, or, a and b, or, a and c, or, b and c, or, a, b, and c.
  • a, b, and c can be single or multiple.
  • instruction or “for instruction” can include both direct instruction and indirect instruction.
  • instruction can include whether the instruction directly instructs A or indirectly instructs A, but does not necessarily mean that the instruction carries A.
  • the indication methods involved in the embodiments of this application should be understood to cover various methods that enable the party to be indicated to obtain the information to be indicated.
  • the information to be indicated can be sent as a whole or divided into multiple sub-information and sent separately.
  • the sending period and/or sending time of these sub-information can be the same or different. This application does not limit the sending method, for example.
  • the "instruction information" in the embodiments of this application can be an explicit instruction, that is, a direct instruction through signaling, or an instruction obtained by combining other rules or parameters with the parameters indicated by the signaling, or by deduction. It can also be an implicit instruction, that is, an instruction obtained based on rules or relationships, or based on other parameters, or by deduction. This application does not specifically limit it in this regard.
  • protocol can refer to a standard protocol in the field of communications, such as the 5th generation (5G) protocol, the new radio (NR) protocol, and related protocols applied to future communication systems. This application does not limit the term “protocol”.
  • Predefined can include predefined terms, such as protocol definitions.
  • Preconfiguration can be implemented by pre-storing corresponding codes, tables, or other means that can be used to indicate relevant information in the device. This application does not limit the implementation method of this feature.
  • Sending information to XX (device) can be understood as the destination of the information being that device. This can include sending information directly or indirectly to that device.
  • Receiviving information from XX (device), or receiving information from XX (device) can be understood as the source of the information being that device. This can include receiving information directly or indirectly from that device.
  • Information may undergo necessary processing between the source and destination, such as format changes, but the destination can understand the valid information from the source. Similar expressions in this application can be interpreted similarly and will not be elaborated upon here.
  • Communication can also be described as “communication,” “information transmission,” “data processing,” etc.
  • Transmission includes “sending” and “receiving.”
  • Transmission can be described as “output.”
  • Send can also be understood as the "output” of a chip interface, and “receiving” can be understood as the "input” of a chip interface.
  • sending or “receiving” can occur between devices, for example, between network devices and terminal devices via an air interface.
  • sending” or “receiving” can also occur within a device, for example, between components, modules, chips, software modules, or hardware modules within a device via a bus, wiring, or interface.
  • the technical solutions of this application can be applied to various communication systems, such as 5G or NR systems, Long Term Evolution (LTE) systems, LTE Frequency Division Duplex (FDD) systems, and LTE Time Division Duplex (TDD) systems.
  • LTE Long Term Evolution
  • FDD Frequency Division Duplex
  • TDD Time Division Duplex
  • the technical solutions of this application can also be applied to device-to-device (D2D) communication, vehicle-to-everything (V2X) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), and Internet of Things (IoT) communication systems or other communication systems.
  • D2D device-to-device
  • V2X vehicle-to-everything
  • M2M machine-to-machine
  • MTC machine-type communication
  • IoT Internet of Things
  • the technical solutions of this application can also be applied to non-terrestrial network (NTN) systems such as inter-satellite communication and satellite communication.
  • NTN
  • the satellite base station provides communication services to the terminal equipment.
  • the satellite base station can also communicate with ground base stations.
  • the satellite can act as a base station or as a terminal equipment.
  • satellites can refer to drones, hot air balloons, low-orbit satellites, medium-orbit satellites, high-orbit satellites, etc., or non-ground base stations or non-ground equipment, etc.
  • a device can send signals to or receive signals from another device. These signals can include information, signaling, or data.
  • the device can also be replaced by an entity, network entity, network element, communication equipment, communication module, node, communication node, etc.; this application uses a device as an example for description.
  • a communication system can include at least one terminal device and at least one network device.
  • the network device can send downlink signals to the terminal device, and/or the terminal device can send uplink signals to the network device.
  • FIG. 1 is a schematic diagram of a communication system provided in an embodiment of this application.
  • the communication system 10 includes a radio access network (RAN) 100 and a core network (CN) 200.
  • RAN 100 includes at least one RAN node (110a and 110b in Figure 1, collectively referred to as 110) and at least one terminal (120a-120j in Figure 1, collectively referred to as 120).
  • RAN 100 may also include other RAN nodes, such as wireless relay devices and/or wireless backhaul devices (not shown in Figure 1).
  • Terminal 120 is wirelessly connected to RAN node 110.
  • RAN node 110 is wirelessly or wired connected to core network 200.
  • the core network device in core network 200 and RAN node 110 in RAN 100 can be different physical devices, or they can be the same physical device integrating core network logical functions and radio access network logical functions.
  • RAN 100 can be a cellular system related to the 3rd Generation Partnership Project (3GPP), such as a 4G mobile communication system, a 5G mobile communication system, or a future-oriented evolution system.
  • 3GPP 3rd Generation Partnership Project
  • RAN 100 can also be an open access network (O-RAN or ORAN), a cloud radio access network (CRAN), or a wireless fidelity (WiFi) system.
  • RAN 100 can also be a communication system that integrates two or more of the above systems.
  • RAN node 110 sometimes also referred to as access network equipment, RAN entity, or access node, constitutes part of the communication system and is used to help terminals achieve wireless access.
  • Multiple RAN nodes 110 in communication system 10 can be of the same type or different types. In some scenarios, the roles of RAN node 110 and terminal 120 are relative.
  • network element 120i in Figure 1 can be a helicopter or drone, which can be configured as a mobile base station.
  • network element 120i accessing RAN 100 through network element 120i
  • network element 120i is a base station; but for base station 110a, network element 120i is a terminal.
  • RAN node 110 and terminal 120 are sometimes both referred to as communication devices.
  • network elements 110a and 110b in Figure 1 can be understood as communication devices with base station functions
  • network elements 120a-120j can be understood as communication devices with terminal functions.
  • a RAN node can be a base station, an evolved NodeB (eNodeB), an access point (AP), a transmission reception point (TRP), a base station in a future mobile communication system, or an access node in a WiFi system.
  • a RAN node can be a macro base station (as shown in Figure 1, 110a), a micro base station or indoor station (as shown in Figure 1, 110b), a relay node or donor node, or a radio controller in a CRAN scenario.
  • a RAN node can also be a server, wearable device, vehicle, or in-vehicle equipment.
  • the access network equipment in vehicle-to-everything (V2X) technology can be a roadside unit (RSU).
  • V2X vehicle-to-everything
  • RSU roadside unit
  • RAN nodes collaborate to assist the terminal in achieving wireless access, with each RAN node performing a portion of the base station's functions.
  • RAN nodes can be CUs, DUs, CUs (control plane, CP), CUs (user plane, UP), or radio units (RUs).
  • CUs and DUs can be separate entities or included in the same network element, such as a baseband unit (BBU).
  • RUs can be included in radio frequency equipment or radio frequency units, such as remote radio units (RRUs), active antenna units (AAUs), or remote radio heads (RRHs).
  • RRUs remote radio units
  • AAUs active antenna units
  • RRHs remote radio heads
  • CU including open CU-CP (O-CU-CP) and open CU-UP (O-CU-UP), DU, or RU
  • CU can also be called an open central unit (O-CU)
  • DU can also be called an open distributed unit (O-DU)
  • CU-CP can also be called O-CU-CP
  • CU-UP can also be called O-CU-UP
  • RU can also be called O-RU.
  • this application uses CU, CU-CP, CU-UP, DU, and RU as examples. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software and hardware modules.
  • Terminal 120 can be a device or module that accesses the aforementioned communication system and has corresponding communication functions.
  • a terminal can also be referred to as user equipment (UE), terminal, user device, access terminal, user unit, user station, mobile station, mobile station (MS), remote station, remote terminal, mobile device, user terminal, terminal unit, terminal station, terminal device, wireless communication equipment, user agent, or user device.
  • UE user equipment
  • MS mobile station
  • remote terminal remote terminal
  • mobile device user terminal, terminal unit, terminal station, terminal device, wireless communication equipment, user agent, or user device.
  • a terminal typically contains a communication module, circuit, or chip that performs the corresponding communication functions.
  • the terminal may also be configured with program instructions for performing these communication functions.
  • the terminal in this application embodiment can be a mobile phone, a personal digital assistant (PDA) computer, a laptop computer, a tablet computer, a drone, a computer with wireless transceiver capabilities, an MTC terminal, a virtual reality (VR) terminal, an augmented reality (AR) terminal, an internet of things (IoT) terminal, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical care, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home (such as a game console, smart TV, smart speaker, smart refrigerator, and fitness equipment), a transportation vehicle with wireless communication capabilities, a communication module, or an RSU with terminal functionality.
  • PDA personal digital assistant
  • laptop computer a laptop computer
  • a tablet computer a drone
  • a computer with wireless transceiver capabilities an MTC terminal
  • VR virtual reality
  • AR augmented reality
  • IoT internet of things
  • RAN 100 and terminal 120 can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; and they can also be deployed in the air on aircraft, balloons, and satellites.
  • the embodiments of this application do not limit the scenarios in which RAN 100 and terminal 120 are located.
  • CN 200 can be a 5G core network or an evolved 5G core network.
  • CN 200 includes access and mobility management (AMF) network elements responsible for mobility management and access management services; session management (SMF) network elements responsible for session management; user plane function (UPF) network elements responsible for user plane packet routing and forwarding and quality of service (QoS) control; and policy control (PCF) network elements.
  • AMF access and mobility management
  • SMF session management
  • UPF user plane function
  • QoS quality of service
  • PCF policy control
  • These core network elements can work independently or be combined to implement certain control functions.
  • AMF, SMF, and PCF can be combined into a single core network device.
  • Figure 1 is merely an example and does not constitute a limitation on the scope of protection of this application.
  • the communication method provided in the embodiments of this application may also involve devices not shown in Figure 1, such as wireless relay devices and/or wireless backhaul devices, etc.
  • the communication method provided in the embodiments of this application may also include only some of the devices shown in Figure 1.
  • native data e.g., native data from the wireless access network or local traffic
  • future wireless communication processes which will bring new demands for transmission.
  • Figure 2 is a schematic diagram of the raw data compression and transmission processing flow.
  • the input data undergoes operations such as data reassembly, data filtering, data transformation, data selection (or data discarding, etc.), quantization, entropy coding, and channel mapping to obtain the bitstream to be transmitted.
  • the input data can be raw data, such as sensor data, AI data, or channel data.
  • a binary bitstream is first obtained through demodulation and channel decoding.
  • the binary bitstream undergoes channel demapping, entropy decoding, inverse quantization, data recovery, inverse data transformation, zero-padding, and data reassembly and recovery operations to output reconstructed data.
  • the reconstructed data and the input data have the same dimension, but the corresponding data values may have slight errors.
  • a given piece of raw data can correspond to one or more data types. For example, different data types may be used during transmission depending on the application scenario.
  • a fixed data reconstruction scheme is typically used for data reconstruction. This cannot support dynamic switching between multiple reconstruction schemes, and therefore cannot adapt to different task or business needs. Therefore, it is urgent to define multiple reconstruction schemes for raw data of specific data types to support a standardized process for physical layer compression.
  • this application provides a data transmission method and apparatus to adapt to different task or business needs and ensure data transmission performance.
  • the embodiments shown below do not specifically limit the structure of the execution subject of the method provided in the embodiments of this application, as long as it is possible to communicate according to the method provided in the embodiments of this application by running the code or program that records the method provided in the embodiments of this application.
  • the method provided in the embodiments of this application can be executed by a first device (or encoding device) and a second device (or decoding device).
  • first device or “second device” in this application can refer to the device itself (e.g., terminal device or network device), or a component in the device (e.g., communication module, processor, circuit, chip (such as modem chip, also known as baseband chip, or SoC chip or SIP chip containing modem core), or chip system, etc.), or it can be a logic module or software that can implement all or part of the device functions.
  • a component in the device e.g., communication module, processor, circuit, chip (such as modem chip, also known as baseband chip, or SoC chip or SIP chip containing modem core), or chip system, etc.)
  • modem chip also known as baseband chip, or SoC chip or SIP chip containing modem core
  • chip system e.g., a logic module or software that can implement all or part of the device functions.
  • FIG 3 is a flowchart illustrating a data transmission method provided in an embodiment of this application. As shown in Figure 3, the method 300 includes the following steps.
  • the first device determines the first recombination mode.
  • the first recombination pattern is used to indicate the grouping method for data of the first data type.
  • the first recombination pattern belongs to P recombination patterns, and each of the P recombination patterns corresponds to the first data type, where P is an integer greater than 1.
  • the first device after determining the first recombination mode, the first device also determines the grouping method corresponding to the data of the first data type (e.g., data #a).
  • the first data type can correspond to multiple recombination modes, each recombination mode indicating a recombination scheme; that is, each recombination mode corresponds to a grouping method for data #a, and the first recombination mode is one of multiple recombination modes.
  • the data of the first data type can be raw data, such as one of the following: sensor data, AI data, or channel data, or other raw data, without limitation.
  • the data type corresponding to data #a can be one or more.
  • the first data type corresponding to the perception data is one of environmental reflection points, environmental patches, environmental reconstruction maps, or radio frequency maps.
  • the first data type corresponding to the AI data is one of AI training data, AI inference data, AI model parameters, AI gradient parameters, AI intermediate feature data, or AI performance data.
  • the data type corresponding to the channel data includes at least one of the channel matrix (such as the H matrix fed back by the device in a multi-antenna system), CSI matrix, or channel indication information (such as a channel quality indicator (CQI) or rank indicator (RI)), then the first data type corresponding to the channel data is one of a channel matrix, CSI matrix, or channel indication information.
  • the recombination mode includes one or more of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, a recombination mode based on spatial position, a recombination mode based on semantic information, or a recombination mode based on importance ranking.
  • the first recombination mode can be one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, a recombination mode based on spatial position, a recombination mode based on semantic information, or a recombination mode based on importance ranking.
  • Figure 4 is a schematic diagram of data grouping based on a sequential splitting reorganization pattern.
  • the first data type is AI intermediate feature data in AI data.
  • the length (or number) of this intermediate feature data is N, and it is divided into M groups equally from top to bottom. That is, the number of the first data groups is M, and each group contains N/M intermediate feature data.
  • the first data type is the channel matrix (e.g., H matrix) in channel data.
  • the dimension of this H matrix is N1*N2.
  • the second dimension it is divided into M groups equally from left to right.
  • Each group contains N2/M columns of data.
  • the dimension of the H matrix is N1*(N2/M).
  • the first dimension of this H matrix it is divided into M groups equally from top to bottom.
  • Each group contains N1/M rows of data.
  • the dimension of the H matrix is (N1/M)*N2/M.
  • the grouping method corresponding to the above reorganization mode is to group the data in an equal division manner, that is, each of the M groups contains data of equal length (or the size of each group).
  • the grouping method corresponding to the above reorganization mode can also be to group the data in an unequal division manner, that is, at least two of the M groups have different data lengths, and there is no restriction on this.
  • the first and second devices need to align the grouping method for the data of the first data type.
  • the number M of groups included in the above grouping method, and the size of each group can be predefined or preconfigured, in which case there is no need to explicitly indicate the recombination indication information #1;
  • the method 300 further includes: the first/second device acquiring or outputting the recombination indication information #1, for Figure 4(a), the recombination indication information #1 indicates the group size N/M or the number of groups M; for Figure 4(b), the recombination indication information #1 indicates the group size N2/M (or N1/M) or the number of groups M.
  • the recombination indication information #1 indicates the size of each of the M groups.
  • the data is first represented as a bit stream (such as using floating-point numbers or after quantization), and then divided into different groups according to the high and low values of the bits, which is convenient for designing in conjunction with bit layering or incremental transmission.
  • Figure 5 is a schematic diagram of data grouping based on bit position recombination patterns.
  • the grouping method corresponding to the bit position recombination patterns in the following examples adopts an equal division method, that is, each of the M groups contains data of equal length.
  • the first data type is the environmental reflection points in the sensory data.
  • the length (or number) of these environmental reflection points is N, for example, P1 , P2 , ..., PN .
  • the N environmental reflection points are sequentially quantized using Q bits, where M ⁇ Q.
  • the N data can be used to obtain an N*Q bit matrix.
  • the second dimension of this bit matrix (corresponding to the quantization bit layer) is divided into M equal groups in sequence. That is, the number of the first data groups is M, and each group contains Q/M bit layers, for a total of N*(Q/M) bits.
  • the first dimension of the bit matrix (corresponding to the quantization bit layer) can also be divided into M equal groups in sequence, with each group containing N/M bit layers, for a total of (N/M)*Q bits.
  • the grouping method corresponding to Figure 5(a) above can also be grouped in an unequal manner, that is, at least two of the M groups have different data lengths, without any restriction.
  • Q ⁇ M which means that each group can contain a different number of bit layers.
  • group 1 can include two bit layers
  • group M can include one bit layer.
  • the first data type is the AI model parameters in the AI data.
  • These AI model parameters are an N1*N2 matrix.
  • Each parameter is quantized to Q bits, where M ⁇ Q. Therefore, N1*N2 data points can yield an N1*N2*Q three-dimensional bit matrix.
  • the third dimension of this bit matrix (corresponding to the quantization bit layer) is sequentially and equally divided into M groups.
  • the first dimension of the bit matrix can also be sequentially and equally divided into M groups, each group containing N1/M bit layers, for a total of (N1/M)*N2*Q bits.
  • the second dimension of the bit matrix can be divided into M equal groups in sequence, each group containing N2/M bit layers, for a total of N1*(N2/M)*Q bits.
  • the grouping method corresponding to (c) in Figure 5 above can also be grouped in an unequal manner, that is, at least two of the M groups have different data lengths, without any restriction.
  • Q ⁇ M indicating that each group can contain a different number of bit layers.
  • group 1 can include two bit layers, that is, two planes;
  • group 4 can include one bit layer, that is, one plane;
  • group M can include three bit layers, that is, three planes, and so on, not shown in the figure.
  • the first device and the second device need to align the grouping method of the data for the first data type.
  • the number of groups M contained in the above grouping method and the size of each group can be predefined or preconfigured, in which case there is no need to explicitly indicate the recombination indication information #2; as another example, the method 300 further includes: the first device/second device acquiring or outputting the recombination indication information #2, for FIG5(a), the recombination indication information #2 indicates the group size Q/M (or N/M) or the number of groups M; for FIG5(c), the recombination indication information #2 indicates the group size Q/M (or N1/M, or N2/M) or the number of groups M.
  • the recombination indication information #2 indicates the size of each of the M groups.
  • Figure 6 is a schematic diagram of the data grouping structure based on spatial location reorganization pattern.
  • the first data type is the radio frequency map (RF map) in the sensing data (e.g., an RF map constructed based on two-dimensional coordinates).
  • This RF map includes RF parameters 1 to RF parameters 14. Multiple RF parameters that are physically close to each other are grouped into 4 groups, that is, the number of the first data groups is M.
  • group 1 includes radio frequency parameters 1 to 4, with corresponding two-dimensional coordinate information of (x1, y1), (x1, y2), (x1, y3), and (x1, y4), respectively;
  • group 2 includes radio frequency parameters 5 to 7, with radio frequency parameter 5 corresponding to two-dimensional coordinate information of (x2, y1) and (x2, y2), radio frequency parameter 6 corresponding to two-dimensional coordinate information of (x2, y3), and radio frequency parameter 7 corresponding to two-dimensional coordinate information of (x2, y4);
  • group 3 includes radio frequency parameters 8, 9, and 12, with radio frequency parameter 8 corresponding to two-dimensional coordinate information of (x3, y1) and (x4, y1), radio frequency parameter 9 corresponding to two-dimensional coordinate information of (x3, y2), and radio frequency parameter 12 corresponding to two-dimensional coordinate information of (x4, y2);
  • group 4 includes radio frequency parameters 10, 11, 13, and 14, with corresponding two-dimensional coordinate information of (x3, y3), (x3, y4), (x4, y3), and
  • the grouping method corresponding to the above recombination mode is to group the data in an unequal manner, that is, at least two of the M groups have different data lengths (or the number of radio frequency parameters).
  • the grouping method corresponding to the above recombination mode can also be to group the data in an equal manner, that is, each of the M groups contains the same number of radio frequency parameters. There is no restriction on this.
  • the first device and the second device need to align the grouping method of the data for the first data type.
  • the number M of groups included in the above grouping method, as well as the radio frequency parameters and coordinate range information of each group can be predefined or preconfigured, in which case there is no need to explicitly indicate the reassembly indication information #3; as another example, the method 300 further includes: the first device/second device acquiring or outputting the reassembly indication information #3, which may include the radio frequency parameter index corresponding to each group, for example, indicated in the form of a set of radio frequency parameter indexes, such as the index set corresponding to group 1 being ⁇ 1,2,3,4 ⁇ , the index set corresponding to group 2 being ⁇ 5,6,7 ⁇ , the index set corresponding to group 3 being ⁇ 8,9,12 ⁇ , and the index set corresponding to group 4 being ⁇ 10,11,13,14 ⁇ .
  • the corresponding geographic coordinate information can be determined by sending the radio
  • different groups are obtained by dividing the data according to the semantic category. It is suitable for data with semantic features (such as environmental reflection points, environmental patches, environmental reconstruction maps, etc. that have undergone semantic segmentation.
  • the semantic information corresponding to these data can indicate different objects, foreground/background, etc.).
  • the data can be grouped and processed by combining the semantic features of the data.
  • Figure 7 is a schematic diagram of data grouping based on semantic information recombination patterns.
  • the first data type is the environmental patch in the perceptual data.
  • This environmental patch includes building patches (e.g., building 1 patch and building 2 patch), street light patches (e.g., street light 1 patch, street light 2 patch and street light 3 patch), and street surface patches.
  • building patches e.g., building 1 patch and building 2 patch
  • street light patches e.g., street light 1 patch, street light 2 patch and street light 3 patch
  • street surface patches e.g., street light 1 patch, street light 2 patch and street light 3 patch
  • the grouping method corresponding to the above reorganization mode is to group the data in an unequal manner, that is, at least two groups in the M groups have different data lengths (or the number of environment patches).
  • the grouping method corresponding to the above reorganization mode can also be to group the data in an equal manner, that is, each group in the M groups contains the same number of environment patches, and there is no restriction on this.
  • the first device and the second device need to align the grouping method of the data for the first data type.
  • the number of groups M contained in the above grouping method, and the environmental patches contained in each group can be predefined or preconfigured; or, the environmental patch data implicitly contains the coordinate information of the environmental patch, in which case there is no need to explicitly indicate the reorganization instruction information #4.
  • This reconfiguration mode can be combined with the specific application scenario of the data to obtain data streams of different importance, and supports unequal protection UEP transmission design, that is, the relatively more important data streams use channel coding or modulation methods with higher reliability; conversely, the relatively less important data streams use channel coding or modulation methods with lower reliability.
  • Method 1 Use signal power or strength to characterize importance. Sort the data of length N according to signal power or strength and divide it into different groups. For example, group the data with high signal power or strength (higher importance) into one group and the data with low signal power or strength (lower importance) into another group.
  • FIG 8 is a structural diagram illustrating data grouping based on an importance-ranking reorganization pattern.
  • the first data type is radio frequency (RF) maps in sensing data (e.g., RF map data in multipath form).
  • Each path in this RF map includes information such as multipath intensity, phase, delay, angle of arrival, and angle of departure.
  • One or more paths with higher multipath intensity are grouped into one group (as relatively more important information), and one or more paths with lower multipath intensity are grouped into another group (as relatively less important information).
  • dividing the data according to multipath intensity yields M groups, meaning the first data group has M groups.
  • Each group contains data for one or more paths.
  • the data in group 1 corresponds to the highest multipath intensity (i.e., the most important group), and so on, with the data in group M corresponding to the lowest multipath intensity (i.e., the least important group).
  • Method 2 Characterize importance by contribution to downstream AI tasks. For example, for AI-related tasks such as target recognition and localization, we can first assess the impact of different data components on task accuracy (the greater the impact, the greater the contribution; conversely, the smaller the contribution), and use this to measure importance and group the data.
  • Figure 9 illustrates another structural diagram of data grouping based on an importance-ranking reorganization pattern.
  • the first data type is intermediate AI feature data within the AI data, with a length of N.
  • Each intermediate feature data is set to zero, and the decrease in AI task accuracy is tested.
  • the N intermediate AI feature data are sorted and grouped into M groups, meaning the number of groups in the first data set is M.
  • the intermediate AI feature data with the largest decrease in accuracy is considered the most important information, and so on, with the intermediate AI feature data with the smallest decrease in accuracy being considered the least important information.
  • the index of each intermediate AI feature data is recorded.
  • Each group can contain the index of the intermediate AI feature data.
  • the number of data items contained in each group is not limited; they can be equal or unequal.
  • the first device and the second device need to align the grouping method for the data of the first data type.
  • the number M of groups included in the above grouping method, and the data included in each group can be predefined or preconfigured, in which case there is no need to explicitly indicate the reorganization instruction information #5; as another example, the method 300 further includes: the first device/second device acquiring or outputting the reorganization instruction information #5, which may include the index or number of the data corresponding to each of the M groups.
  • the recombination patterns and corresponding grouping methods provided in Figures 4 to 9 above are merely examples for ease of understanding, and other recombination schemes are not excluded.
  • this application does not limit the number of recombination patterns, the number M of groups in each recombination pattern, or the number of data items contained in each of the M groups.
  • the recombination scheme of this application can group data evenly or unevenly; for example, data #a of length N can be grouped randomly.
  • the data corresponding to each of the M groups is determined by a random seed, and the first device and the second device can achieve synchronization by sending the number or index corresponding to the random seed.
  • the first device and the second device can also negotiate to align the groups in any other way, which is within the scope of protection of this application and is not limited thereto.
  • the following provides examples illustrating the rules and effects for determining the first data type and its corresponding first recombination pattern.
  • Scenario 1 The data of the first data type is perceptual data.
  • the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit position, a reconstruction mode based on spatial position, or a reconstruction mode based on semantic information.
  • the corresponding first reconstruction mode can be determined according to the specific application scenario of the environmental reflection points.
  • a reconstruction mode based on spatial position or semantic information can be selected to achieve better task accuracy; as another example, when environmental reflection point data is used for environmental reconstruction, a reconstruction mode based on sequential splitting can be selected to achieve better reconstruction performance; and as yet another example, when environmental reflection point data is used for incremental transmission, a reconstruction mode based on bit position can be selected to flexibly match different transmission resources.
  • Incremental transmission refers to obtaining multiple bit-layer information.
  • the terminal device can send one or more bit layers at a time. When the scheduled transmission resources change, higher-order bit-layer (higher importance) information can be sent first.
  • the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, or a reconstruction mode based on semantic information.
  • the first reconstruction mode can be determined according to the specific application scenario of the environment patch. For example, when environment patch data is used for target and object recognition tasks, a reconstruction mode based on semantic information can be selected to achieve better task accuracy; as another example, when environment patch data is used for environment reconstruction, a reconstruction mode based on sequential splitting can be selected to achieve better reconstruction performance.
  • the first reconstruction mode includes one of the following: a reconstruction mode based on sequential splitting, a reconstruction mode based on bit location, a reconstruction mode based on spatial location, or a reconstruction mode based on semantic information.
  • the first reconstruction mode can be determined according to the specific application scenario of the environment reconstruction map.
  • a reconstruction mode based on spatial location or semantic information can be selected to achieve better task accuracy; as another example, when the environment reconstruction map data is used for environment reconstruction, a reconstruction mode based on sequential splitting can be selected to achieve better reconstruction performance; when the environment reconstruction map data is used for incremental transmission, a reconstruction mode based on bit location can be selected to flexibly match different transmission resources.
  • the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit location, a reassembly mode based on spatial location, or a reassembly mode based on importance ranking.
  • the first reassembly mode can be determined according to the specific application scenario of the RF map.
  • a reassembly mode based on spatial location or semantic information can be selected to achieve better task accuracy
  • a reassembly mode based on sequential splitting can be selected to achieve better reconstruction performance
  • a reassembly mode based on bit location can be selected to flexibly match different transmission resources.
  • Scenario 2 The data of the first data type is AI data.
  • the first recombination mode includes one of the following: a recombination mode based on sequential splitting, a recombination mode based on bit position, a recombination mode based on semantic information, or a recombination mode based on importance ranking.
  • the first recombination mode can be determined according to the specific application scenario of the AI training data or AI inference data.
  • a recombination mode based on semantic information can be selected to achieve better task accuracy
  • a recombination mode based on sequential splitting can be selected to achieve better reconstruction performance
  • a recombination mode based on bit position can be selected to flexibly match different transmission resources.
  • the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on importance ranking.
  • the first reassembly mode can be determined according to the specific application scenario of the AI model parameters or AI gradient parameters. For example, in a UEP transmission scenario, when it is necessary to obtain data streams of different importance of AI model parameters or AI gradient parameters, a reassembly mode based on importance ranking can be selected to achieve better transmission performance.
  • Another example is in a non-UEP transmission scenario, when the data streams corresponding to AI model parameters or AI gradient parameters adopt the same coding and modulation configuration, a reassembly mode based on sequential splitting can be selected to achieve lower complexity. Yet another example is when AI model parameters or AI gradient parameters are used for incremental transmission, a reassembly mode based on bit position can be selected to flexibly match different transmission resources.
  • the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on semantic information.
  • the first reassembly mode can be determined according to the specific application scenario of the AI intermediate feature data.
  • a reassembly mode based on semantic information can be selected to achieve better task accuracy
  • a reassembly mode based on sequential splitting can be selected to achieve better reconstruction performance
  • a reassembly mode based on bit position can be selected to flexibly match different transmission resources.
  • Example 4 If the first data type is AI performance data, then the first reorganization pattern includes a reorganization pattern based on sequential splitting.
  • Scenario 3 The data of the first data type is channel data.
  • the first reassembly mode includes one of the following: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, or a reassembly mode based on importance ranking.
  • the first reassembly mode can be determined according to the specific application scenario of the channel matrix or CSI matrix. For example, in a UEP transmission scenario, when it is necessary to obtain data streams of different importance in the channel matrix or CSI matrix, a reassembly mode based on importance ranking can be selected to achieve better transmission performance.
  • a reassembly mode based on sequential splitting can be selected to achieve lower complexity.
  • a reassembly mode based on bit position can be selected to flexibly match different transmission resources.
  • Example 2 If the first data type is channel indication information, then the first reassembly mode includes a reassembly mode based on sequential splitting.
  • the following describes how the first device determines the first recombination mode in step S310.
  • the first recombination pattern can be predefined or preconfigured. Predefinition can include pre-defined parameters, such as protocol definitions, while preconfiguration can be achieved by pre-storing relevant code, tables, functions, text, strings, or other means that can be used to indicate related information (e.g., the first recombination pattern) in the first device (and/or the second device).
  • predefinition can include pre-defined parameters, such as protocol definitions, while preconfiguration can be achieved by pre-storing relevant code, tables, functions, text, strings, or other means that can be used to indicate related information (e.g., the first recombination pattern) in the first device (and/or the second device).
  • the first device determines the first recombination mode based on the first data type and the first mapping relationship.
  • the first mapping relationship indicates the correspondence between K recombination patterns and L data types.
  • the K recombination patterns include the first recombination pattern
  • the L data types include the first data type
  • the K recombination patterns contain P recombination patterns, where P is an integer less than or equal to K, and L is an integer greater than or equal to 1.
  • the method 300 further includes: the first device acquiring the first mapping relationship.
  • the first mapping relationship may be predefined or preconfigured; or, the first device acquires indication information #a (i.e., an example of first indication information), which indicates the first mapping relationship, i.e., the first mapping relationship may be indicated or configured by a network device via signaling.
  • the second device sends indication information #a to the first device, which indicates (or includes) the first mapping relationship; correspondingly, the first device receives the indication information #a and determines the first mapping relationship based on the indication information #a.
  • the method 300 further includes: the first device acquiring the first data type.
  • the first data type may be predefined or preconfigured; or, the first device acquires indication information #b, which indicates the first data type, i.e., the first data type may be indicated or configured by the network device through signaling.
  • the second device sends indication information #b to the first device, which indicates the first data type, and correspondingly, the first device receives the indication information #b and determines the first data type based on the indication information #b.
  • the first device can feed back the determined first recombination mode to the second device.
  • the first device sends indication information #f (i.e., an example of second indication information) to the second device, which indicates the first recombination mode.
  • indication information #f i.e., an example of second indication information
  • the method 300 further includes: the second device sending one or more reassembly modes supported by the network side.
  • the one or more reassembly modes include a first reassembly mode.
  • the second device can indicate the one or more reassembly modes supported by the network side by sending a system message, a broadcast message, or a multicast message. That is, the first device can determine the first reassembly mode based on the one or more reassembly modes supported by the network side, combined with a first data type and a first mapping relationship.
  • the corresponding multiple reassembly modes include: a reassembly mode based on sequential splitting, a reassembly mode based on bit position, and a reassembly mode based on importance ranking.
  • the reassembly modes supported by the network side are the reassembly mode based on sequential splitting and the reassembly mode based on importance ranking
  • the first reassembly mode finally determined by the first device can be either the reassembly mode based on sequential splitting or the reassembly mode based on importance ranking.
  • the first device acquires indication information #c (an example of second indication information), which indicates a first reconfiguration mode. That is, the first reconfiguration mode can be indicated or configured by the network device via signaling.
  • the second device sends indication information #c to the first device, which indicates (or includes) an index or identifier of the first reconfiguration mode.
  • the first device receives the indication information #c and determines the first reconfiguration mode based on it.
  • the method 300 further includes: the second device determining a first reconstruction mode.
  • the first reconstruction mode may be predefined or preconfigured; or, the second device determines the first reconstruction mode based on a first data type and a first mapping relationship; or, the second device determines the first reconstruction mode based on the capability information of the terminal device, the capability information indicating one or more reconstruction modes supported by the terminal device, the one or more reconstruction modes including the first reconstruction mode.
  • the method 300 further includes: the second device acquiring the first mapping relationship.
  • the first mapping relationship may be predefined or preconfigured, wherein predefinition includes pre-defined features, such as protocol definitions. Preconfiguration can be achieved by pre-saving corresponding codes, tables, or other means that can be used to indicate relevant information in the first device and/or the second device; or, the second device acquires indication information #d (i.e., an example of first indication information), which indicates the first mapping relationship, i.e., the first mapping relationship may be indicated or configured by the terminal device via signaling.
  • the first device sends indication information #d to the second device, which indicates the first mapping relationship; correspondingly, the second device receives indication information #d and determines the first mapping relationship based on the indication information #d.
  • the method 300 further includes: the second device acquiring the first data type.
  • the first data type may be predefined or preconfigured; or, the second device acquires indication information #e, which indicates the first data type, i.e., the first data type may be indicated or configured by the terminal device via signaling.
  • the first device sends indication information #e to the second device, which indicates the first data type, and correspondingly, the second device receives the indication information #e and determines the first data type based on the indication information #e.
  • the method 300 further includes: the first device sending indication information #g (i.e., an example of third indication information) to the second device.
  • This indication information #g indicates the capability information of the first device, or one or more reassembly modes supported by the first device, wherein the one or more reassembly modes include the first reassembly mode.
  • the second device can determine one or more reassembly modes supported by the terminal device based on the received indication information #g, and then determine the first reassembly mode by combining the first data type and the first mapping relationship.
  • the corresponding multiple reassembly modes include: a reassembly mode based on sequential splitting and a reassembly mode based on semantic information.
  • the reassembly mode supported by the terminal side is the reassembly mode based on sequential splitting
  • the first reassembly mode ultimately determined by the second device is the reassembly mode based on sequential splitting.
  • the indication information #b, #a, or #c may be carried in the first signaling or the first resource.
  • the first signaling includes radio resource control (RRC) signaling, downlink control information (DCI) signaling, or medium access control-control element (MAC CE).
  • RRC radio resource control
  • DCI downlink control information
  • MAC CE medium access control-control element
  • the first resource includes physical downlink control channel (PDCCH) or physical downlink shared channel (PDSCH).
  • the indication information #d, #e, #f, or #g may be carried in a second signaling or a second resource.
  • the second signaling includes RRC signaling, MAC CE signaling, or uplink control information (UCI) signaling.
  • the second resource includes a physical uplink control channel (PUCCH) or a physical uplink shared channel (PUSCH).
  • this application does not limit the specific implementation of the first device determining the first recombination mode, as long as the first device and the second device align the first recombination mode.
  • the first device after determining the first recombination mode, can indicate the selected first recombination mode to the second device (or the first device), as shown in the following specific example.
  • a recombination mode is introduced to indicate different recombination patterns.
  • the recombination mode can be equal to 1, 2, ..., P, each indicating one of the P recombination patterns.
  • the corresponding reassembly mode can be determined when the first data type is determined, thus determining the reassembly mode supported by the first data type (i.e., the first reassembly mode); when the first data type supports multiple reassembly modes, the first device or the second device can configure or indicate the reassembly mode number through signaling, thereby determining the first reassembly mode corresponding to the first data type.
  • the first or second device can add a reorganizeMode field to the RRC signaling, which can carry an index or number of the reorganize mode to indicate the corresponding reorganization mode.
  • reorganizeMode can either directly indicate the reorganization mode number or indicate the element index in a specific mode set.
  • the first data type has 5 reorganization modes
  • these 5 reorganization modes correspond one-to-one with 5 reorganization schemes, meaning the first data type supports 5 reorganization schemes.
  • the reorganization modes supported by a certain data type have reorganization mode numbers 1, 3, 4, and 5, optionally represented as a set ⁇ 1, 3, 4, 5 ⁇ , indicating that the data type supports reorganization modes including: reorganization mode #1, reorganization mode #3, reorganization mode #4, and reorganization mode #5.
  • a reorganizeMode value of 1 corresponds to reorganization mode #1
  • a reorganizeMode value of 2 corresponds to reorganization mode #3
  • a reorganizeMode value of 3 corresponds to reorganization mode #4
  • a reorganizeMode value of 4 corresponds to reorganization mode #5. That is, the value of reorganizeMode indicates which reorganization scheme in the corresponding set.
  • This first mapping relationship indicates the correspondence between K recombination patterns and L data types.
  • This first mapping relationship can exist in the form of a table, set, function, text, or string, such as for storage or transmission. This application does not limit its specific implementation. For ease of understanding, this application provides an example of the first mapping relationship in tabular form, as shown in Table 1 or Table 2 below. In the table, "1" indicates that the data type supports the corresponding recombination pattern, and "0" indicates that the data type does not support the corresponding recombination pattern.
  • the various data types corresponding to perceptual data mainly support: recombination patterns based on sequence splitting, recombination patterns based on bit position, recombination patterns based on spatial position, and recombination patterns based on semantic information.
  • the various data types corresponding to AI data mainly support: recombination patterns based on sequence splitting, recombination patterns based on bit position, and recombination patterns based on importance ranking.
  • the various data types corresponding to channel data mainly support recombination patterns based on sequence splitting.
  • the corresponding P reconstruction patterns include: reconstruction patterns based on sequence splitting, reconstruction patterns based on bit position, reconstruction patterns based on spatial position, and reconstruction patterns based on semantic information, where the first reconstruction pattern can be a reconstruction pattern based on sequence splitting;
  • the corresponding P reconstruction patterns include reconstruction patterns based on sequence splitting, that is, the first reconstruction pattern is a reconstruction pattern based on sequence splitting;
  • this first mapping relationship can be stored or indicated using a bitmap.
  • "1" and "0" represent bits 1 and 0, respectively, indicating whether the data type supports the reassembly mode.
  • environmental reflection points support reassembly modes based on sequence splitting, bit position, spatial location, and semantic information, with corresponding reassembly modes of 1, 2, 3, and 4.
  • AI training data supports reassembly modes based on sequence splitting, bit position, semantic information, and importance ranking, with corresponding reassembly modes of 1, 2, 4, and 5.
  • channel matrices support reassembly modes based on sequence splitting, bit position, and importance ranking, with corresponding reassembly modes of 1, 2, and 5, and so on.
  • this first mapping relationship can be stored or indicated by enumeration, where the correspondence between multiple data types and multiple recombination modes can be represented by a set of recombination modes.
  • the channel matrix/CSI matrix in the channel data is represented as ⁇ 1,2,5 ⁇ , indicating that the channel matrix/CSI matrix supports the reassembly modes based on sequential splitting, bit position, and importance order, corresponding to reassembly modes 1, 2, and 5; similarly, the channel indication information is represented as ⁇ 1 ⁇ .
  • the various data types corresponding to sensing data mainly support: recombination patterns based on sequence splitting, recombination patterns based on bit position, and recombination patterns based on spatial position.
  • the various data types corresponding to AI data mainly support: recombination patterns based on sequence splitting, recombination patterns based on bit position, and recombination patterns based on importance ranking.
  • the various data types corresponding to channel data mainly support recombination patterns based on sequence splitting.
  • the corresponding P recombination patterns include: recombination patterns based on sequential splitting, recombination patterns based on bit position, and recombination patterns based on spatial position, where the first recombination pattern can be a recombination pattern based on bit position;
  • the corresponding P recombination patterns include recombination patterns based on sequential splitting, recombination patterns based on bit position, and recombination patterns based on importance ranking, where the first recombination pattern can be a recombination pattern based on importance ranking;
  • this first mapping relationship can be stored or indicated using a bitmap.
  • "1" and “0" represent bits 1 and 0, respectively, indicating whether the data type supports the reassembly mode.
  • environmental reflection points support reassembly modes based on sequence splitting, bit position, and spatial position, with corresponding reassembly modes of 1, 2, and 3; similarly, AI training data supports reassembly modes based on sequence splitting, bit position, and importance ranking, with corresponding reassembly modes of 1, 2, and 5; and channel indication information supports reassembly modes based on sequence splitting, with a corresponding reassembly mode of 1, and so on.
  • this first mapping relationship can be stored or indicated by enumeration.
  • the correspondence between multiple data types and multiple recombination modes can be represented by a recombination mode set.
  • this application does not limit the number of correspondences (e.g., one row in the table) in any of the above tables; for example, one or more rows can be added or removed.
  • this application does not limit the number of original data, the data types corresponding to the original data, or the number of recombination patterns in the above tables, nor does it limit the correspondence between multiple data types and multiple recombination patterns; in other words, this application does not limit the number of rows or columns in the tables.
  • any of the above tables can be split into multiple independent tables, and this application does not limit the splitting method.
  • splitting at the granularity of original data can be independently divided into three new tables; or, the first mapping relationship between the sensing data and AI data in Table 2 can be independently divided into a new table, and the first mapping relationship corresponding to the channel data in Table 2 can be independently divided into a new table.
  • the mapping relationships can be split at the data type level. For instance, the first three rows of Table 1 could be separated into a new table, the middle four rows into a new table, and the last three rows into a new table; or, the first four rows of Table 2 could be separated into a new table, and the last two rows into a new table.
  • Table 1 can be split into multiple other tables for illustrative purposes.
  • the first mapping relationships between the first three recombination patterns and various data types in Table 1 and the first mapping relationships between the last two recombination patterns and various data types in Table 2 could each be separated into a new table, and so on.
  • These examples are provided for ease of understanding only and do not exclude other solutions. Any variations of the first mapping relationships fall within the scope of this application.
  • the first device groups the data of the first data type according to the first recombination mode to obtain the first data group.
  • this application does not limit the number of groups in the first data group or the size of the data contained in each group.
  • grouping methods please refer to the relevant descriptions in Figures 4 to 9 above. For the sake of brevity, they will not be described here.
  • the length of the first data type is N
  • the number of first data groups obtained by grouping the first data type according to the first recombination mode is M, where M is an integer greater than 1 and less than N.
  • M is an integer greater than 1 and less than N.
  • the first device can divide this data into 4 groups. These 4 groups can contain an equal amount (or equal length, or the same number) of data, that is, each group contains 2 data; or, these 4 groups can contain unequal amounts (or unequal lengths) of data, such as the 4 groups containing 1, 2, 2, and 3 data blocks respectively.
  • the first device outputs first data, which is obtained by processing the first data group.
  • the second device receives the first data.
  • the first data is obtained by processing first data groups.
  • the first type of data can undergo at least one of the following operations: data filtering, data transformation, data selection (or data discarding), quantization, entropy coding, channel mapping, channel coding, or modulation, to obtain the first data.
  • data filtering data transformation
  • data selection or data discarding
  • quantization quantization
  • entropy coding channel mapping
  • channel coding or modulation
  • the first device may send the first data to the second device directly or indirectly.
  • the first device sends the first data to the second device; or, for another example, the first device sends the first data to the second device through a third device, which may be a core network device, a network device, or a terminal device.
  • the second device processes the first data to obtain the first data group.
  • the second device demodulates and decodes the first data to obtain a binary code stream, and then performs operations such as channel demapping, entropy decoding, inverse quantization, data recovery, and data reassembly on the binary code stream to obtain the first data packet.
  • the implementation methods for demodulation, channel decoding, channel demapping, entropy decoding, inverse quantization, data recovery, inverse data transformation, zero-padding, or data reassembly can be found in existing descriptions and are not elaborated here for brevity.
  • the second device performs data recovery on the first data group according to the first reconstruction mode to obtain reconstructed data for the first data type.
  • the first recombination mode is clearly known to the second device and can be determined by predefinition or preconfiguration, or it can be indicated by the first device through signaling. This application does not limit this.
  • the first device can determine the grouping method of the data of the first data type according to the first recombination mode. Therefore, the second device can reconstruct the first data grouping according to the first recombination mode to obtain reconstructed data for the first data type.
  • the dimensions of the reconstructed data and the data before grouping in step S310 are consistent. However, due to calculation or processing errors, the values of the reconstructed first data type may deviate slightly from those of the data before grouping.
  • the above solution primarily targets grouping data of a first data type using a first reassembly mode to obtain a first data packet, thereby supporting the transmission of a specific service.
  • the first data type corresponds to P reassembly modes, it can adapt to different task or service requirements, meaning the first data type supports flexible switching between different reassembly modes, thus ensuring transmission performance.
  • the service type on the network side changes—for instance, the initial feedback of the first data is used for data fusion, while the subsequent feedback of the second data is used to perform a specific detection task—different reassembly modes can be used to group the data of the first data type to meet different task or service requirements and improve end-to-end performance.
  • method 300 further includes steps S301 to S305.
  • the first device acquires fourth indication information.
  • the second device sends fourth indication information to the first device.
  • the first device receives fourth indication information from the second device.
  • This fourth indication information is used to indicate a second recombination mode, which is used to indicate the grouping method for data of the first data type.
  • a second recombination mode which is used to indicate the grouping method for data of the first data type.
  • method 300 further includes: the first device obtaining fifth indication information, for example, the second device sending the fifth indication information to the first device; correspondingly, the first device receiving the fifth indication information from the second device, the fifth indication information being used to indicate a second mapping relationship, the second mapping relationship being used to indicate the correspondence between K’ recombination patterns and L’ data types, the K’ recombination patterns including the second recombination pattern, the L’ data types including the first data type, K’ being an integer greater than 1, and L’ being an integer greater than or equal to 1.
  • the second mapping relationship please refer to the relevant description of the first mapping relationship above; for brevity, it will not be explained here.
  • the second device can update the correspondence between data types and recombination modes to the first device by sending a fifth instruction message.
  • this could be an instruction to update from a first mapping relationship to a second mapping relationship, or an instruction to update from a third mapping relationship to a second mapping relationship, where the third mapping relationship can be obtained by updating the first mapping relationship.
  • the first device determines the second recombination mode by obtaining the fourth instruction message and the updated second mapping relationship.
  • the K' recombination patterns may be equal to or unequal to the P recombination patterns in step S310 above.
  • the K' recombination patterns may include the P recombination patterns, and there may also be some identical recombination patterns. This is not limited.
  • the K' recombination patterns may be equal to or unequal to the K recombination patterns in the first mapping relationship above, and there may also be some identical recombination patterns. This is not limited.
  • the L’ recombination patterns can be equal to or unequal to the L data types in the first mapping relationship described above, and there may also be partially identical recombination patterns; this is not limited.
  • method 300 further includes: the first device obtaining the sixth indication information, for example, the second device sending the sixth indication information to the first device, correspondingly, the first device receiving the sixth indication information from the second device, the sixth indication information being used to indicate Q recombination patterns corresponding to the first data type, the Q recombination patterns including the second recombination pattern, where Q is an integer greater than or equal to 2.
  • the Q recombination patterns may be equal to or unequal to the P recombination patterns in step S310 above.
  • the Q recombination patterns may include the P recombination patterns, and there may also be some identical recombination patterns. This is not limited.
  • the Q recombination patterns may be equal to or unequal to the K recombination patterns in the first mapping relationship above. There may also be some identical recombination patterns. This is not limited.
  • the second device can indicate Q recombination modes corresponding to the first data type to the first device by sending a sixth indication message. Further, the first device can determine the updated second recombination mode by obtaining the fourth indication message.
  • the sixth indication information may not be sent.
  • the first device groups the data of the first data type according to the second recombination mode to obtain the second data group.
  • the second data group is different from the first data group, which can be understood as follows: the number of groups contained in the first data group and the second data group are different. For example, the number of groups contained in the first data group and the second data group are M and M” respectively; and/or, the data content corresponding to the a groups contained in the first data group is different from the data content corresponding to the b groups contained in the second data group. It can be completely different or partially different. In other words, at least one of the a groups and the b groups has different data content.
  • the first data group and the second data group may contain the same number of groups. If the first data group and the second data group contain the same number of groups, for example, M, then at least one of the a groups in the first data group and the b groups in the second data group corresponds to different data content.
  • this application does not limit the number of second data packets or the data size contained in each packet.
  • Specific grouping methods can be referred to the relevant descriptions in Figures 4 to 9 above, which will not be elaborated here for brevity. It is understood that the first reassembly mode and the second reassembly mode are different, and the first data packets and the second data packets are also different, to adapt to different service transmission requirements.
  • the number of second data groups obtained by grouping the first data type according to the second recombination mode is M", where M is an integer greater than 1 and less than N.
  • M is an integer greater than 1 and less than N.
  • the first device can divide this data into 4 groups. These 4 groups can contain an equal amount (or equal length, or the same number) of data, that is, each group contains 2 data; or, these 4 groups can contain unequal amounts (or unequal lengths) of data, such as the 4 groups containing 1, 1, 2, and 4 data blocks respectively, without limitation.
  • the first device outputs second data, wherein the second data is obtained by processing the second data group.
  • the second device acquires the second data.
  • the second device processes the second data to obtain a second data group.
  • the second device performs data recovery on the second data group according to the second reconstruction mode to obtain reconstructed data for the first data type.
  • steps S304 and S305 please refer to the descriptions of steps S340 and S350 above. For the sake of brevity, these details will not be repeated here.
  • the first device can determine a first reassembly mode and then use this first reassembly mode to group data of the first data type.
  • This implementation design allows for multiple reassembly modes for the same data type, enabling the first device to flexibly select the appropriate reassembly mode to group data based on different business or task transmission requirements. This supports diverse data transmission needs and ensures data transmission performance.
  • the first device can switch reassembly modes based on specific compression and transmission requirements, thereby improving end-to-end performance.
  • the first device can process the grouped data, such as data selection, data discarding, or quantization, supporting dynamic matching of transmission resource constraints. This allows limited transmission resources to be used to send more important data packets, achieving better signal discarding effects and supporting flexible and efficient data compression and transmission at the physical layer, ensuring data transmission performance.
  • Figures 3 to 9 above mainly describe the data reassembly scheme, enabling flexible selection or switching of reassembly modes for specific data types to adapt to different business transmission needs. Furthermore, to support the standardized process of physical layer compression, meet the bitrate adjustment requirements in AI compression scenarios, and achieve dynamic matching of transmission resource constraints, compressed or discarded data after grouping can also be performed.
  • Figures 10 to 15 we will explain the data selection scheme (or data discarding scheme) and quantization scheme involved in the first device's statement in step S330 above that "the first data is obtained by processing the first data group.”
  • the first data is obtained by processing the first data groups, including: selecting (or discarding) data from M first data groups to obtain the first data. That is, the first device can filter or select M' first data groups from the M first data groups, or in other words, the first device discards M-M' first data groups from the M first data groups. In other words, the first data includes the M' first data groups.
  • the first device selects M' first data packets from the M first data packets based on the importance of the M first data packets.
  • the importance of the M’ first data packets is higher than or equal to the importance of the other M-M’ first data packets among the M first data packets.
  • the first device can select the important M’ first data packets from the M first data packets for transmission.
  • the importance of the first data packet can be determined by calculating the energy or L1 norm of the data contained in the first data packet, or other importance indicators.
  • the importance indicators can be predefined or preconfigured, or indicated by signaling.
  • Figure 10 is a schematic diagram of the grouping and filtering scheme provided in an embodiment of this application.
  • the first data type e.g., AI intermediate feature data
  • N the length of the first data type
  • the value of M' can be determined by resource constraints.
  • the value of M' can be predefined or preconfigured by the protocol, or the value of M' can be configured by the network device through signaling, or the value of M' can be indicated by the terminal device, etc.
  • the first device may also send indication information #A to the second device, which indicates M' first data packets. That is, the first device can indicate to the second device, by sending indication information #A, M' first data packets selected from the M first data packets.
  • the indication information #A may include a first bitmap, wherein the first bitmap indicates M' first data packets, and the size of the first bitmap is M bits; or, the indication information #A may include a packet sequence number indication corresponding to the M' first data packets, and the size of the packet sequence number indication is... Bits.
  • the first device prioritizes sending the data corresponding to packets 2 and 4.
  • Each bit can be carried in the indication information #A. For example, "00" indicates group 1, "01" indicates group 2, "10" indicates group 3, and "11" indicates group 4.
  • the specific implementation method by which the first device indicates the M' packets to the second device can be predefined by the protocol.
  • the protocol instructs the first device and the second device to always use the implementation method described in the first or second example to indicate the M' packets.
  • the first device can use a dynamic switching method to indicate the M' packets to the second device.
  • the header of the indication information #A may include a first bit, such as 1 bit, to indicate that the current first device uses the packet sequence number to indicate the M' first data packets (such as bit "1", in which case the indication information #A carries "1 10 11"), or, a first bit map to indicate the M' first data packets (such as bit "0", in which case the indication information #A carries "0 0101").
  • M' when M' is large, it is less costly to indicate the M' packets using the first bitmap method; conversely, it is more effective to indicate the M' packets using the packet sequence number method.
  • the first device when the first device needs to mark the priority order of the actually transmitted packets (such as in a scenario using unequal quantization), the first device can use the packet sequence number method.
  • the first device may implicitly indicate the use of a first bitmap or a packet sequence number indication method to indicate the M' first data packets. For example, this can be determined by the data type of the currently transmitted first data.
  • the first device uses a first bitmap to indicate the M' packets; when the data type of the first data is sensing data, imaging data, or AI data, it can be understood that the first device uses a packet sequence number indication method to indicate the M' packets.
  • the first data is obtained by processing the first data groups, including: quantizing the M first data groups to obtain the first data.
  • the quantization process can be performed after a data selection or data discarding scheme, that is, the first device quantizes the M' first data groups to obtain the first data.
  • the first device may employ scalar quantization (SQ) or vector quantization (VQ), wherein M1 of the M’ first data packets are quantized based on the quantization bit length Q1 to obtain the first data, where M1 is an integer greater than 0 and less than or equal to M’, and Q1 is an integer greater than 0.
  • SQL scalar quantization
  • VQ vector quantization
  • the SQ or VQ method is used, where M2 of the M’ first data packets are quantized based on the quantization bit length Q2 to obtain the first data, where M2 is an integer greater than 0 and less than or equal to M’, and Q2 is an integer greater than or equal to Q1.
  • the first device can use the SQ method to quantize M1 first data packets based on the quantization bit length Q1 to obtain a portion of the first data, and use the SQ method to quantize M2 first data packets based on the quantization bit length Q2 to obtain another portion of the first data; or, the first device can use the SQ method to quantize M1 first data packets based on the quantization bit length Q1 to obtain a portion of the first data, and use the VQ method to quantize M2 first data packets based on the quantization bit length Q2 to obtain a portion of the first data.
  • the first device may use VQ to quantize M1 first data packets based on quantization bit length Q1 to obtain a portion of the first data, and use VQ to quantize M2 first data packets based on quantization bit length Q2 to obtain another portion of the first data; or, the first device may use VQ to quantize M1 first data packets based on quantization bit length Q1 to obtain a portion of the first data, and use SQ to quantize M2 first data packets based on quantization bit length Q2 to obtain another portion of the first data.
  • the quantization bit length can be predefined or preconfigured, or it can be determined automatically based on resource constraints.
  • this application does not specifically limit the quantity and size relationship of M1 and M2.
  • the first device sends indication information #B to the second device, or the first device receives indication information #B from the second device; wherein, indication information #B is used to indicate Q, where Q represents the quantization bit length used to quantize M’ first data packets, and Q is an integer greater than 0.
  • indication information #B is used to indicate Q1, where Q1 represents the quantization bit length used to quantize M1 of the M’ first data packets, and/or, indication information #B is used to indicate Q2, where Q2 represents the quantization bit length used to quantize M2 of the M’ first data packets.
  • the quantization bit length can be replaced by the number of quantization bits, quantization length, quantization bits, or quantization resolution.
  • the quantization resolution can be expressed as the number of SQ quantization intervals or VQ quantization codewords. When the quantization bit length is longer, the corresponding number of quantization bits, quantization length, or quantization bits is also more, and the quantization resolution is higher.
  • Q2 when Q2 equals Q1, it means that the M' first data packets are quantized using the same quantization bit length, i.e., the M1 and M2 first data packets are quantized using the same quantization resolution, meaning the quantization precision of the M first data packets is the same.
  • Q2 is greater than Q1
  • Q2 the quantization bit length used by the M2 first data packets is greater than that used by the M1 first data packets. Understandably, the importance of the M2 first data packets is higher than that of the M1 first data packets.
  • dynamic non-uniform quantization of the M' first data packets can be implemented to improve the overall quantization precision of the reconstructed data.
  • the first device may also send indication information #C to the second device.
  • This indication information #C is used to indicate quantization configuration, which includes quantization mode and quantization parameters.
  • the quantization parameters include quantization interval information
  • the quantization parameters include VQ codebook matrix information
  • the quantization parameters include quantization interval information and VQ codebook matrix information.
  • this application does not limit the quantization method used for the M' first data groups, which can be SQ method and/or VQ method.
  • the quantization parameters can include quantization interval information and/or VQ codebook matrix information.
  • FIG 11 is a schematic diagram of the structure of the SQ and VQ codebooks provided in the embodiments of this application.
  • the SQ method is used to quantize M' first data groups.
  • the binary data stream obtained by quantization is represented as a group of Q bits (each group corresponds to one data), totaling Group, Bits.
  • the M' first data groups are quantized using the VQ method.
  • the binary data stream obtained through quantization i.e., the second data
  • QL bits each group corresponds to L data points, with a length of L
  • the first device can dynamically select appropriate quantization parameters, such as using mean-square error (MSE) or mean absolute error (MAE) as evaluation indicators, and select the quantization parameter with the smallest MSE or MAE for quantization.
  • MSE mean-square error
  • MAE mean absolute error
  • quantization method and/or quantization parameters described above may be predefined or preconfigured, or may be indicated by signaling; this application does not limit this.
  • the SQ method shown in Figure 11(a) and the VQ method shown in Figure 11(b) can be implemented independently or in combination.
  • the specific implementation method can be referred to the above description, which will not be repeated here.
  • Figure 12 is a schematic diagram of a packet filtering scheme under fixed resource constraints and quantization bit length provided in an embodiment of this application.
  • the scheduled resources and quantization bit length Q can be floating-point numbers (such as quarter-precision floating-point number fp8, half-precision floating-point number fp16, or single-precision floating-point number fp32, etc.) or fixed-point numbers, etc.
  • the first device can determine the number M’ of the first data packets to be sent in reality based on the scheduled transmission resources and quantization bit length.
  • N is the length of the data
  • M is the number of first data packets.
  • the initial calculation of the actual number of first data packets M' to be sent is performed, i.e., determining whether the data length N is less than or equal to K.
  • packet filtering information i.e., indication information #A
  • Figure 13 is a schematic diagram of a packet filtering and quantization scheme under fixed resource constraints provided in an embodiment of this application. As shown in Figure 13, assuming that the scheduled resources are fixed in advance, the first device can determine the number M' of the first data packets to be sent and the quantization bit length Q based on the scheduled transmission resources.
  • the first device can estimate the total number of bits B that can be transmitted based on the scheduled transmission resources (e.g., the number of time-frequency resources, channel code rate, or modulation order). It is assumed that a minimum quantization bit length Qmin is predefined or pre-configured, which is considered to be negligible due to the poor performance of AI encoding and decoding below Qmin .
  • N is the length of the data
  • M is the number of the first data packets.
  • the number of first data packets M' and the quantization bit length Q are initially calculated, i.e., it is determined whether the bit length N ⁇ Qmin is less than or equal to B when using the minimum quantization bit length Qmin .
  • the first device sends the packet filtering information (i.e., indication information #A) and indication information #B to the second device, indicating the M' first data packets and the quantization bit length Q.
  • the second device can correspondingly perform inverse quantization processing, data group recovery, and data reconstruction operations on the acquired first data to obtain reconstructed first data type data.
  • the specific implementation method is as follows.
  • the second device in step S340 above processes the first data to obtain the first data group, including: the second device performs inverse quantization on the first data to obtain M' reconstructed first data groups, wherein the M' reconstructed first data groups belong to the M reconstructed first data groups, and M' is a positive integer less than or equal to M.
  • the implementation method of inverse quantization of the first data to obtain the reconstructed M' feature groups in this application is not limited, and relevant descriptions of existing solutions can be referenced. For brevity, it will not be described further here.
  • the second device in step S340 above processes the first data to obtain the first data group, including: the second device performs zero-padding operation on M-M’ first data groups to obtain M reconstructed first data groups, where M is an integer greater than 1.
  • FIG 14 is a schematic diagram of the processing flow of the encoding unit provided in an embodiment of this application. As shown in Figure 14, the processing flow of the encoding unit includes, but is not limited to, the following steps. The following example illustrates the process using AI intermediate feature data as the first data type and a sequential split-based recombination mode as the first recombination mode.
  • the encoding unit After receiving feature data of length N, the encoding unit reassembles the feature data. That is, the encoding unit groups the features of length N, for example, according to the first reassembly mode, the N features are divided into M groups at equal intervals, namely data group 1, data group 2, data group 3, ..., data group M. Then the length of each data group is N/M, or in other words, each data group contains N/M feature data.
  • the M' groups belong to the M groups, and the importance of the M' groups is greater than that of the M-M' groups, so as to ensure data transmission accuracy while reducing transmission overhead.
  • the encoding unit sorts the M groups in step S1 according to their importance and selects M' groups from them.
  • the importance of each group can be determined by calculating the energy or L1 norm corresponding to that group.
  • the encoding unit can determine the quantization bit length Q according to the scheduled resources, and then quantize the M’ groups based on Q to obtain the first data, where Q is an integer greater than or equal to 1.
  • the signal to be transmitted includes a data stream (i.e., first data) and indication information #A, wherein indication information #A represents group filtering information (corresponding to S2, i.e., second indication information).
  • indication information #A represents group filtering information (corresponding to S2, i.e., second indication information).
  • the signal to be transmitted also includes recombination indication information #1 (corresponding to S1, see Figure 4 above) for indicating a first recombination mode.
  • the signal to be transmitted may also include codebook indication information (corresponding to S3, i.e., indication information #C) indicating quantization configuration.
  • FIG 15 is a schematic diagram of the processing flow of the decoding unit provided in an embodiment of this application. As shown in Figure 15, the processing flow of the decoding unit includes the following steps.
  • the first data type is AI intermediate feature data
  • the first recombination mode is a recombination mode based on sequential splitting, used as an example for illustration.
  • the decoding unit performs dequantization on the data stream according to the quantization configuration indicated by the instruction information #C, and obtains data features of M’ groups.
  • the decoding unit performs zero-padding on the M-M' groups filtered out by the first device, and then restores the order of the M groups according to the group filtering information carried in the instruction information #A to obtain the reconstructed data groups, such as reconstructed data group 1, reconstructed data group 2, reconstructed data group 3, ..., reconstructed data group M.
  • the decoding unit determines the grouping method for the AI intermediate feature data according to the recombination instruction information #1, that is, the recombination mode based on sequential splitting, and then performs a reverse mapping operation on the reconstructed data grouping (i.e., restores the order of feature data) to obtain the reconstructed AI intermediate feature data (e.g., a vector of length N).
  • the reconstructed data grouping i.e., restores the order of feature data
  • the first device groups the data of the first data type into M first data groups according to the determined first recombination mode. Then, it selects M' first data groups from these M groups (or discards M-M' first data groups), performs quantization processing, and outputs the first data.
  • M' first data groups from these M groups (or discards M-M' first data groups)
  • the first device can also process the M first data groups, such as performing data selection and quantization operations, to adapt to different AI models and/or transmission scenarios.
  • the examples are mainly based on devices in existing network architectures (e.g., a first device or a second device, etc.), and it should be understood that the specific form of the device is not limited in the embodiments of this application. For example, any device that can achieve the same function in the future is applicable to the embodiments of this application.
  • the methods and operations implemented by the device can also be implemented by components of the device (e.g., a chip or circuit).
  • the data transmission method provided in the embodiments of this application has been described in detail above with reference to Figures 1 to 16.
  • the above data transmission method is mainly described from the perspective of the interaction between the first device and the second device. It is understood that, in order to achieve the above functions, the first device and the second device include hardware structures and/or software modules corresponding to the execution of each function.
  • this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is implemented in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
  • This application embodiment can divide the communication device into functional modules according to the above method example.
  • each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module.
  • the integrated modules can be implemented in hardware, software, or a combination of both.
  • the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.
  • the following description uses the division of functional modules according to each function as an example.
  • Figure 16 is an exemplary block diagram of a communication device provided in an embodiment of this application.
  • the communication device 1000 may include a chip system 1100, a memory 1200, a bus 1300, a power management module 1400, or a transceiver 1500, etc.
  • the chip system 1100 can be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed through integrated logic circuits in the hardware of the chip system 1100 or through software instructions.
  • the chip system 1100 may include circuitry or chips responsible for signal processing (such as a modem chip, also known as a baseband chip, or a system-on-a-chip (SoC) or SIP chip containing a modem core).
  • a modem chip also known as a baseband chip, or a system-on-a-chip (SoC) or SIP chip containing a modem core.
  • SoC system-on-a-chip
  • the chip system 1100 may also include a memory (such as a cache) for storing instructions and data.
  • a memory such as a cache
  • the memory in the chip system 1100 is a cache memory. This memory can store instructions or data that the chip system 1100 has just used or that are used repeatedly. If the chip system 1100 needs to use the instruction or data again, it can directly retrieve it from the memory. This avoids repeated accesses, reduces the waiting time of the chip system 1100, and thus improves the efficiency of the system.
  • the chip system 1100 may include one or more interfaces.
  • Interfaces may include an inter-integrated circuit (I2C) interface, an inter-integrated circuitsound (I2S) interface, a pulse code modulation (PCM) interface, a universal asynchronous receiver/transmitter (UART) interface, a mobile industry processor interface (MIPI), a general-purpose input/output (GPIO) interface, a subscriber identity module (SIM) interface, and/or a universal serial bus (USB) interface, etc.
  • I2C inter-integrated circuit
  • I2S inter-integrated circuitsound
  • PCM pulse code modulation
  • UART universal asynchronous receiver/transmitter
  • MIPI mobile industry processor interface
  • GPIO general-purpose input/output
  • SIM subscriber identity module
  • USB universal serial bus
  • the memory 1200 may include random access memory (RAM) and read-only memory (ROM).
  • RAM random access memory
  • ROM read-only memory
  • the memory 1200 may store computer-readable, computer-executable code, including instructions that, when executed, cause the processor to perform the various functions described in this application.
  • the code may include instructions for implementing various aspects of the embodiments of this application, including instructions for supporting the generation or parsing of first information.
  • the code may be stored in a non-transitory computer-readable medium such as system memory or other types of memory.
  • the code may not be directly executable by the chip system 1100, but may enable a computer (e.g., at compile and execution time) to perform the functions described in this application.
  • memory 1200 may in particular contain a basic input/output (I/O) system that controls basic hardware or software operations, such as interaction with peripheral components or devices.
  • I/O basic input/output
  • the chip system 1100 executes various functional applications and data processing of the communication device 1000 by running instructions stored in the memory 1200. For instance, when the communication device 1000 transfers files with other devices (e.g., terminal devices, network devices, or core network devices), the chip system 1100 of the communication device 1000 can call the computer-executable program code stored in the memory 1200 to implement the data and/or signaling transmission methods provided in the embodiments of this application.
  • the chip system 1100 of the communication device 1000 can call the computer-executable program code stored in the memory 1200 to implement the data and/or signaling transmission methods provided in the embodiments of this application.
  • the memory 1200 can be integrated into the aforementioned chip system 1100, or it can be independent of the chip system 1100.
  • Bus 1300 can be USB, used to support communication between various parts of communication device 1000.
  • the power management module 1400 is used to receive charging input from the charger.
  • the power management module 1400 can also supply power to the communication device 1000 while charging it (e.g., the battery module of the communication device 1000).
  • the power management module 1400 can also supply power to other devices besides the communication device 1000.
  • Transceiver 1500 can communicate bidirectionally via one or more antennas, wired links, or wireless links.
  • transceiver 1500 can represent a wireless transceiver and can communicate bidirectionally with another wireless transceiver.
  • Transceiver 1500 may also include a modem for modulating packets and providing the modulated packets to the antenna for transmission, and for demodulating packets received from the antenna.
  • Transceiver 1500 may include a receiver and a transmitter, the receiver performing the function of receiving information and the transmitter performing the function of transmitting information.
  • a wireless device may include a single antenna. However, in other cases, the device may have more than one antenna, such as antenna 1 and antenna 2 shown in FIG. 16, which may be capable of simultaneously transmitting or receiving multiple wireless transmissions. Exemplarily, antenna 1 and antenna 2 are used to transmit and receive electromagnetic wave signals.
  • Each antenna in the communication device 1000 can be used to cover one or more communication frequency bands. Different antennas can also be multiplexed to improve antenna utilization. For example, antenna 1 can be multiplexed as a diversity antenna for a wireless local area network. In other embodiments, the antennas can be used in conjunction with a tuning switch.
  • the communication device 1000 can transfer files to other devices via wireless communication functions.
  • the communication device 1000 may correspond to the first device in the above method embodiment.
  • the device 1000 can implement the steps or processes corresponding to those performed by the first device in the above method embodiments, wherein the transceiver 1500 can be used to perform the transmission and reception related operations of the first device in the above method embodiments; and the chip system 1100 can be used to perform the processing related operations of the first device in the above method embodiments.
  • the communication device 1000 may correspond to the second device in the above method embodiment.
  • the device 1000 can implement the steps or processes corresponding to those performed by the second device in the above method embodiments, wherein the transceiver 1500 can be used to perform transmission and reception related operations of the second device in the above method embodiments; and the chip system 1100 can be used to perform processing related operations of the second device in the above method embodiments.
  • the communication device 1000 may include modules such as the short-range communication module 1640, sensor 1610, display 1620, or camera 1630 as shown in Figure 16.
  • the short-range communication module 1640 may include a wireless network (WI-FI, or WIFI), or a module that supports short-range communication such as Bluetooth.
  • WI-FI wireless network
  • Bluetooth a module that supports short-range communication
  • Sensor 1610 may include pressure sensors, gyroscope sensors, barometric pressure sensors, magnetic sensors, accelerometers, distance sensors, proximity sensors, fingerprint sensors, temperature sensors, touch sensors, ambient light sensors, bone conduction sensors, etc.
  • Display 1620 is used to display images, videos, etc.
  • the display includes a display panel.
  • the display panel can be a liquid crystal display (LCD), an organic light-emitting diode (OLED), an active-matrix organic light-emitting diode (AMOLED), a flexible light-emitting diode (FLED), a miniature LED, a microLED, a quantum dot light-emitting diode (QLED), etc.
  • the display can be used to display the interface required by the communication device 1000.
  • the communication device 1000 implements display functions through a graphics processing unit (GPU), a display, and an application processor.
  • the GPU is a microprocessor for image processing, connected to the display and the application processor.
  • the GPU is used to perform mathematical and geometric calculations and for graphics rendering.
  • the chip system 1100 may include one or more GPUs that execute program instructions to generate or change display information.
  • the camera 1630 is used to acquire images, videos, etc.
  • the structure shown in Figure 16 does not constitute a specific limitation on the communication device 1000, and the specific structure of the terminal device and/or network device can be referred to Figure 16.
  • the communication device 1000 may also include more or fewer components than shown in Figure 16, or combine some components, or split some components, or have different component arrangements, etc.
  • some components shown in Figure 16 may be implemented in hardware, software, or a combination of software and hardware, and the terminal device and/or network device may add or reduce components based on the structure given in Figure 16.
  • FIG 17 is a schematic block diagram of a communication device provided in an embodiment of this application.
  • the communication device 2000 may include a baseband unit 2100, which can communicate with external devices via a cellular RF transceiver 2200 (e.g., if the communication device 2000 is a terminal device, the baseband unit 2100 can communicate with network devices via the cellular RF transceiver 2200; or, if the communication device 200 is a network device, the baseband unit 2100 can communicate with terminal devices and/or core network devices via the cellular RF transceiver 2200).
  • a baseband unit 2100 can communicate with external devices via a cellular RF transceiver 2200 (e.g., if the communication device 2000 is a terminal device, the baseband unit 2100 can communicate with network devices via the cellular RF transceiver 2200; or, if the communication device 200 is a network device, the baseband unit 2100 can communicate with terminal devices and/or core network devices via the cellular RF transceiver 2200
  • Baseband unit 2100 may include computer-readable medium/memory.
  • Baseband unit 2100 is responsible for general processing, including the execution of software stored on the computer-readable medium/memory. When executed by baseband unit 2100, the software causes baseband unit 2100 to perform the various functions described above.
  • the computer-readable medium/memory may also be used to store data manipulated by baseband unit 2100 during software execution.
  • the baseband unit 2100 further includes a receiving unit 2010, a management unit 2020, and a transmitting unit 2030.
  • the management unit 2020 includes one or more sub-units shown in FIG. 17 (e.g., an encoding unit and/or a decoding unit, wherein the encoding unit may refer to the relevant description in FIG. 12 above, for example, for performing steps S410 to S430 in the method embodiment 400 above, and the decoding unit may refer to the relevant description in FIG. 13 above, for example, for performing steps S430 to S470 in the method embodiment 400 above).
  • the units within the management unit 2020 may be stored in a computer-readable medium/memory and/or configured as hardware within the baseband unit 2100.
  • the receiving unit 2010 and the transmitting unit 2030 may be referred to as transceiver units.
  • the receiving unit 2010 is used to execute the receiving step of the first device
  • the sending unit 2030 is used to execute the sending step of the first device
  • the management unit 2020 is used to execute the processing step of the first device.
  • the receiving unit 2010 can be used to perform the step of receiving information in the method
  • the management unit 2020 can be used to perform the processing step in the method
  • the sending unit 2030 can be used to perform the step of sending information in the method.
  • the receiving unit 2010 is used to execute the receiving step of the second device
  • the sending unit 2030 is used to execute the sending step of the second device
  • the management unit 2020 is used to execute the processing step of the second device.
  • the receiving unit 2010 can be used to perform the step of receiving information in the method
  • the management unit 2020 can be used to perform the processing step in the method
  • the sending unit 2030 can be used to perform the step of sending information in the method.
  • Figure 18 is a schematic block diagram of a chip system 3000 provided in an embodiment of this application.
  • the chip system includes, but is not limited to: a modem chip, also known as a baseband chip, or a system-on-a-chip (SoC) chip or a system-in-package (SIP) chip containing a modem core.
  • SoC system-on-a-chip
  • SIP system-in-package
  • the chip system (or processing system) includes a processor 3100, a memory 3200, and an input/output interface 3300.
  • the processor 3100 can be a processing circuit in the chip system (including at least one processor, such as processor 1 and processor 2 as shown in FIG. 18).
  • the processor 3100 can be coupled to the memory 3200, and call the instructions in the memory 3200, so that the chip system can implement the methods and functions of the embodiments of this application.
  • the input/output interface 3300 can be an input/output circuit in the chip system, which outputs the information processed by the chip system, or inputs the data or signaling information to be processed into the chip system for processing.
  • the chip system is used to implement the operations performed by the first or second device in the various method embodiments described above.
  • the processor 3100 is used to implement the processing-related operations performed by the first device or the second device in the above method embodiments, as described in the foregoing embodiments;
  • the input/output interface 3300 is used to implement the sending and/or receiving-related operations performed by the first device or the second device in the above method embodiments, as described in the foregoing embodiments.
  • FIG 19 is a schematic block diagram of another chip system 4000 provided in an embodiment of this application.
  • the chip system (or processing system) includes an input/output interface 4100 and logic circuits 4200.
  • the input/output interface 4100 can be an input/output circuit in the chip system, outputting processed information or inputting data or signaling information to be processed into the chip system for processing; specific details can be found in the descriptions of the foregoing embodiments.
  • the logic circuits 4200 are used to execute the aforementioned data transmission method; specific details can also be found in the descriptions of the foregoing embodiments.
  • the chip system is used to implement the operations performed by the first or second device in the various method embodiments described above.
  • logic circuit 4200 is used to implement processing-related operations performed by the first or second device in the above method embodiments
  • input/output interface 4100 is used to implement sending and/or receiving-related operations performed by the first or second device in the above method embodiments.
  • This application also provides a computer-readable storage medium storing computer instructions for implementing the methods executed by the apparatus in the above-described method embodiments.
  • the computer program when executed by a computer, it enables the computer to implement the methods performed by the first or second device in the various embodiments of the above methods.
  • This application also provides a computer program product comprising instructions which, when executed by a computer, implement the methods performed by the first or second device in the above-described method embodiments.
  • This application also provides a communication system, including the aforementioned first device and/or second device.
  • the disclosed systems, apparatuses, and methods can be implemented in other ways.
  • the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods.
  • multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.
  • the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
  • the units described as separate components may or may not be physically separate.
  • the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
  • the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
  • the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.
  • This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.
  • the aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

一种数据传输方法、装置和系统，方法包括：确定第一重组模式，根据第一重组模式对第一数据类型的数据进行分组得到第一数据分组；输出第一数据，第一数据是由第一数据分组经过处理得到的。其中，第一重组模式用于指示针对第一数据类型的数据的分组方式，第一重组模式属于P个重组模式，P个重组模式与第一数据类型对应，P为大于1的整数。本申请技术方案，支持针对同一种数据类型下的多种重组模式的选择和切换，以期适应不同任务或业务需求，保证传输性能。

Description

数据传输方法、装置和系统
本申请要求在2024年06月29日提交中国国家知识产权局、申请号为202410874141.X，发明名称为“数据传输方法、装置和系统”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及通信技术领域，并且更具体地，涉及一种数据传输方法、装置和系统。
背景技术
随着无线通信应用场景的日益丰富，无线通信过程中会产生很多面向新场景的数据，例如感知/成像数据、信道数据、人工智能(artificial intelligence，AI)数据等，这些数据具有数据量大、冗余较多、存在时域、频域、或空域相关性等特性，可以利用这些特性对待传输的数据进行压缩，以降低传输开销。
示例性地，针对原生数据，例如信道状态信息(channel state information，CSI)压缩、感知数据压缩、或AI数据(如训练、模型参数等)等，可以进行数据重组、数据过滤、数据变换、数据选择、量化、熵编码、或信道映射等操作，得到待发送的码流，然后进行信道编码、调制等处理后通过无线空口发送。然而，当前的数据重组方案可能无法适应不同任务或业务需求，也就无法保证传输性能。
发明内容
本申请提供一种数据传输方法、装置和系统，以期适应不同任务或业务需求，保证传输性能。
第一方面，提供了一种数据传输方法。该方法可以由第一装置(或编码设备)执行，在不特殊说明的情况下，本申请中的“第一装置”既可以指通信设备本身(例如，终端设备或网络设备)，也可以是通信设备中的组件(例如，通信模组、处理器、电路、芯片、或芯片系统等)，或者也可以是能实现全部或部分通信设备功能的逻辑模块或软件。
该方法包括：确定第一重组模式，第一重组模式用于指示针对第一数据类型的数据的分组方式，第一重组模式属于P个重组模式，P个重组模式与第一数据类型对应，P为大于1的整数；根据第一重组模式对第一数据类型的数据进行分组，得到第一数据分组；输出第一数据，第一数据是由第一数据分组经过处理得到的。
基于上述方案，第一装置可以确定第一重组模式，然后使用该第一重组模式对第一数据类型的数据进行分组。该实现方式，针对同一数据类型(例如第一数据类型)设计多种重组模式(例如P种重组模式)，使得第一装置可以根据不同的业务或任务传输需求，灵活选择满足传输需求的第一重组模式进行数据分组，即支持多样化的数据传输需求，保证数据传输性能。同时，第一装置还可以结合具体压缩及传输需求进行重组模式的切换，从而提升端到端性能。另外，第一装置还可以将分组后的数据进行处理，例如数据选择、压缩、或量化等，支持动态匹配传输资源约束，可以将有限的传输资源用于发送更重要的分组数据，达到较好的信号丢弃效果，用于支持在物理层实现灵活、高效地数据压缩和传输，保证数据传输性能。
可选地，第一数据是由第一数据分组经过处理得到的，可以理解为：第一类型的数据经过第一重组模式进行分组后，可以再经过数据过滤、数据变换、数据选择(或者说数据丢弃)、量化、熵编码、信道映射、信道编码、或调制中的至少一项操作，得到第一数据。
示例性地，第一数据类型包括以下一项：环境反射点、环境面片、环境重构地图、射频地图、AI训练数据、AI推理数据、AI模型参数、AI梯度参数、AI中间特征数据、AI性能数据、信道矩阵、信道状态信息(channel state information，CSI)矩阵、或信道指示信息。
示例性地，重组模式包括以下一项或多项：基于顺序拆分的重组模式；基于比特位置的重组模式；基于空间位置的重组模式；基于语义信息的重组模式；或者，基于重要性排序的重组模式。
其中，基于顺序拆分的重组模式是指：根据数据存储或者记录的顺序进行划分得到不同的分组，能够实现复杂度低，可以支持多种数据类型。基于比特位置的重组模式是指：根据比特位的高低进行划分得到不同的分组，便于与比特分层、或增量传输等结合设计。基于空间位置的重组模式是指：根据数据所在的位置信息进行划分得到不同的分组，适用于具有地理位置信息的感知数据，可以结合地理位置信息对数据进行联合处理。基于语义信息的重组模式是指：根据数据对应的语义类别进行划分得到不同的分组，适用于划分了语义特征的数据(如进行了语义分割的环境反射点、环境面片、环境重构地图等数据，这些数据对应的语义信息可以指示不同的对象、前景/背景等)，可以结合数据语义特征对数据进行分组处理。基于重要性排序的重组模式是指：相对更重要的码流使用可靠度较高的信道编码、或调制方式(信道码率、调制阶数更低)；反之，相对不重要的码流使用可靠度较低的信道编码、或调制方式(信道码率、或调制阶数更高)，可以结合数据具体应用场景，获得不同重要性的数据码流，支持非均等保护(unequal error protection，UEP)传输设计。
在一种可能的设计中，当第一数据类型的数据为感知数据时，若第一数据类型为环境反射点，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式。该示例中，可以根据环境反射点的具体应用场景，确定对应的第一重组模式。例如，环境反射点数据被用于进行目标、物体识别任务时，可以选择基于空间位置或语义信息的重组模式，以实现更好的任务精度；再例如，环境反射点数据被用于进行环境重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；又例如，环境反射点数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。其中，增量传输是指获得多份比特层信息，终端设备可以每次发一个或多个比特层，当调度的传输资源发生变化时，可以优先发送高位比特层(重要性更高)信息。
在一种可能的设计中，当第一数据类型的数据为感知数据时，若第一数据类型为环境面片，则第一重组模式包括以下一项：基于顺序拆分的重组模式，或者，基于语义信息的重组模式。该示例中，可以根据环境面片的具体应用场景，确定第一重组模式。例如，环境面片数据被用于进行目标、物体识别任务时，可以选择基于语义信息的重组模式，以实现更好的任务精度；再例如，环境面片数据被用于进行环境重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能。
在一种可能的设计中，当第一数据类型的数据为感知数据时，若第一数据类型为环境重构地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式。该示例中，可以根据环境重构地图的具体应用场景，确定第一重组模式。例如，环境重构地图数据被用于进行目标、物体识别任务时，可以选择基于空间位置或语义信息的重组模式，以实现更好的任务精度；再例如，环境重构地图数据被用于进行环境重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；环境重构地图数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
在一种可能的设计中，当第一数据类型的数据为感知数据时，若第一数据类型为射频地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于重要性排序的重组模式。该示例中，可以根据射频地图的具体应用场景，确定第一重组模式。例如，射频地图数据被用于进行目标、物体识别或波束方向获取等任务时，可以选择基于空间位置或语义信息的重组模式，以实现更好的任务精度；再例如，射频地图数据被用于进行环境重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；又例如，射频地图数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
在一种可能的设计中，当第一数据类型的数据为AI数据时，若第一数据类型为AI训练数据或AI推理数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于语义信息的重组模式，或者，基于重要性排序的重组模式。该示例中，可以根据AI训练数据或AI推理数据的具体应用场景，确定第一重组模式。例如，AI训练数据或AI推理数据被用于进行目标、物体识别或波束方向获取等任务时，可以选择基于语义信息的重组模式，以实现更好的任务精度；再例如，AI训练数据或AI推理数据被用于进行数据重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；又例如，AI训练数据或AI推理数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
在一种可能的设计中，当第一数据类型的数据为AI数据时，若第一数据类型为AI模型参数或AI梯度参数，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式。该示例中，可以根据AI模型参数或AI梯度参数的具体应用场景，确定第一重组模式。例如，UEP传输场景下，需要获得AI模型参数或AI梯度参数不同重要性数据码流时，可以选择基于重要性排序的重组模式，以实现更好的传输性能；再例如，非UEP传输场景下，AI模型参数或AI梯度参数对应数据码流采取相同的编码调制配置时，可以选择基于顺序拆分的重组模式，以实现复杂度较低；又例如，AI模型参数或AI梯度参数被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
在一种可能的设计中，当第一数据类型的数据为AI数据时，若第一数据类型为AI中间特征数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于语义信息的重组模式。该示例中，可以根据AI中间特征数据的具体应用场景，确定第一重组模式。例如，AI中间特征数据被用于进行目标、物体识别或波束方向获取等任务时，可以选择基于语义信息的重组模式，以实现更好的任务精度；再例如，AI中间特征数据被用于做数据重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；又例如，AI中间特征数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
在一种可能的设计中，当第一数据类型的数据为AI数据时，若第一数据类型为AI性能数据，则第一重组模式包括基于顺序拆分的重组模式。
在一种可能的设计中，当第一数据类型的数据为信道数据时，若第一数据类型为信道矩阵或CSI矩阵，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式。该示例中，可以根据信道矩阵或CSI矩阵的具体应用场景，确定第一重组模式。例如，UEP传输场景下，需要获得信道矩阵或CSI矩阵不同重要性数据码流时，可以选择基于重要性排序的重组模式，以实现更好的传输性能；再例如，非UEP传输场景下，信道矩阵或CSI矩阵对应数据码流采取相同的编码调制配置时，以选择基于顺序拆分的重组模式，以实现复杂度较低；又例如，信道矩阵或CSI矩阵被用于进行增量传输时，以选择基于比特位置的重组模式，可以灵活匹配不同传输资源。
在一种可能的设计中，当第一数据类型的数据为信道数据时，若第一数据类型为信道指示信息，则第一重组模式包括基于顺序拆分的重组模式。
在一种可能的设计中，获取或输出第一指示信息，第一指示信息用于指示第一映射关系，第一映射关系用于指示K个重组模式与L个数据类型之间的对应关系，K个重组模式包括第一重组模式，L个数据类型包括第一数据类型，K个重组模式包含P个重组模式，P为小于或等于K的整数，L为大于或等于1的整数。
可选地，该第一映射关系可以是预定义或预配置的。
基于上述方案，第一装置和第二装置可以确定K个重组模式与L个数据类型之间的对应关系，进而可以确定第一数据类型对应的一个或多个重组模式，进而选择合适的第一重组模式对数据进行分组，实现业务的传输需求，满足不同的业务传输场景。
在一种可能的设计中，确定第一重组模式，包括：根据第一数据类型和第一映射关系，确定第一重组模式。
可选地，该第一重组模式可以是预定义或预配置的。
基于上述方案，第一装置可以根据第一数据类型和第一映射关系确定第一重组模式，进而根据第一重组模式对数据进行分组，实现业务的传输需求，满足不同的业务传输场景。可以理解的，第一装置和第二装置同时支持该第一重组模式，有助于保证数据传输性能。
在一种可能的设计中，该方法包括：获取或输出第二指示信息，第二指示信息用于指示第一重组模式。
基于上述方案，第一装置和第二装置可以通过传输第二指示信息，确定第一重组模式，进而第一装置可以根据第一重组模式对第一数据类型的数据进行分组得到第一数据分组，并且输出第一数据，以适应业务传输需求，保证传输性能。
在一种可能的设计中，在获取第二指示信息之前，方法还包括：发送第三指示信息，第三指示信息用于指示第一装置的能力信息，或者，第一装置支持的一个或多个重组模式，其中，一个或多个重组模式包括第一重组模式。
基于上述方案，通过第三指示信息确定第一装置支持的一个或多个重组模式，进而有利于后续第一装置或第二装置确定第一重组模式，例如综合考虑第一装置支持的一个或多个重组模式，第二装置支持的一个或多个重组模式，以及第一数据类型对应的一个或多个重组模式，最终确定合适的第一重组模式，有利于保障传输性能，提升用户体验。
在一种可能的设计中，该方法还包括：获取第四指示信息，第四指示信息用于指示第二重组模式，第二重组模式用于指示针对第一数据类型的数据的分组方式；根据第二重组模式对第一数据类型的数据进行分组，得到第二数据分组，第二数据分组与第一数据分组不同；输出第二数据，第二数据是由第二数据分组经过处理得到的。
基于上述方案，考虑到第一数据类型对应P个重组模式，可以适应不同任务或业务需求，即第一数据类型支持不同的重组模式的灵活切换，从而保证传输性能。例如，针对同一数据类型，若网络侧的业务类型发生了变化，例如初始时反馈的第一数据用于数据融合，后续待反馈的第二数据用于执行特定检测任务，则可以通过发送第四指示信息指示采用不同的重组模式对第一数据类型的数据进行分组，以满足不同任务或业务需求，提升端到端性能。
在一种可能的设计中，在获取第四指示信息之前，方法还包括：获取第五指示信息，第五指示信息用于指示第二映射关系，第二映射关系用于指示K’个重组模式与L’个数据类型之间的对应关系，K’个重组模式包括第二重组模式，L’个数据类型包括第一数据类型，K’为大于1的整数，L’为大于或等于1的整数。
该实现方式中，第二装置可以通过发送第五指示信息，向第一装置更新数据类型和重组模式之间的对应关系，例如可以是指示从第一映射关系更新为第二映射关系，或者，也可以是指示从第三映射关系更新为第二映射关系，该第三映射关系可以是更新第一映射关系得到的。进一步地，第一装置通过获取第四指示信息和更新后的第二映射关系确定第二重组模式，支持针对同一种数据类型下的多种重组模式的选择和切换，以期适应不同任务或业务需求，保证传输性能。
在一种可能的设计中，在获取第四指示信息之前，方法还包括：获取第六指示信息，第六指示信息用于指示第一数据类型对应的Q个重组模式，Q个重组模式包括第二重组模式，Q为大于或等于2的整数。
可选地，如果该第一数据类型对应的Q个重组模式等于上述P个重组模式，则该第六指示信息可以不发送。
该实现方式中，第二装置可以通过发送第六指示信息，向第一装置指示第一数据类型对应的Q个重组模式，进一步地，第一装置通过获取第四指示信息可以确定更新后的第二重组模式，支持针对同一种数据类型下的多种重组模式的选择和切换，以期适应不同任务或业务需求，保证传输性能。例如，Q个重组模式可以以集合的形式存在，假设Q个重组模式＝4，对应重组mode＝{1,2,3,4}，假设第一重组模式对应重组mode＝1，则该第六指示信息可以包括重组mode＝3，表示第二重组模式对应重组mode 3，例如基于空间位置的重组模式。
在一种可能的设计中，第一装置获取或输出重组指示信息，该重组指示信息指示所述第一数据分组的个数M，和/或，M个第一数据分组中的每个第一数据分组的大小(或者说长度)。
该实现方式中，通过获取或输出重组指示信息，使得第二装置可以确定第一数据分组的个数M，和/或，M个第一数据分组中的每个第一数据分组的大小(或者说长度)，便于后续的数据分组重构，保证数据端到端的传输性能。
第二方面，提供了一种数据传输方法。该方法可以由第二装置(或解码设备)执行，在不特殊说明的情况下，本申请中的“第二装置”既可以指通信设备本身(例如，网络设备或终端设备)，也可以是通信设备中的组件(例如，通信模组、处理器、电路、芯片、或芯片系统等)，或者也可以是能实现全部或部分通信设备功能的逻辑模块或软件。
该方法包括：获取第一数据；对第一数据进行处理得到第一数据分组根据第一重组模式对第一数据分组进行数据恢复，得到针对第一数据类型的重构数据；其中，第一重组模式用于指示针对第一数据类型的数据的分组方式，第一重组模式属于P个重组模式，P个重组模式与第一数据类型对应，P为大于1的整数。
在一种可能的设计中，获取或输出第一指示信息，第一指示信息用于指示第一映射关系，第一映射关系用于指示K个重组模式与L个数据类型之间的对应关系，K个重组模式包括第一重组模式，L个数据类型包括第一数据类型，K个重组模式包含P个重组模式，P为小于或等于K的整数，L为大于或等于1的整数。
在一种可能的设计中，第一重组模式是根据第一数据类型和第一映射关系确定的。
在一种可能的设计中，方法还包括：获取或输出第二指示信息，第二指示信息用于指示第一重组模式。
在一种可能的设计中，在获取第二指示信息之前，方法还包括：接收第三指示信息，第三指示信息用于指示第一装置的能力信息，或者，第一装置支持的一个或多个重组模式，其中，一个或多个重组模式包括第一重组模式。
在一种可能的设计中，重组模式包括以下一项或多项：基于顺序拆分的重组模式基于比特位置的重组模式；基于空间位置的重组模式；基于语义信息的重组模式；或者，基于重要性排序的重组模式。
在一种可能的设计中，第一数据类型包括以下一项：环境反射点、环境面片、环境重构地图、射频地图、人工智能AI训练数据、AI推理数据、AI模型参数、AI梯度参数、AI中间特征数据、AI性能数据、信道矩阵、信道状态信息CSI矩阵、或信道指示信息。
在一种可能的设计中，当第一数据类型的数据为感知数据时，若第一数据类型为环境反射点，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式若第一数据类型为环境面片，则第一重组模式包括以下一项：基于顺序拆分的重组模式，或者，基于语义信息的重组模式；若第一数据类型为环境重构地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式若第一数据类型为射频地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于重要性排序的重组模式。
在一种可能的设计中，当第一数据类型的数据为AI数据时，若第一数据类型为AI训练数据或AI推理数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于语义信息的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为AI模型参数或AI梯度参数，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为AI中间特征数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于语义信息的重组模式；若第一数据类型为AI性能数据，则第一重组模式包括基于顺序拆分的重组模式。
在一种可能的设计中，当第一数据类型的数据为信道数据时，若第一数据类型为信道矩阵或CSI矩阵，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为信道指示信息，则第一重组模式包括基于顺序拆分的重组模式。
在一种可能的设计中，该方法还包括：获取第二数据；对第二数据进行处理得到第二数据分组；根据第二重组模式对第二数据分组进行数据恢复，得到针对第一数据类型的重构数据；其中，第二重组模式用于指示针对第一数据类型的数据的分组方式。
在一种可能的设计中，在获取第二数据之前，方法还包括：发送第五指示信息，第五指示信息用于指示第二映射关系，第二映射关系用于指示K’个重组模式与L’个数据类型之间的对应关系，K’个重组模式包括第二重组模式，L’个数据类型包括第一数据类型，K’为大于1的整数，L’为大于或等于1的整数。
在一种可能的设计中，在获取第二数据之前，方法还包括：发送第六指示信息，第六指示信息用于指示第一数据类型对应的Q个重组模式，Q个重组模式包括第二重组模式，Q为大于或等于2的整数。
在一种可能的设计中，第二装置获取或输出重组指示信息，该重组指示信息指示所述第一数据分组的个数M，和/或，M个第一数据分组中的每个第一数据分组的大小(或者说长度)。
该实现方式中，通过获取或输出重组指示信息，使得第一装置可以确定第一数据分组的个数M，和/或，M个第一数据分组中的每个第一数据分组的大小(或者说长度)，便于后续的数据分组，保证数据端到端的传输性能。
上述第二方面及第二方面的某些实现方式以及对应的有益效果可以参考第一方面相关的描述，在此不予赘述。
第三方面，提供了一种通信装置，通信装置具备实现上述第一方面的功能，比如，通信装置包括执行上述第一方面涉及操作所对应的模块或单元或手段(means)，模块或单元或手段具体可以通过软件实现，或者通过硬件实现，也可以通过软件结合硬件的方式实现。
示例性地，该通信装置可以为上述第一装置，例如可以是用于执行上述第一方面所描述的方法或操作或步骤或动作所一一对应的模块或单元(例如芯片，或者芯片系统，或者电路)。
一种可能的实现中，该通信装置包括：收发单元(或者通信模块)，以及与收发单元连接的处理单元(或者处理模块)。
示例性地，处理单元，用于确定第一重组模式，第一重组模式用于指示针对第一数据类型的数据的分组方式，第一重组模式属于P个重组模式，P个重组模式与第一数据类型对应，P为大于1的整数；处理单元，还用于根据第一重组模式对第一数据类型的数据进行分组，得到第一数据分组；收发单元，用于输出第一数据，第一数据是由第一数据分组经过处理得到的。
在一种可能的设计中，收发单元，还用于获取或输出第一指示信息，第一指示信息用于指示第一映射关系，第一映射关系用于指示K个重组模式与L个数据类型之间的对应关系，K个重组模式包括第一重组模式，L个数据类型包括第一数据类型，K个重组模式包含P个重组模式，P为小于或等于K的整数，L为大于或等于1的整数。
在一种可能的设计中，处理单元，还用于根据第一数据类型和第一映射关系，确定第一重组模式。
在一种可能的设计中，收发单元，还用于获取或输出第二指示信息，第二指示信息用于指示第一重组模式。
在一种可能的设计中，收发单元，还用于发送第三指示信息，第三指示信息用于指示第一装置的能力信息，或者，第一装置支持的一个或多个重组模式，其中，一个或多个重组模式包括第一重组模式。
在一种可能的设计中，重组模式包括以下一项或多项：基于顺序拆分的重组模式；基于比特位置的重组模式基于空间位置的重组模式；基于语义信息的重组模式；或者，基于重要性排序的重组模式。
在一种可能的设计中，第一数据类型包括以下一项：环境反射点、环境面片、环境重构地图、射频地图、人工智能AI训练数据、AI推理数据、AI模型参数、AI梯度参数、AI中间特征数据、AI性能数据、信道矩阵、信道状态信息CSI矩阵、或信道指示信息。
在一种可能的设计中，当第一数据类型的数据为感知数据时，若第一数据类型为环境反射点，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式；若第一数据类型为环境面片，则第一重组模式包括以下一项：基于顺序拆分的重组模式，或者，基于语义信息的重组模式；若第一数据类型为环境重构地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式；若第一数据类型为射频地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于重要性排序的重组模式。
在一种可能的设计中，当第一数据类型的数据为AI数据时，若第一数据类型为AI训练数据或AI推理数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于语义信息的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为AI模型参数或AI梯度参数，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为AI中间特征数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于语义信息的重组模式；若第一数据类型为AI性能数据，则第一重组模式包括基于顺序拆分的重组模式。
在一种可能的设计中，当第一数据类型的数据为信道数据时，若第一数据类型为信道矩阵或CSI矩阵，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为信道指示信息，则第一重组模式包括基于顺序拆分的重组模式。
在一种可能的设计中，收发单元，还用于获取第四指示信息，第四指示信息用于指示第二重组模式，第二重组模式用于指示针对第一数据类型的数据的分组方式；处理单元，还用于根据第二重组模式对第一数据类型的数据进行分组，得到第二数据分组，第二数据分组与第一数据分组不同；收发单元，还用于输出第二数据，第二数据是由第二数据分组经过处理得到的。
在一种可能的设计中，收发单元，还用于获取第五指示信息，第五指示信息用于指示第二映射关系，第二映射关系用于指示K’个重组模式与L’个数据类型之间的对应关系，K’个重组模式包括第二重组模式，L’个数据类型包括第一数据类型，K’为大于1的整数，L’为大于或等于1的整数。
在一种可能的设计中，收发单元，还用于获取第六指示信息，第六指示信息用于指示第一数据类型对应的Q个重组模式，Q个重组模式包括第二重组模式，Q为大于或等于2的整数。
第四方面，提供了一种通信装置，通信装置具备实现上述第二方面的功能，比如，通信装置包括执行上述第二方面涉及操作所对应的模块或单元或手段(means)，模块或单元或手段具体可以通过软件实现，或者通过硬件实现，也可以通过软件结合硬件的方式实现。
示例性地，该通信装置可以为上述第二装置，例如可以是用于执行上述第二方面所描述的方法或操作或步骤或动作所一一对应的模块或单元(例如芯片，或者芯片系统，或者电路)。
一种可能的实现中，该通信装置包括：收发单元(或者通信模块)，以及与收发单元连接的处理单元(或者处理模块)。
示例性地，处理单元，用于获取第一数据；处理单元，还用于对第一数据进行处理得到第一数据分组；处理单元，还用于根据第一重组模式对第一数据分组进行数据恢复，得到针对第一数据类型的重构数据；其中，第一重组模式用于指示针对第一数据类型的数据的分组方式，第一重组模式属于P个重组模式，P个重组模式与第一数据类型对应，P为大于1的整数。
在一种可能的设计中，收发单元，还用于获取或输出第一指示信息，第一指示信息用于指示第一映射关系，第一映射关系用于指示K个重组模式与L个数据类型之间的对应关系，K个重组模式包括第一重组模式，L个数据类型包括第一数据类型，K个重组模式包含P个重组模式，P为小于或等于K的整数，L为大于或等于1的整数。
在一种可能的设计中，处理单元，还用于根据第一数据类型和第一映射关系，确定第一重组模式。
在一种可能的设计中，收发单元，还用于获取或输出第二指示信息，第二指示信息用于指示第一重组模式。
在一种可能的设计中，收发单元，还用于接收第三指示信息，第三指示信息用于指示第一装置的能力信息，或者，第一装置支持的一个或多个重组模式，其中，一个或多个重组模式包括第一重组模式。
在一种可能的设计中，重组模式包括以下一项或多项：基于顺序拆分的重组模式；基于比特位置的重组模式；基于空间位置的重组模式；基于语义信息的重组模式；或者，基于重要性排序的重组模式。
在一种可能的设计中，收发单元，还用于第一数据类型包括以下一项：环境反射点、环境面片、环境重构地图、射频地图、人工智能AI训练数据、AI推理数据、AI模型参数、AI梯度参数、AI中间特征数据、AI性能数据、信道矩阵、信道状态信息CSI矩阵、或信道指示信息。
在一种可能的设计中，当第一数据类型的数据为感知数据时，若第一数据类型为环境反射点，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式；若第一数据类型为环境面片，则第一重组模式包括以下一项：基于顺序拆分的重组模式，或者，基于语义信息的重组模式；若第一数据类型为环境重构地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式；若第一数据类型为射频地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于重要性排序的重组模式。
在一种可能的设计中，当第一数据类型的数据为AI数据时，若第一数据类型为AI训练数据或AI推理数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于语义信息的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为AI模型参数或AI梯度参数，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为AI中间特征数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于语义信息的重组模式；若第一数据类型为AI性能数据，则第一重组模式包括基于顺序拆分的重组模式。
在一种可能的设计中，当第一数据类型的数据为信道数据时，若第一数据类型为信道矩阵或CSI矩阵，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；若第一数据类型为信道指示信息，则第一重组模式包括基于顺序拆分的重组模式。
在一种可能的设计中，收发单元，还用于获取第二数据；处理单元，还用于对第二数据进行处理得到第二数据分组；处理单元，还用于根据第二重组模式对第二数据分组进行数据恢复，得到针对第一数据类型的重构数据；其中，第二重组模式用于指示针对第一数据类型的数据的分组方式。
在一种可能的设计中，收发单元，还用于发送第五指示信息，第五指示信息用于指示第二映射关系，第二映射关系用于指示K’个重组模式与L’个数据类型之间的对应关系，K’个重组模式包括第二重组模式，L’个数据类型包括第一数据类型，K’为大于1的整数，L’为大于或等于1的整数。
在一种可能的设计中，收发单元，还用于发送第六指示信息，第六指示信息用于指示第一数据类型对应的Q个重组模式，Q个重组模式包括第二重组模式，Q为大于或等于2的整数。
第五方面，提供了一种通信装置。该通信装置可以是上述第一装置或第二装置。该通信装置包括收发器、处理器和存储器，该处理器用于控制收发器收发信号，该存储器用于存储计算机程序，该处理器用于从存储器中调用并运行该计算机程序，使得该通信装置执行上述第一方面或第二方面中任一种可能实现方式中的方法。
可选地，处理器为一个或多个，存储器为一个或多个。
可选地，存储器可以与处理器集成在一起，或者存储器与处理器分离设置。
可选地，该通信装置还包括发射机(发射器)和接收机(接收器)。
第六方面，提供了一种通信装置，所述通信装置包括一个或多个处理器，所述一个或多个处理器用于执行计算机程序或指令，当所述计算机程序或指令被执行时，使得所述通信装置实现上述第一方面或第二方面中任意可能的设计或实现方式中的方法。可选的，所述通信装置还包括存储器，所述存储器用于存储实现上述第一方面或第二方面涉及的功能的所述计算机程序或指令的部分或全部。
在一种可能的设计中，所述通信装置还可以包括接口电路，其中，处理器用于通过所述接口电路与其它装置或组件通信。
上述通信装置可以是终端，或终端中的通信模组，或终端中负责通信功能的芯片如Modem芯片(又称基带芯片)，或包含modem模块的片上系统(system on chip，SoC)芯片或系统级封装(system in apackage，SIP)芯片。
上述通信装置可以是网络设备，或者网络设备中的通信模组，或者网络设备中负责通信功能的电路或芯片，或者网络设备中能够调用并执行程序的功能模块。
第七方面，提供了一种通信系统。该通信系统包括第一装置和/或第二装置，其中，第一装置用于执行上述第一方面中任一种可能实现方式中的方法，第二装置用于执行上述第二方面中任一种可能实现方式中的方法。
示例性地，第一装置或第二装置可以是终端设备，或者，终端设备中的芯片或电路，或者，终端设备中能够调用并执行程序的功能模块；或者，第一装置或第二装置可以是网络设备，或者，网络设备中的芯片或电路，或者，网络设备中的集中式单元(central unit，CU)或分布式单元(distributed unit，DU)，或者，网络设备中能够调用并执行程序的功能模块。
第八方面，提供了一种计算机可读存储介质。该计算机可读存储介质存储有计算机程序代码或指令，以使得上述第一方面或第二方面中任一种可能实现方式中的方法被执行，例如当计算机读取并执行所述计算机程序代码或指令被运行时，使得上述第一方面或第二方面中任一种可能实现方式中的方法被实现。
第九方面，提供了一种计算机程序产品。该计算机程序产品包括：计算机程序代码或指令，以使得上述第一方面或第二方面中任一种可能实现方式中的方法被实现。例如，当计算机读取并执行所述计算机程序产品时，使得上述第一方面或第二方面中任一种可能实现方式中的方法被实现。
第十方面，提供了一种计算机程序。当计算机程序被运行时，使得上述第一方面或第二方面中任一种可能实现方式中的方法被实现。
应理解，上述第三方面至第十方面的有益效果可以参考上述第一方面或第二方面及其任一种可能的实现方式，在此不赘述。
附图说明
图1是适用于本申请实施例的一种通信系统的示意图；
图2是一种原生数据压缩及传输处理流程的示意图；
图3是本申请实施例提供的通信方法的流程示意图；
图4是本申请实施例提供的基于顺序拆分的重组模式对数据进行分组的结构示意图；
图5是本申请实施例提供的基于比特位置的重组模式对数据进行分组的结构示意图；
图6是本申请实施例提供的基于空间位置的重组模式对数据进行分组的结构示意图；
图7是本申请实施例提供的是基于语义信息的重组模式对数据进行分组的结构示意图；
图8和图9是本申请实施例提供的基于重要性排序的重组模式对数据进行分组的结构示意图；
图10是本申请实施例提供的分组筛选方案的示意图；
图11是本申请实施例提供的标量量化(scalar quantization，SQ)及矢量量化(vector quantization，VQ)码本的结构示意图；
图12是本申请实施例提供的固定资源约束和量化比特长度的情况下的分组筛选方案的示意图；
图13是本申请实施例提供的固定资源约束情况下的分组筛选和量化方案的示意图；
图14是本申请实施例提供的编码单元的处理流程的示意图；
图15是本申请实施例提供的解码单元的处理流程的示意图；
图16是本申请实施例提供的一种通信装置的示意性框图；
图17是本申请实施例提供的另一种通信装置的示意性框图；
图18是本申请实施例提供的一种芯片系统的示意性框图；
图19是本申请实施例提供的另一种芯片系统的示意性框图。
具体实施方式
下面将结合附图，对本申请中的技术方案进行描述。
为便于理解本申请实施例，做出以下几点说明。
(1)在本申请中，如果没有特殊说明以及逻辑冲突，不同的实施例之间的术语和/或描述具有一致性、且可以相互引用，不同的实施例中的技术特征根据其内在的逻辑关系可以组合形成新的实施例。
(2)在本申请中，“至少一个”是指一个或者多个，“多个”是指两个或两个以上。“和/或”，描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B的情况，其中A，B可以是单数或者复数。在本申请的文字描述中，字符“/”一般表示前后关联对象是一种“或”的关系。“以下至少一项(个)”或其类似表达，是指的这些项中的任意组合，包括单项(个)或复数项(个)的任意组合。例如，a、b和c中的至少一项(个)，可以表示：a，或，b，或，c，或，a和b，或，a和c，或，b和c，或，a、b和c。其中a、b和c分别可以是单个，也可以是多个。
(3)在本申请中，“第一”、“第二”以及各种数字编号指示为了描述方便进行的区分，并不用来限制本申请实施例的范围。例如，区分不同的消息等，而不是用于描述特定的顺序或先后次序。应理解，这样描述的对象在适当情况下可以互换，以便能够描述本申请的实施例以外的方案。
(4)在本申请中，“指示”或“用于指示”可以包括用于直接指示和用于间接指示。当描述某一指示信息用于指示A时，可以包括该指示信息直接指示A或间接指示A，而并不代表该指示信息中一定携带有A。
本申请实施例涉及的指示方式应理解为涵盖可以使得待指示方获知待指示信息的各种方法。待指示信息可以作为整体一起发送，也可以分成多个子信息分开发送，而且这些子信息的发送周期和/或发送时机可以相同，也可以不同，本申请对例如发送方法不作限定。
本申请实施例中的“指示信息”可以是显式指示，即通过信令直接指示，或者根据信令指示的参数，结合其他规则或结合其他参数或通过推导获得。也可以是隐式指示，即根据规则或关系，或根据其他参数，或推导获得。本申请对此不作具体限定。
(5)在本申请中，“协议”可以是指通信领域的标准协议，例如可以包括(5th generation，5G)协议、新空口(new radio，NR)协议、以及应用于未来的通信系统中的相关协议，本申请对此不作限定。“预定义”可以包括预先定义。例如，协议定义。“预配置”可以通过在设备中预先保存相应的代码、表格或其他可用于指示相关信息的方式来实现，本申请对于其实现方式不作限定。
(6)在本申请中，“消息”、“信息”、“信号”或“信息元素(information element，IE)”等可以替换使用，对于消息或信息的名称不作任何限定，能够实现相应功能即可。
“向XX(设备)发送信息”可以理解为该信息的目的端是该设备。可以包括直接或间接地向该设备发送信息。“从XX(设备)接收信息，或者接收来自XX(设备)的信息”可以理解为该信息的源端是该设备，可以包括直接或间接地从该设备接收信息。信息在信息发送的源端和目的端之间可能会被进行必要的处理，例如格式变化等，但目的端可以理解来自源端的有效信息。本申请中类似的表述可以做类似的理解，在此不作说明。
“通信”还可以描述为“通信”、“信息传输”、“数据处理”等。“传输”包括“发送”和“接收”。“传输”可以描述为“输出”。“发送”也可以理解为芯片接口的“输出”，“接收”也可以理解为芯片接口的“输入”。换言之，“发送”或“接收”可以是在设备之间进行的，例如，网络设备和终端设备之间通过空口分别进行发送或接收，“发送”或“接收”也可以是在设备内进行的，例如，通过总线、走线或接口在设备内的部件之间、模组之间、芯片之间、软件模块或者硬件模块之间发送或接收。
(7)在本申请中，“示例性地”、“比如”等词语用于表示例子、例证或说明。本申请中被描述为“示例”的任何实施例或设计方案不应被解释为比其它实施例或设计方案更优选或更具优势。确切而言，使用“示例”一词旨在以具体方式呈现概念。本申请实施例中，“的(of)”，“相应的(corresponding，relevant)”、“对应的(corresponding)”和“关联的(associate)”有时可以混用，应当指出的是，在不强调其区别时，其所要表达的含义是一致的。
本申请的技术方案可以应用于各种通信系统，例如5G或NR系统、长期演进(long term evolution，LTE)系统、LTE频分双工(frequency division duplex，FDD)系统、LTE时分双工(time division duplex，TDD)系统等。本申请的技术方案还可以应用于设备到设备(device to device，D2D)通信，车到万物(vehicle-to-everything，V2X)通信，机器到机器(machine to machine，M2M)通信，机器类型通信(machine type communication，MTC)，以及物联网(internet of things，IoT)通信系统或者其他通信系统。本申请的技术方案还可以应用于星间通信和卫星通信等非陆地通信网络(non-terrestrial network，NTN)系统。作为示例，卫星通信系统包括卫星基站和终端设备。该卫星基站为终端设备提供通信服务。卫星基站也可以与地面基站进行通信。卫星可作为基站，也可作为终端设备。其中，卫星可以是指无人机，热气球，低轨卫星，中轨卫星，高轨卫星等，也可以是指非地面基站或非地面设备等。
通信系统中的一个设备可以向另一个设备发送信号或从另一个设备接收信号。其中信号可以包括信息、信令或者数据等。其中，设备也可以被替换为实体、网络实体、网元、通信设备、通信模块、节点、通信节点等等，本申请中以设备为例进行描述。例如，通信系统可以包括至少一个终端设备和至少一个网络设备。网络设备可以向终端设备发送下行信号，和/或，终端设备可以向网络设备发送上行信号。
图1是适用于本申请实施例提供的一种通信系统的示意图。如图1所示，该通信系统10包括无线接入网(radio access network，RAN)100和核心网(core network，CN)200。RAN 100包括至少一个RAN节点(如图1中的110a和110b，统称为110)和至少一个终端(如图1中的120a-120j，统称为120)。RAN100中还可以包括其它RAN节点，例如，无线中继设备和/或无线回传设备(图1中未示出)等。终端120通过无线的方式与RAN节点110相连。RAN节点110通过无线或有线方式与核心网200连接。核心网200中的核心网设备与RAN 100中的RAN节点110可以分别是不同的物理设备，也可以是集成了核心网逻辑功能和无线接入网逻辑功能的同一个物理设备。
RAN 100可以为第三代合作伙伴计划(3rd generation partnership project，3GPP)相关的蜂窝系统，例如，4G移动通信系统、5G移动通信系统、或面向未来的演进系统。RAN 100还可以是开放式接入网(open RAN，O-RAN或ORAN)、云无线接入网络(cloud radio access network，CRAN)、或者无线保真(wireless fidelity，WiFi)系统。RAN100还可以是以上两种或两种以上系统融合的通信系统。
RAN节点110，有时也可以称为接入网设备，RAN实体或接入节点等，构成通信系统的一部分，用以帮助终端实现无线接入。通信系统10中的多个RAN节点110可以为同一类型的节点，也可以为不同类型的节点。在一些场景下，RAN节点110和终端120的角色是相对的，例如，图1中网元120i可以是直升机或无人机，其可以被配置成移动基站，对于那些通过网元120i接入到RAN 100的终端120j来说，网元120i是基站；但对于基站110a来说，网元120i是终端。RAN节点110和终端120有时都称为通信装置，例如图1中网元110a和110b可以理解为具有基站功能的通信装置，网元120a-120j可以理解为具有终端功能的通信装置。
在一种可能的场景中，RAN节点可以是基站(base station)、演进型基站(evolved NodeB，eNodeB)、接入点(access point，AP)、发送接收点(transmission reception point，TRP)、未来移动通信系统中的基站、或WiFi系统中的接入节点等。RAN节点可以是宏基站(如图1中的110a)、微基站或室内站(如图1中的110b)、中继节点或施主节点、或者是CRAN场景下的无线控制器。可选的，RAN节点还可以是服务器，可穿戴设备，车辆或车载设备等。例如，车辆外联(vehicle to everything，V2X)技术中的接入网设备可以为路侧单元(road side unit，RSU)。
在另一种可能的场景中，由多个RAN节点协作协助终端实现无线接入，不同RAN节点分别实现基站的部分功能。例如，RAN节点可以是CU，DU，CU-控制面(control plane，CP)，CU-用户面(user plane，UP)，或者无线单元(radio unit，RU)等。CU和DU可以是单独设置，或者也可以包括在同一个网元中，例如基带单元(baseband unit，BBU)中。RU可以包括在射频设备或者射频单元中，例如包括在射频拉远单元(remote radio unit，RRU)、有源天线处理单元(active antenna unit，AAU)或远程射频头(remote radio head，RRH)中。
在不同系统中，CU(包括开放式CU-CP(open CU-CP，O-CU-CP)和开放式CU-UP(open CU-UP，O-CU-UP)、DU或RU也可以有不同的名称，但是本领域的技术人员可以理解其含义。例如，在ORAN系统中，CU也可以称为开放式集中式单元(open central unit，O-CU)，DU也可以称为开放式分布式单元(open distributed unit，O-DU)，CU-CP也可以称为O-CU-CP，CU-UP也可以称为O-CU-UP，RU也可以称为O-RU。为描述方便，本申请中以CU，CU-CP，CU-UP、DU和RU为例进行描述。本申请中的CU(或CU-CP、CU-UP)、DU和RU中的任一单元，可以是通过软件模块、硬件模块、或者软件模块与硬件模块结合来实现。
终端(terminal)120，可以为接入上述通信系统、且具有相应通信功能的设备或模组。终端也可以称为用户设备(user equipment，UE)、终端、用户装置、接入终端、用户单元、用户站、移动站、移动台(mobile station，MS)、远方站、远程终端、移动设备、用户终端、终端单元、终端站、终端装置、无线通信设备、用户代理或用户装置。终端内通常设置有执行相应通信功能的通信模组、电路或芯片。终端内还以配置用于执行相应通信功能的程序指令。
例如，本申请实施例中的终端可以是手机(mobile phone)、个人数字助理(personal digital assistant，PDA)电脑、膝上型电脑(laptop computer)、平板电脑(Pad)、无人机、带无线收发功能的电脑、MTC终端、虚拟现实(virtual reality，VR)终端、增强现实(augmented reality，AR)终端、物联网(internet of things，IoT)终端、工业控制(industrial control)中的无线终端、无人驾驶(self driving)中的无线终端、远程医疗(remote medical)中的无线终端、智能电网(smart grid)中的无线终端、运输安全(transportation safety)中的无线终端、智慧城市(smart city)中的无线终端、智慧家庭(smart home)中的无线终端(例如游戏机、智能电视、智能音箱、智能冰箱和健身器材等)、具有无线通信功能的运输载具、通信模组、具有终端功能的RSU。
RAN 100和终端120可以部署在陆地上，包括室内或室外、手持或车载；也可以部署在水面上；还可以部署在空中的飞机、气球和卫星上。本申请实施例中对RAN 100和终端120所处的场景不作限定。
CN 200可以是5G核心网，也可以是演进的5G核心网。以5G核心网为例，CN 200中包括负责移动性管理、接入管理等服务的接入和移动性管理功能(access and mobility management function，AMF)网元、负责会话管理的会话管理功能(session management function，SMF)网元、负责用户面的数据包路由转发及服务质量(quality of service，QoS)控制的用户面功能(user plane function，UPF)网元、策略控制功能(policy control function，PCF)网元等。上述核心网元可以独立工作，也可以组合在一起实现某些控制功能，如：AMF、SMF和PCF可以组合在一起作为一个核心网设备。
应理解，上述命名仅为便于区分不同的功能而定义，不应对本申请构成任何限定。本申请并不排除在5G网络以及未来其它的网络中采用其他命名的可能。例如，在未来网络中，上述各个网元中的部分或全部可以沿用5G中的术语，也可能采用其他名称等。
可以理解，图1只是一种示例，对本申请的保护范围不构成限定。本申请实施例提供的通信方法还可以涉及图1中未示出的设备，如无线中继设备和/或无线回传设备等，当然本申请实施例提供的通信方法也可以只包括图1示出的部分设备。
随着无线通信应用场景的日益丰富，未来无线通信过程中会产生大量的原生数据(例如，无线接入网原生数据native data，或本地数据local traffic)，这些原生数据对传输带来了新的需求。
图2是原生数据压缩及传输处理流程的示意图。如图2所示，在发送端，输入数据经过数据重组、数据过滤、数据变换、数据选择(或者说数据丢弃等)、量化、熵编码、信道映射等操作，得到待发送的码流，然后进行信道编码、调制等处理后通过无线空口传输。其中，输入数据可以是原生数据，例如感知数据、AI数据、或信道数据等。相应地，在接收端，首先通过解调、信道解码等操作得到二进制码流，再对二进制码流进行信道解映射、熵解码、反量化、数据恢复、数据反变换、数据补零、数据重组恢复等操作后输出重构数据。通常情况下，重构数据和输入数据的维度相等，对应的数据取值可能存在些许误差。
可以理解的，某一原生数据可以对应一种或多种数据类型。例如，针对不同的应用场景，传输过程可能会使用不同的数据类型。然而，当前针对某一原生数据对应的某一数据类型，通常采用固定的数据重组方案进行数据重组，无法支持在多种重组方案之间动态切换，也就无法适应不同任务或业务需求。因此，亟需定义针对特定数据类型的原生数据的多种重组方案，以支撑物理层压缩的标准化流程。
为了解决上述技术问题，本申请提供一种数据传输方法和装置，以期适应不同任务或业务需求，保证数据的传输性能。
下文将结合附图详细说明本申请实施例提供的数据传输方法，可以应用于上述图1所示的通信系统中。应理解，本申请实施例可以适用于发送端和接收端通信的场景。
还应理解，下文示出的实施例并未对本申请实施例提供的方法的执行主体的具体结构特别限定，只要能够通过运行记录有本申请实施例的提供的方法的代码或程序，以根据本申请实施例提供的方法进行通信即可。例如，本申请实施例提供的方法可以由第一装置(或编码设备)和第二装置(或解码设备)执行，在并不特殊说明的情况下，本申请中的“第一装置”或“第二装置”既可以指装置本身(例如，终端设备或网络设备)，也可以是装置中的组件(例如，通信模组、处理器、电路、芯片(如调制解调(modem)芯片，又称基带(baseband)芯片，或包含modem核的SoC芯片或SIP芯片)、或芯片系统等)，或者也可以是能实现全部或部分装置功能的逻辑模块或软件。
图3是本申请实施例提供的一种数据传输方法的流程示意图。如图3所示，该方法300包括如下多个步骤。
S310，第一装置确定第一重组模式。
其中，第一重组模式用于指示针对第一数据类型的数据的分组方式，第一重组模式属于P个重组模式，P个重组模式与第一数据类型对应，P为大于1的整数。
也就是说，第一装置在确定第一重组模式之后，也就确定第一数据类型的数据(例如，数据#a)对应的分组方式。其中，第一数据类型可以对应多个重组模式，每一个重组模式指示一种重组方案，即每一个重组模式对应数据#a的一种分组方式，第一重组模式是多个重组模式中的一个重组模式。
在本申请中，第一数据类型的数据(例如，数据#a)可以是原生数据，例如可以是感知数据、AI数据、或信道数据中的一种，或者也可以是其他原生数据，不予限定。数据#a对应的数据类型可以是一个或多个。例如，感知数据对应的数据类型包括：环境反射点、环境面片、环境重构地图、或射频地图中的至少一个，则感知数据对应的第一数据类型为环境反射点、环境面片、环境重构地图、或射频地图中的一个；再例如，AI数据对应的数据类型包括：AI训练数据、AI推理数据、AI模型参数、AI梯度参数、AI中间特征数据、或AI性能数据的至少一个，则AI数据对应的第一数据类型为AI训练数据、AI推理数据、AI模型参数、AI梯度参数、AI中间特征数据、或AI性能数据中的一个；又例如，信道数据对应的数据类型包括：信道矩阵(如多天线系统中设备反馈的H矩阵)、CSI矩阵、或信道指示信息(如信道质量指示(channel quality indicator，CQI)、或秩指示(rank indicator，RI))中的至少一个，则信道数据对应的第一数据类型为信道矩阵、CSI矩阵、或信道指示信息中的一个。
在本申请中，重组模式包括以下一项或多项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，基于语义信息的重组模式，或者，基于重要性排序的重组模式。那么，第一重组模式可以是：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式、基于语义信息的重组模式，或者，基于重要性排序的重组模式中的一个。
下面结合图4至图9，针对本申请实施例中涉及的重组模式及其对应的分组方式进行示例说明。
(1)基于顺序拆分的重组模式。
该重组模式中，根据数据存储或者记录的顺序进行划分得到不同的分组，实现复杂度低，可以支持多种数据类型。
图4是基于顺序拆分的重组模式对数据进行分组的结构示意图。如图4的(a)所示，第一数据类型为AI数据中的AI中间特征数据，该中间特征数据的长度(或个数)为N，按顺序从上到下依次均等划分为M个分组，即第一数据分组的个数为M，每个分组包含N/M个中间特征数据。如图4的(b)所示，第一数据类型为信道数据中的信道矩阵(例如H矩阵)，该H矩阵的维度为N1*N2，例如针对第二个维度，按顺序从左到右依次均等划分为M个组，每个分组包含N2/M列数据，此时H矩阵的维度为N1*(N2/M))。可选地，针对该H矩阵的第一个维度，按顺序从上到下依次均等划分为M个组，每个分组包含N1/M行数据，此时H矩阵的维度为(N1/M)*N2/M。
可以理解的，上述重组模式对应的分组方式是采用均等划分的方式进行分组的，即M个分组中的每个分组包含的数据长度(或者说每个分组大小)相等，可选地，上述重组模式对应的分组方式也可以采用不均等划分的方式进行分组，即M个分组中至少存在两个分组的数据长度不等，对此不作限制。
可选地，针对图4所示的基于顺序拆分的重组模式，第一装置和第二装置需对齐针对第一数据类型的数据的分组方式。例如，针对上述分组方式所包含的分组个数M，以及每个分组的大小，可以是预定义或预配置的，此时无需显式地指示重组指示信息#1；再例如，该方法300还包括：第一装置/第二装置获取或输出重组指示信息#1，对于图4的(a)，该重组指示信息#1指示分组的大小N/M或组数M；对于图4的(b)，该重组指示信息#1指示分组的大小N2/M(或N1/M)或组数M。可选地，如果考虑非均匀大小的分组，则该重组指示信息#1指示M个分组中的每个分组的大小。
(2)基于比特位置的重组模式。
该重组模式中，首先将数据表示为比特流形式(如使用浮点数表示，或经过量化后表示)，然后根据比特位的高低进行划分得到不同的分组，便于与比特分层、或增量传输等结合设计。
图5是基于比特位置的重组模式对数据进行分组的结构示意图。为便于描述和理解，以下示例中基于比特位置的重组模式对应的分组方式是采用均等划分的方式进行分组的，即M个分组中的每个分组包含的数据长度相等。
如图5的(a)所示，第一数据类型为感知数据中的环境反射点，该环境反射点的长度(或个数)为N，例如P1,P2,…PN，首先对N个环境反射点依次进行Q比特量化，M＜Q，则N个数据可以得到一个N*Q的比特矩阵，然后对该比特矩阵的第二个维度(对应量化比特层)按顺序进行均等划分为M个分组，即第一数据分组的个数为M，每个分组包含Q/M个比特层，共计N*(Q/M)个比特。例如，以均等划分的方式对N个数据进行分组，此时Q＝M，表示每个分组包括一个比特层，如图5的(a)的虚线框所示，即一列代表一个比特层。可选地，也可以对该比特矩阵的第一个维度(对应量化比特层)按顺序进行均等划分为M个分组，每个分组包含N/M个比特层，共计(N/M)*Q个比特。
可选地，上述图5的(a)对应的分组方式也可以采用不均等划分的方式进行分组，即M个分组中至少存在两个分组的数据长度不等，对此不作限制。此时Q≠M，表示每个分组可以包含不同个数的比特层，例如图5的(b)的虚线框所示，分组1可以包括两个比特层，分组M可以包括一个比特层。
如图5的(c)所示，第一数据类型为AI数据中的AI模型参数，该AI模型参数为N1*N2的矩阵，对每个参数进行Q比特量化，M＜Q，则N1*N2个数据可以得到一个N1*N2*Q的三维比特矩阵，然后对该比特矩阵的第三个维度(对应量化比特层)按顺序进行均等划分为M个分组，每个分组包含Q/M个比特层，共计N1*N2*(Q/M)个比特。例如，以均等划分的方式对N个数据进行分组，此时Q＝M，表示每个分组包括一个比特层，如图5的(c)的虚线框所示，即一个平面代表一个比特层。可选地，也可以对该比特矩阵的第一个维度按顺序进行均等划分为M个分组，每个分组包含N1/M个比特层，共计(N1/M)*N2*Q个比特。可选地，也可以对该比特矩阵的第二个维度按顺序进行均等划分为M个分组，每个分组包含N2/M个比特层，共计N1*(N2/M)*Q个比特。
可选地，上述图5的(c)对应的分组方式也可以采用不均等划分的方式进行分组，即M个分组中至少存在两个分组的数据长度不等，对此不作限制。此时Q≠M，表示每个分组可以包含不同个数的比特层，例如分组1可以包括两个比特层，也就是包括两个平面，分组4可以包括一个比特层，也就是包括一个平面，分组M可以包括三个比特层，也就是包括三个平面，等等，图中未示出。
可选地，针对图5所示的基于比特位置的重组模式，第一装置和第二装置需对齐针对第一数据类型的数据的分组方式。例如，针对上述分组方式所包含的分组个数M，以及每个分组的大小，可以是预定义或预配置的，此时无需显式地指示重组指示信息#2；再例如，该方法300还包括：第一装置/第二装置获取或输出重组指示信息#2，对于图5的(a)，该重组指示信息#2指示分组的大小Q/M(或者N/M)或组数M；对于图5的(c)，该重组指示信息#2指示分组的大小Q/M(或者N1/M、或N2/M)或组数M。可选地，如果考虑非均匀大小的分组，对于图5的(b)，该重组指示信息#2指示M个分组中的每个分组的大小。
(3)基于空间位置的重组模式。
该重组模式中，根据数据所在的位置信息进行划分得到不同的分组，适用于具有地理位置信息的感知数据，可以结合地理位置信息对数据进行联合处理。
图6是基于空间位置的重组模式对数据进行分组的结构示意图。如图6所示，第一数据类型为感知数据中的射频地图(例如，基于二维坐标构建的射频地图)，该射频地图包括射频参数1至射频参数14，将物理位置临近的多个射频参数化分为一组，得到4个分组，即第一数据分组的个数为M。例如，分组1包括射频参数1至射频参数4，分别对应的二维坐标信息是(x1,y1)、(x1,y2)、(x1,y3)和(x1,y4)；分组2包括射频参数5至射频参数7，射频参数5对应的二维坐标信息是(x2,y1)和(x2,y2)，射频参数6对应的二维坐标信息是(x2,y3)，射频参数7对应的二维坐标信息是(x2,y4)；分组3包括射频参数8、射频参数9和射频参数12，射频参数8对应的二维坐标信息是(x3,y1)和(x4,y1)，射频参数9对应的二维坐标信息是(x3,y2)，射频参数12对应的二维坐标信息是(x4,y2)；分组4包括射频参数10、射频参数11、射频参数13和射频参数14，分别对应二维坐标信息(x3,y3)、(x3,y4)、(x4,y3)和(x4,y4)。
可以理解的，上述重组模式对应的分组方式是采用不均等划分的方式进行分组的，即M个分组中至少存在两个分组的数据长度(或者说射频参数的个数)不等，可选地，上述重组模式对应的分组方式也可以采用均等划分的方式进行分组的，即M个分组中的每个分组包含的射频参数的个数相等，对此不作限制。
可选地，针对图6所示的基于空间位置的重组模式，第一装置和第二装置需对齐针对第一数据类型的数据的分组方式。例如，针对上述分组方式所包含的分组个数M，以及每个分组包含的射频参数及其所在的坐标范围信息，可以是预定义或预配置的，此时无需显式地指示重组指示信息#3；再例如，该方法300还包括：第一装置/第二装置获取或输出重组指示信息#3，该重组指示信息#3可以包括每个分组对应的射频参数索引，例如以射频参数索引集合的形式指示，如分组1对应的索引集合为{1,2,3,4}，分组2对应的索引集合为{5,6,7}，分组3对应的索引集合为{8,9,12}，分组4对应的索引集合为{10,11,13,14}，通过发送射频参数索引可以确定对应的地理坐标信息。
(4)基于语义信息的重组模式。
该重组模式中，根据数据对应的语义类别进行划分得到不同的分组，适用于划分了语义特征的数据(如进行了语义分割的环境反射点、环境面片、环境重构地图等数据，这些数据对应的语义信息可以指示不同的对象、前景/背景等)，可以结合数据语义特征对数据进行分组处理。
图7是基于语义信息的重组模式对数据进行分组的结构示意图。如图7所示，第一数据类型为感知数据中的环境面片，该环境面片包括大楼面片(例如，大楼1面片和大楼2面片)，路灯面片(例如，路灯1面片、路灯2面片和路灯3面片)和街区路面面片。针对不同的对象，如大楼1、大楼2、路灯1、路灯2、路灯3和街区路面对应的面片进行区分，得到不同的分组环境面片数据，例如3个分组，即第一数据分组的个数为M＝3。
可以理解的，上述重组模式对应的分组方式是采用不均等划分的方式进行分组的，即M个分组中至少存在两个分组的数据长度(或者说环境面片的个数)不等，可选地，上述重组模式对应的分组方式也可以采用均等划分的方式进行分组的，即M个分组中的每个分组包含的环境面片的个数相等，对此不作限制。
可选地，针对图7所示的基于语义信息的重组模式，第一装置和第二装置需对齐针对第一数据类型的数据的分组方式。例如，针对上述分组方式所包含的分组个数M，以及每个分组包含的环境面片，可以是预定义或预配置的；或者，环境面片数据隐含环境面片所在的坐标信息，此时无需显式地指示重组指示信息#4。
(5)基于重要性排序的重组模式。
该重组模式可以结合数据具体应用场景，获得不同重要性的数据码流，支持非均等保护UEP传输设计，即相对更重要的码流使用可靠度较高的信道编码、或调制方式；反之，相对不重要的码流使用可靠度较低的信道编码、或调制方式。
方式一：以信号功率或强度表征重要性，按照信号功率或强度对长度为N的数据进行排序，划分得到不同的分组，例如，将信号功率或强度高(重要性更高)的数据划分到一个分组，将信号功率或强度低(重要性较低)的数据划分到一个分组。
图8是基于重要性排序的重组模式对数据进行分组的一结构示意图。如图8所示，第一数据类型为感知数据中的射频地图(例如，以多径形式的射频地图数据)，该射频地图中的每个径包括多径强度、相位、时延、到达角和离去角等信息，将多径强度较大的一个或多个径划分到一个分组中(作为相对更重要的信息)，将多径强度较小的一个或多个径划分到一个分组中(作为相对不重要的信息)。例如，根据多径强度对数据进行划分得到M个分组，即第一数据分组的个数为M，每个分组包含一个或多个径的数据，比如分组1中的数据对应的多径强度最高(即最重要的分组)，依次类推，分组M中的数据对应的多径强度最低(即最不重要的分组)。
方式二：以对下游AI任务的贡献度表征重要性，如针对AI相关任务如目标识别、定位等，可以先评估不同数据分量对任务精度的影响程度(影响程度越大，贡献度也越大；反之贡献度越小)，用于衡量重要性并进行分组划分。
图9是基于重要性排序的重组模式对数据进行分组的另一结构示意图。如图9所示，第一数据类型为AI数据中的AI中间特征数据，长度为N，分别将每个中间特征数据置零后测试AI任务精度的下降值，根据精度下降值对N个AI中间特征数据进行排序分组得到M个分组，即第一数据分组的个数为M。例如，精度下降最大的AI中间特征数据作为最重要的信息，依次类推，精度下降最小的AI中间特征数据作为最不重要的信息，记录各AI中间特征数据的索引。每个分组可以包含AI中间特征数据的索引，例如分组1＝{1,3,…,N}，表示将第1,3,…,N个AI中间特征数据划分为一个组，…，分组M＝{4,7,…,N-2}，表示将第4,7,…,N-2个AI中间特征数据划分为一个组。
可选地，针对上述方式一或方式二，每个分组包含的数据的个数不作限定，可以相等，也可以不等，例如存在至少两个分组所包含的数据的个数不等。
可选地，针对图9所示的基于重要性排序的重组模式，第一装置和第二装置需对齐针对第一数据类型的数据的分组方式。例如，针对上述分组方式所包含的分组个数M，以及每个分组包含的数据，可以是预定义或预配置的，此时无需显式地指示重组指示信息#5；再例如，该方法300还包括：第一装置/第二装置获取或输出重组指示信息#5，该重组指示信息#5可以包括M个分组中的每个分组对应的数据的索引或编号。
上述图4至图9提供的重组模式及其对应的分组方式，仅是为便于理解给出的示例，不排除其他重组方案。可选地，本申请对重组模式的个数、每个重组模式划分的分组个数M，以及M个分组中的每个分组所包含的数据的个数不作限定。例如，本申请的重组方案可以均匀分组，也可以不均匀分组，例如针对长度为N的数据#a可以采用随机划分的方式进行分组。示例性地，通过随机种子确定M个分组中的每个分组对应的数据，第一装置和第二装置之间可以通过发送随机种子对应的编号或索引实现同步。可选地，第一装置和第二装置之间也可以协商，以其他任一方式对齐分组的方式，均属于本申请的保护范围，对此不作限定。
下面，针对第一数据类型及其对应的第一重组模式的确定规则和效果进行示例说明。
情形一：第一数据类型的数据为感知数据。
示例1，若第一数据类型为环境反射点，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式。该示例中，可以根据环境反射点的具体应用场景，确定对应的第一重组模式。例如，环境反射点数据被用于进行目标、物体识别任务时，可以选择基于空间位置或语义信息的重组模式，以实现更好的任务精度；再例如，环境反射点数据被用于进行环境重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；又例如，环境反射点数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。其中，增量传输是指获得多份比特层信息，终端设备可以每次发一个或多个比特层，当调度的传输资源发生变化时，可以优先发送高位比特层(重要性更高)信息。
示例2，若第一数据类型为环境面片，则第一重组模式包括以下一项：基于顺序拆分的重组模式，或者，基于语义信息的重组模式。该示例中，可以根据环境面片的具体应用场景，确定第一重组模式。例如，环境面片数据被用于进行目标、物体识别任务时，可以选择基于语义信息的重组模式，以实现更好的任务精度；再例如，环境面片数据被用于进行环境重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能。
示例3，若第一数据类型为环境重构地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式。该示例中，可以根据环境重构地图的具体应用场景，确定第一重组模式。例如，环境重构地图数据被用于进行目标、物体识别任务时，可以选择基于空间位置或语义信息的重组模式，以实现更好的任务精度；再例如，环境重构地图数据被用于进行环境重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；环境重构地图数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
示例4，若第一数据类型为射频地图，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于重要性排序的重组模式。该示例中，可以根据射频地图的具体应用场景，确定第一重组模式。例如，射频地图数据被用于进行目标、物体识别或波束方向获取等任务时，可以选择基于空间位置或语义信息的重组模式，以实现更好的任务精度；再例如，射频地图数据被用于进行环境重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；又例如，射频地图数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
情形二：第一数据类型的数据为AI数据。
示例1，若第一数据类型为AI训练数据或AI推理数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于语义信息的重组模式，或者，基于重要性排序的重组模式。该示例中，可以根据AI训练数据或AI推理数据的具体应用场景，确定第一重组模式。例如，AI训练数据或AI推理数据被用于进行目标、物体识别或波束方向获取等任务时，可以选择基于语义信息的重组模式，以实现更好的任务精度；再例如，AI训练数据或AI推理数据被用于进行数据重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；又例如，AI训练数据或AI推理数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
示例2，若第一数据类型为AI模型参数或AI梯度参数，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式。该示例中，可以根据AI模型参数或AI梯度参数的具体应用场景，确定第一重组模式。例如，UEP传输场景下，需要获得AI模型参数或AI梯度参数不同重要性数据码流时，可以选择基于重要性排序的重组模式，以实现更好的传输性能；再例如，非UEP传输场景下，AI模型参数或AI梯度参数对应数据码流采取相同的编码调制配置时，可以选择基于顺序拆分的重组模式，以实现复杂度较低；又例如，AI模型参数或AI梯度参数被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
示例3，若第一数据类型为AI中间特征数据，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于语义信息的重组模式。该示例中，可以根据AI中间特征数据的具体应用场景，确定第一重组模式。例如，AI中间特征数据被用于进行目标、物体识别或波束方向获取等任务时，可以选择基于语义信息的重组模式，以实现更好的任务精度；再例如，AI中间特征数据被用于做数据重构时，可以选择基于顺序拆分的重组模式，以实现更好的重构性能；又例如，AI中间特征数据被用于进行增量传输时，可以选择基于比特位置的重组模式，以灵活匹配不同传输资源。
示例4，若第一数据类型为AI性能数据，则第一重组模式包括基于顺序拆分的重组模式。
情形三：第一数据类型的数据为信道数据。
示例1，若第一数据类型为信道矩阵或CSI矩阵，则第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式。该示例中，可以根据信道矩阵或CSI矩阵的具体应用场景，确定第一重组模式。例如，UEP传输场景下，需要获得信道矩阵或CSI矩阵不同重要性数据码流时，可以选择基于重要性排序的重组模式，以实现更好的传输性能；再例如，非UEP传输场景下，信道矩阵或CSI矩阵对应数据码流采取相同的编码调制配置时，以选择基于顺序拆分的重组模式，以实现复杂度较低；又例如，信道矩阵或CSI矩阵被用于进行增量传输时，以选择基于比特位置的重组模式，可以灵活匹配不同传输资源。
示例2，若第一数据类型为信道指示信息，则第一重组模式包括基于顺序拆分的重组模式。
可以理解的，上述提供的不同原生数据对应的一个或多个数据类型，第一数据类型及其对应的第一重组模式的确定规则和效果，仅是为便于理解本申请给出的示例说明，不排除其他方案。
下面，针对上述步骤S310中第一装置确定第一重组模式的实现方式进行说明。
在一种实现方式中，该第一重组模式可以是预定义或预配置的。其中，预定义可以包括预先定义，例如协议定义，预配置可以通过在第一装置(和/或第二装置)预先保存相应的代码、表格、函数、文本、字符串或其他可用于指示相关信息(例如，第一重组模式)的方式来实现。
在另一种实现方式中，第一装置根据第一数据类型和第一映射关系确定第一重组模式。
其中，该第一映射关系用于指示K个重组模式与L个数据类型之间的对应关系，K个重组模式包括第一重组模式，L个数据类型包括第一数据类型，K个重组模式包含P个重组模式，P为小于或等于K的整数，L为大于或等于1的整数。有关第一映射关系的表现形式在下文具体描述，这里先不作说明。
可选地，在第一装置根据第一数据类型和第一映射关系确定第一重组模式之前，该方法300还包括：第一装置获取第一映射关系。例如，该第一映射关系可以是预定义或预配置的；或者，第一装置获取指示信息#a(即，第一指示信息的一例)，该指示信息#a用于指示第一映射关系，即该第一映射关系可以是网络设备通过信令指示或配置的。例如，第二装置向第一装置发送指示信息#a，该指示信息#a指示(或包括)第一映射关系，对应地，第一装置接收指示信息#a，根据该指示信息#a确定该第一映射关系。
可选地，在第一装置根据第一数据类型和第一映射关系确定第一重组模式之前，该方法300还包括：第一装置获取第一数据类型。例如，该第一数据类型可以是预定义或预配置的；或者，第一装置获取指示信息#b，该指示信息#b用于指示第一数据类型，即该第一数据类型可以是网络设备通过信令指示或配置的。例如，第二装置向第一装置发送指示信息#b，该指示信息#b指示第一数据类型，对应地，第一装置接收指示信息#b，根据该指示信息#b确定该第一数据类型。
可选地，基于上述两种实现方式，第一装置可以将确定的第一重组模式反馈给第二装置。例如，第一装置向第二装置发送指示信息#f(即第二指示信息的一例)，该指示信息#f用于指示第一重组模式。
可选地，在第一装置向第二装置发送指示信息#f之前，该方法300还包括：第二装置发送网络侧支持的一个或多个重组模式。其中，一个或多个重组模式包括第一重组模式。可选地，第二装置可以通过发送系统消息、广播消息或多播消息，指示网络侧支持的一个或多个重组。也就是说，第一装置根据网络侧支持的一个或多个重组模式，结合第一数据类型和第一映射关系可以确定第一重组模式。例如，假设第一数据类型为AI模型参数，对应的多个重组模式包括：基于顺序拆分的重组模式，基于比特位置的重组模式和基于重要性排序的重组模式，考虑到网络侧支持的重组模式为基于顺序拆分的重组模式和基于重要性排序的重组模式，则第一装置最终确定的第一重组模式可以是基于顺序拆分的重组模式或基于重要性排序的重组模式。
在又一种实现方式中，第一装置获取指示信息#c(即第二指示信息的一例)，该指示信息#c用于指示第一重组模式。即该第一重组模式可以是网络设备通过信令指示或配置的。例如，第二装置向第一装置发送指示信息#c，该指示信息#c指示(或包括)第一重组模式的索引或标识，对应地，第一装置接收指示信息#c，根据该指示信息#c确定该第一重组模式。
可选地，在第二装置向第一装置发送指示信息#c之前，该方法300还包括：第二装置确定第一重组模式。例如，该第一重组模式可以是预定义或预配置的；或者，第二装置根据第一数据类型和第一映射关系确定第一重组模式；或者，第二装置根据终端设备的能力信息确定第一重组模式，该能力信息指示终端设备支持的一个或多个重组模式，该一个或多个重组模式包括第一重组模式。
可选地，在第二装置根据第一数据类型和第一映射关系确定第一重组模式之前，该方法300还包括：第二装置获取第一映射关系。例如，该第一映射关系可以是预定义或预配置的，其中，预定义包括预先定义，例如协议定义。预配置可以通过在第一装置和/或第二装置中预先保存相应的代码、表格或其他可用于指示相关信息的方式来实现；或者，第二装置获取指示信息#d(即，第一指示信息的一例)，该指示信息#d用于指示第一映射关系，即该第一映射关系可以是终端设备通过信令指示或配置的。例如，第一装置向第二装置发送指示信息#d，该指示信息#d指示第一映射关系，对应地，第二装置接收指示信息#d，根据该指示信息#d确定该第一映射关系。
可选地，在第二装置根据第一数据类型和第一映射关系确定第一重组模式之前，该方法300还包括：第二装置获取第一数据类型。例如，该第一数据类型可以是预定义或预配置的；或者，第二装置获取指示信息#e，该指示信息#e用于指示第一数据类型，即该第一数据类型可以是终端设备通过信令指示或配置的。例如，第一装置向第二装置发送指示信息#e，该指示信息#e指示第一数据类型，对应地，第二装置接收指示信息#e，根据该指示信息#e确定该第一数据类型。
可选地，在第一装置获取指示信息#c之前，该方法300还包括：第一装置向第二装置发送指示信息#g(即，第三指示信息的一例)，该指示信息#g用于指示第一装置的能力信息，或者，第一装置支持的一个或多个重组模式，其中，一个或多个重组模式包括第一重组模式。也就是说，第二装置根据接收的指示信息#g可以确定终端设备支持的一个或多个重组模式，进而结合第一数据类型和第一映射关系可以确定第一重组模式。例如，假设第一数据类型为感知数据中的环境面片，对应的多个重组模式包括：基于顺序拆分的重组模式和基于语义信息的重组模式，考虑到终端侧支持的重组模式为基于顺序拆分的重组模式，则第二装置最终确定的第一重组模式即为基于顺序拆分的重组模式。
可选地，该指示信息#b或指示信息#a或指示信息#c可以承载于第一信令或第一资源，第一信令包括无线资源控制(radio resource control，RRC)信令、下行控制信息(downlink control information，DCI)信令、或媒体接入控制控制元素(medium access control-control element，MAC CE)，第一资源包括物理下行控制信道(physical downlink control channel，PDCCH)、或物理下行数据信道(physical downlink shared channel，PDSCH)。
可选地，该指示信息#d或指示信息#e或指示信息#f或指示信息#g可以承载于第二信令或第二资源，第二信令包括RRC信令、MAC CE信令、或上行控制信息(uplink control information，UCI)信令，第二资源包括物理上行控制信道(physical uplink control channel，PUCCH)、或物理上行数据信道(physical uplink shared channel，PUSCH)。
可以理解的，本申请对于第一装置确定第一重组模式的具体实现方式不作限定，只要第一装置和第二装置对齐该第一重组模式即可。
针对上述实现方式，第一装置(或第二装置)在确定第一重组模式之后，可以向第二装置(或第一装置)指示所选择的第一重组模式，具体示例如下。
作为一种示例，通过引入重组mode指示不同的重组模式。例如，假设定义了P种重组模式，则重组mode可以等于1,2,…,P，分别用于指示P种重组模式。
可选地，当第一数据类型只支持一种重组模式(即第一重组模式)时，则在确定第一数据类型时即可确定对应的重组mode，也就确定了该第一数据类型所支持的一种重组模式(即第一重组模式)；当第一数据类型支持多种重组模式时，第一装置或第二装置可以通过信令配置或指示重组mode编号，进而确定第一数据类型对应的第一重组模式。
例如，第一装置或第二装置可以在RRC信令中新增一个重组模式reorganizeMode字段，该字段中可以携带重组mode的索引或编号，用于指示对应的重组模式。
举例来说，当P＝5时，表示第一数据类型对应5个重组模式，5个重组模式一一对应5种重组方案，也即第一数据类型支持5种重组方案。其中，reorganizeMode中可以包含重组mode编号，如1、2、3、4、5，可以理解为：重组mode＝1对应重组模式#1，重组mode＝2对应重组模式#2，重组mode＝3对应重组模式#3，重组mode＝4对应重组模式#4，重组mode＝5对应重组模式#5。即通过指示重组mode编号即可确定对应的重组模式。
再例如，reorganizeMode既可以直接指示重组mode编号，也可以指示特定mode集合中的元素索引。
举例来说，当P＝5时，表示第一数据类型存在5个重组模式，5个重组模式一一对应5种重组方案，也即第一数据类型支持5种重组方案。假设某一数据类型支持的重组模式对应的重组mode编号分别为1,3,4,5，可选地，以集合的形式表征为{1,3,4,5}，表示该数据类型支持的重组模式包括：重组模式#1、重组模式#3、重组模式#4和重组模式#5。那么，reorganizeMode的取值1～4分别对应重组mode＝1、3、4、5，可以理解为：reorganizeMode的取值为1对应重组模式#1，reorganizeMode的取值为2对应重组模式#3，reorganizeMode的取值为3对应重组模式#4，reorganizeMode的取值为4对应重组模式#5。即reorganizeMode的取值即为对应的集合中的第几个重组方案。
作为另一种示例，通过比特信息指示不同的重组模式。例如，假设定义了P种重组模式，则第一装置可以使用个比特指示第一装置确定的第一重组模式，表示向上取整，P为大于或等于1的整数。比如，P＝2，表示存在两种重组模式，则可以使用log2P＝1比特信息指示确定的重组模式，比如“0”指示第一个重组模式，“1”指示第二个重组模式，假设第一装置选择第一个重组模式(即第一重组模式)对该第一数据类型的数据进行分组，则可以发送携带“0”指示信息，用于指示其选定的重组模式。
下面，针对上文提及的第一映射关系的表现形式进行示例说明。其中，该第一映射关系用于指示K个重组模式与L个数据类型之间的对应关系，该第一映射关系可以以表格，集合，函数，文本，或者，字符串的形式存在，如存储或传输，本申请对于其具体的实现方式不作限定。为便于理解，本申请以表格的形式对第一映射关系进行示例，如下表1或表2所示，表格中的“1”表示该数据类型支持对应的重组模式，“0”表示该数据类型不支持对应的重组模式。
表1

由表1可知，第一映射关系用于指示K个重组模式(即K＝5)与L个数据类型(即L＝13)之间的对应关系。其中，感知数据对应的多种数据类型主要支持：基于顺序拆分的重组模式、基于比特位置的重组模式、基于空间位置的重组模式和基于语义信息的重组模式，AI数据对应的多种数据类型主要支持：基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式，信道数据对应的多种数据类型主要支持基于顺序拆分的重组模式。
举例来说，假设第一数据类型为环境重构地图，则对应的P个重组模式(即P＝4)包括：基于顺序拆分的重组模式、基于比特位置的重组模式、基于空间位置的重组模式和基于语义信息的重组模式，其中第一重组模式可以为基于顺序拆分的重组模式；再例如，假设第一数据类型为AI性能数据，则对应的P个重组模式(即P＝1)包括基于顺序拆分的重组模式，即第一重组模式为基于顺序拆分的重组模式；又例如，假设第一数据类型为CSI矩阵，则对应的P个重组模式(即P＝3)包括：基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式，其中第一重组模式可以为基于比特位置的重组模式。
作为一示例，该第一映射关系可以通过采用bitmap的方式存储或指示。其中，表1中的“1”和“0”分别记录为比特1和0，用于该数据类型是否支持该重组模式。例如，环境反射点支持基于顺序拆分的重组模式、基于比特位置的重组模式、基于空间位置的重组模式和基于语义信息的重组模式，对应的重组mode可以是1、2、3和4；再例如，AI训练数据支持基于顺序拆分的重组模式、基于比特位置的重组模式、基于语义信息的重组模式和基于重要性排序的重组模式，对应的重组mode可以是1、2、4和5；又例如，信道矩阵支持基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式，对应的重组mode可以是1、2和5，等等。
作为另一示例，该第一映射关系可以通过枚举的方式存储或指示，此时多个数据类型和多个重组模式之间的对应关系可以通过重组mode集合的方式表示。例如，感知数据中的环境反射点表示为{1,2,3,4}，说明环境反射点支持重组mode＝1,2,3,4对应的基于顺序拆分的重组模式、基于比特位置的重组模式、基于空间位置的重组模式和基于语义信息的重组模式；类似的，环境面片表示为{1,4}，环境重构地图表示为{1,2,3,4}，射频地图表示为{1,2,3,5}。再例如，AI数据中的AI中间特征数据表示为{1,2,5}，说明AI中间特征数据支持重组mode＝1,2,5对应的基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式；类似的，AI模型参数/AI梯度参数表示为{1,2,4}，AI训练数据/AI推理数据表示为{1,2,4,5}，AI性能数据表示为{1}。又例如，信道数据中的信道矩阵/CSI矩阵表示为{1,2,5}，说明信道矩阵/CSI矩阵支持重组mode＝1,2,5对应的基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式；类似的，信道指示信息表示为{1}。
表2
由表2可知，第一映射关系用于指示K个重组模式(即K＝4)与L个数据类型(即L＝6)之间的对应关系。其中，感知数据对应的多种数据类型主要支持：基于顺序拆分的重组模式、基于比特位置的重组模式和基于空间位置的重组模式，AI数据对应的多种数据类型主要支持：基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式，信道数据对应的多种数据类型主要支持基于顺序拆分的重组模式。
举例来说，假设第一数据类型为环境反射点，则对应的P个重组模式(即P＝3)包括：基于顺序拆分的重组模式、基于比特位置的重组模式和基于空间位置的重组模式，其中第一重组模式可以为基于比特位置的重组模式；再例如，假设第一数据类型为AI模型参数，则对应的P个重组模式(即P＝3)包括基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式，其中第一重组模式可以为基于重要性排序的重组模式；又例如，假设第一数据类型为信道指示信息，则对应的P个重组模式(即P＝1)包括：基于顺序拆分的重组模式，即第一重组模式为基于比特位置的重组模式。
作为一示例，该第一映射关系可以通过采用bitmap的方式存储或指示。其中，表1中的“1”和“0”分别记录为比特1和0，用于该数据类型是否支持该重组模式。例如，环境反射点支持基于顺序拆分的重组模式、基于比特位置的重组模式和基于空间位置的重组模式，对应的重组mode可以是1、2和3；再例如，AI训练数据支持基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式，对应的重组mode可以是1、2和5；又例如，信道指示信息支持基于顺序拆分的重组模式，对应的重组mode可以是1，等等。
作为另一示例，该第一映射关系可以通过枚举的方式存储或指示，此时多个数据类型和多个重组模式之间的对应关系可以通过重组mode集合的方式表示。例如，感知数据中的环境反射点表示为{1,2,3}，说明环境反射点支持重组mode＝1,2,3对应的基于顺序拆分的重组模式、基于比特位置的重组模式和基于空间位置的重组模式；类似的，环境重构地图和射频地图均可以表示为{1,2,3}。再例如，AI数据中的AI训练数据表示为{1,2,5}，说明AI中间特征数据支持重组mode＝1,2,5对应的基于顺序拆分的重组模式、基于比特位置的重组模式和基于重要性排序的重组模式；类似的，AI模型参数表示为{1,2,5}。又例如，信道数据中的信道指示信息表示为{1}，说明信道指示信息支持重组mode＝1对应的基于顺序拆分的重组模式。
可选地，本申请对上述任一表格中的对应关系的数量(例如表格中的一行)不作限定，例如可以增加或减少一行或多行。可选地，本申请对上述表格中的原生数据的个数、原生数据对应的数据类型以及重组模式的个数不作限定，对于多个数据类型与多个重组模式之间的对应关系不作限定，或者说，本申请对表格中的行数或列数不作限定。可选地，上述任一表格可以拆分成多个独立的表格，本申请对拆分方式不限定。例如，以原生数据为粒度进行拆分，比如表1中的感知数据、AI数据和信道数据对应的第一映射关系可以分别独立成三个新的表格；或者，表2中的感知数据和AI数据的第一映射关系可以独立成新的表格，表2中的信道数据对应的第一映射关系可以独立成新的表格。再例如，以数据类型为粒度进行拆分，比如表1中的前三行对应的第一映射关系独立成新的表格，中间四行对应的第一映射关系独立成新的表格，后三行对应的第一映射关系独立成新的表格；或者，表2中的前四行对应的第一映射关系独立成新的表格，后两行对应的第一映射关系独立成新的表格。可选地，上述表1可以拆分成多个其他表格进行示例，例如表1中的前三种重组模式与多种数据类型之间的第一映射关系，以及后两种重组模式与多种数据类型之间的第一映射关系分别独立成新的表格，再例如表2中的前两种重组模式与多种数据类型之间的第一映射关系，以及后两种重组模式与多种数据类型之间的第一映射关系分别独立成新的表格，等等，以上示例仅是为便于理解给出的，不排除其他方案，针对第一映射关系的任何变形均落入本申请保护范围。
S320，第一装置根据第一重组模式对第一数据类型的数据进行分组，得到第一数据分组。
可选地，本申请对第一数据分组的分组个数，以及每个分组包含的数据大小不作限定，具体的分组方式可参考上述图4至图9的相关描述，为了简洁，此处不作说明。
例如，假设第一数据类型的数据的长度为N，根据第一重组模式对第一数据类型的数据进行分组得到的第一数据分组的个数为M，M为大于1且小于N的整数。例如N＝8，M＝4，说明存在长度为8的第一类型的数据，第一装置可以将该数据分为4组，则这4组可以包含等量(或等长度、或相同个数)的数据，即每组包含2个数据；或者，这4组可以包含不等量(或不等长度)的数据，如4组分别包含1,2,2,3个数据块。
S330，第一装置输出第一数据，该第一数据是由第一数据分组经过处理得到的。
对应地，第二装置接收第一数据。
其中，第一数据是由第一数据分组经过处理得到的，可以理解为：第一类型的数据经过第一重组模式进行分组后，可以再经过数据过滤、数据变换、数据选择(或者说数据丢弃)、量化、熵编码、信道映射、信道编码、或调制中的至少一项操作，得到第一数据。针对所述至少一项操作的具体实现方式，可以参考现有方案的相关描述，为了简洁，此处不作说明。
示例性地，第一装置可以直接或间接地向第二装置发送第一数据。例如，第一装置向第二装置发送第一数据；再例如，第一装置通过第三装置向第二装置发送第一数据，该第三装置可以是核心网设备、网络设备或终端设备。
S340，第二装置对第一数据进行处理得到第一数据分组。
可选地，第二装置对第一数据经过解调、信道解码等操作得到二进制码流，再对二进制码流进行信道解映射、熵解码、反量化、数据恢复、数据重组恢复等操作后，得到第一数据分组。其中，针对解调、信道解码、信道解映射、熵解码、反量化、数据恢复、数据反变换、数据补零、或数据重组恢复等操作的实现方式，可以参考现有方案的相关描述，为了简洁，此处不作说明。
S350，第二装置根据第一重组模式对第一数据分组进行数据恢复，得到针对第一数据类型的重构数据。
可以理解的，对于第二装置来说，第一重组模式是明确知道的，可以通过预定义或预配置确定，或者，可以是第一装置通过信令指示的，本申请对此不作限定。
作为示例，第一装置根据第一重组模式，可以确定第一装置针对第一数据类型的数据进行的分组方式，因此，第二装置根据第一重组模式可以对第一数据分组进行重构，得到针对第一数据类型的重构数据。
通常情况下，重构数据和步骤S310中未进行分组之前的数据的维度是一致的，由于计算误差或处理误差，使得重构的第一数据类型的数据与未进行分组之前的数据的取值可能存在些许偏差。
可以理解的，上述方案主要针对使用第一重组模式对第一数据类型的数据进行分组，得到第一数据分组，以支持某一业务传输。可选地，考虑到第一数据类型对应P个重组模式，可以适应不同任务或业务需求，即第一数据类型支持不同的重组模式的灵活切换，从而保证传输性能。示例性地，针对同一数据类型，若网络侧的业务类型发生了变化，例如初始时反馈的第一数据用于数据融合，后续待反馈的第二数据用于执行特定检测任务，则可以采用不同的重组模式对第一数据类型的数据进行分组，满足不同任务或业务需求，提升端到端性能。
可选地，方法300还包括如下步骤S301至步骤S305。
S301，第一装置获取第四指示信息，例如，第二装置向第一装置发送第四指示信息，对应地，第一装置接收来自第二装置的第四指示信息，该第四指示信息用于指示第二重组模式，该第二重组模式用于指示针对第一数据类型的数据的分组方式。有关第二重组模式的具体释义和表现形式可以参考上文第一重组模式的相关描述，为了简洁，此处不作说明。
可选地，作为一种示例，在获取第四指示信息之前，方法300还包括：第一装置获取第五指示信息，例如，第二装置向第一装置发送第五指示信息，对应地，第一装置接收来自第二装置的第五指示信息，该第五指示信息用于指示第二映射关系，第二映射关系用于指示K’个重组模式与L’个数据类型之间的对应关系，K’个重组模式包括第二重组模式，L’个数据类型包括所述第一数据类型，K’为大于1的整数，L’为大于或等于1的整数。有关第二映射关系的具体释义和表现形式可以参考上文第一映射关系的相关描述，为了简洁，此处不作说明。
该实现方式中，第二装置可以通过发送第五指示信息，向第一装置更新数据类型和重组模式之间的对应关系，例如可以是指示从第一映射关系更新为第二映射关系，或者，也可以是指示从第三映射关系更新为第二映射关系，该第三映射关系可以是更新第一映射关系得到的。进一步地，第一装置通过获取第四指示信息和更新后的第二映射关系确定第二重组模式。
可选地，K’个重组模式与上述步骤S310中的P个重组模式可以相等，也可以不等，例如K’个重组模式包括P个重组模式，还可以存在部分相同的重组模式，对此不予限定。可选地，K’个重组模式与上述第一映射关系中的K个重组模式可以相等，也可以不等，还可以存在部分相同的重组模式，对此不予限定。
可选地，L’个重组模式与上述第一映射关系中的L个数据类型可以相等，也可以不等，还可以存在部分相同的重组模式，对此不予限定。
可选地，作为另一种示例，在获取第四指示信息之前，方法300还包括：第一装置获取第六指示信息，例如，第二装置向第一装置发送第六指示信息，对应地，第一装置接收来自第二装置的第六指示信息，该第六指示信息用于指示第一数据类型对应的Q个重组模式，Q个重组模式包括第二重组模式，Q为大于或等于2的整数。
可选地，Q个重组模式与上述步骤S310中的P个重组模式可以相等，也可以不等，例如Q个重组模式包括P个重组模式，还可以存在部分相同的重组模式，对此不予限定。可选地，Q个重组模式与上述第一映射关系中的K个重组模式可以相等，也可以不等，还可以存在部分相同的重组模式，对此不予限定。
该实现方式中，第二装置可以通过发送第六指示信息，向第一装置指示第一数据类型对应的Q个重组模式，进一步地，第一装置通过获取第四指示信息可以确定更新后的第二重组模式。可选地，Q个重组模式可以以集合的形式存在，例如Q＝4，对应重组mode＝{1,2,3,4}，假设第一重组模式对应重组mode＝1，则该第六指示信息可以包括重组mode＝3，表示第二重组模式对应重组mode 3，例如基于空间位置的重组模式。
可选地，如果该第一数据类型对应的Q个重组模式等于步骤S310中的P个重组模式，则该第六指示信息可以不发送。
S302，第一装置根据第二重组模式对第一数据类型的数据进行分组，得到第二数据分组。
其中，第二数据分组与第一数据分组不同，可以理解为：第一数据分组和第二数据分组所包含的分组个数不同，例如，第一数据分组和第二数据分组所包含的分组个数分别为M和M”；和/或，第一数据分组所包含的a个分组所对应的数据内容与第二数据分组所包含的b个分组所对应的数据内容不同，可以是完全不同，也可以是部分不同，换言之，a个分组和b个分组中至少存在一个分组对应的数据内容不同。
可选地，第一数据分组和第二数据分组所包含的分组个数可以相同。如果第一数据分组和第二数据分组所包含的分组个数相同，例如M，则第一数据分组所包含的a个分组，与第二数据分组所包含的b个分组中至少存在一个分组对应的数据内容不同。
可选地，本申请对第二数据分组的分组个数，以及每个分组包含的数据大小不作限定，具体的分组方式可参考上述图4至图9的相关描述，为了简洁，此处不作说明。可以理解的，第一重组模式与第二重组模式不同，第一数据分组和第二数据分组也不同，以适应不同的业务传输需求。
例如，假设第一数据类型的数据的长度为N，根据第二重组模式对第一数据类型的数据进行分组得到的第二数据分组的个数为M”，M”为大于1且小于N的整数。例如N＝8，M”＝4，说明存在长度为8的第一类型的数据，第一装置可以将该数据分为4组，则这4组可以包含等量(或等长度、或相同个数)的数据，即每组包含2个数据；或者，这4组可以包含不等量(或不等长度)的数据，如4组分别包含1,1,2,4个数据块，对此不予限定。
S303，第一装置输出第二数据，其中，第二数据是由第二数据分组经过处理得到的。
对应地，第二装置获取第二数据。
上述步骤S301至S303的具体实现方式，可以参考上述步骤S310至S330的相关描述，为了简洁，此处不再赘述。
S304，第二装置对第二数据进行处理得到第二数据分组。
S305，第二装置根据第二重组模式对第二数据分组进行数据恢复，得到针对第一数据类型的重构数据。
上述步骤S304和S305的具体实现方式，可以参考上述步骤S340和S350的相关描述，为了简洁，此处不再赘述。
基于上述方案，第一装置可以确定第一重组模式，然后使用该第一重组模式对第一数据类型的数据进行分组。该实现方式，针对同一数据类型设计多种重组模式，使得第一装置可以根据不同的业务或任务传输需求，灵活选择满足传输需求的第一重组模式进行数据分组，即支持多样化的数据传输需求，保证数据传输性能。同时，第一装置还可以结合具体压缩及传输需求进行重组模式的切换，从而提升端到端性能。另外，第一装置还可以将分组后的数据进行处理，例如数据选择、或数据丢弃、或量化等，支持动态匹配传输资源约束，可以将有限的传输资源用于发送更重要的分组数据，达到较好的信号丢弃效果，用于支持在物理层实现灵活、高效地数据压缩和传输，保证数据传输性能。
上述图3至图9主要针对数据重组方案的相关描述，使得针对特定数据类型的数据，可以灵活选择或切换重组模式，以适应不同的业务传输需求。另外，为了支撑物理层压缩的标准化流程，满足对AI压缩场景下的码率调节需求，实现动态匹配传输资源约束，还可以针对分组后的数据进行压缩或丢弃。下面，结合图10至图15，针对上述步骤S330中第一装置的“该第一数据是由第一数据分组经过处理得到的”所涉及的数据选择方案(或者说，数据丢弃方案)以及量化方案进行说明。
可选地，该第一数据是由第一数据分组经过处理得到的，包括：对M个第一数据分组进行数据选择(或者说数据丢弃)，得到第一数据。即，第一装置可以从M个第一数据分组中筛选或选择M’个第一数据分组，或者说，第一装置丢弃M个第一数据分组中的M-M’个第一数据分组。即该第一数据包括所述M’个第一数据分组。
在一种实现方式中，第一装置根据M个第一数据分组对应的重要性，从M个第一数据分组中选择M’个第一数据分组。
可选地，该M’个第一数据分组对应的重要性高于或等于M个第一数据分组中的其他M-M’个第一数据分组对应的重要性，也就是说，第一装置可以从M个第一数据分组中选择重要的M’个第一数据分组进行发送。
可选地，第一数据分组对应的重要性可以通过计算该第一数据分组包含的数据所对应的能量、或L1范数等重要性指标确定。可选地，重要性指标可以是预定义或预配置的，或者通过信令指示的。
例如，假设M＝4，包括分组1、分组2、分组3和分组4，M’＝2，说明第一装置可以根据分组对应的重要性，从这M个组中选择M’个组进行发送。如果这4个分组对应的重要性依次递减，则第一装置可以选择分组1和分组2发送。
再例如，假设M＝4，包括分组1、分组2、分组3和分组4，M’＝2，说明第一装置可以根据分组对应的重要性，从这M个组中选择M’个组进行发送。如果分组1对应的重要性最高，分组2、分组3和分组4对应的重要相等，则第一装置可以选择分组1和分组2发送，即存在多个分组对应的重要性相等的情况下，顺序选择其中的分组进行发送；或者，第一装置可以从分组2至分组4中随机选择一个分组，例如分组4，将分组1和分组4发送给第二装置，此时第一装置通过信令分组4即可，例如通过bitmap＝“1001”或分组序号指示，例如“11”指示分组4。
图10是本申请实施例提供的分组筛选方案的示意图。如图10所示，假设第一数据类型的数据(例如，AI中间特征数据)的长度为N＝8，根据确定第一重组模式对数据进行分组得到M＝4个第一数据分组，例如包括分组1、分组2、分组3和分组4，假设M’＝2，则说明第一装置根据这4个分组对应的重要性指标Im排序(如使用能量、或L1范数等度量各个分组的重要性)等信息从中筛选M’＝2个分组。假设长度为8个数据的取值分别为xn，n＝1,2,…,8，以能量度量分组的重要性为例，则有Im＝x2m-1 2+x2m 2，m＝1,2,3,4。例如，若I2≥I4≥I3≥I1，则说明分组2、分组4、分组3和分组1对应的重要性依次递减，即第一装置在确定M’＝2的情况下，可以优先发送分组2和分组4对应的数据，即第一数据。
可选地，M’的取值可以通过资源约束确定，例如M’的取值可以是协议预定义或预配置的，或者，M’的取值可以是网络设备通过信令配置的，或者M’的取值可以是终端设备指示的等。
可选地，第一装置还可以向第二装置发送指示信息#A，该指示信息#A用于指示M’个第一数据分组。也就是说，第一装置可以通过发送指示信息#A，向第二装置指示从M个第一数据分组中选择的M’个第一数据分组。
举例来说，该指示信息#A可以包括第一比特位图(bitmap)，其中，第一比特位图指示M’个第一数据分组，第一比特位图的大小为M比特；或者，该指示信息#A可以包括M’个第一数据分组对应的分组序号指示，分组序号指示的大小为比特。以图10为例，第一装置优先发送分组2和分组4对应的数据，在一种示例中，第一装置可以通过分组序号指示其选择的M’＝2个分组，使用个比特开销，表示向上取整，该个比特可以携带在指示信息#A中，例如“00”指示分组1，“01”指示分组2，“10”指示分组3，“11”指示分组4，则可以在指示信息#A中携带“01”和“11”，指示第一装置发送分组2和分组4；或者，在另一种示例中，第一装置可以通过使用bitmap指示其选择的M’＝2个分组，则该bitmap的大小为4比特，可以在指示信息#A中携带bitmap＝‘0101’，指示第一装置发送分组2和分组4。
可选地，第一装置向第二装置指示M’个分组的具体实现方式，可以是协议预定义的，例如，协议指示第一装置与第二装置之间始终使用上述第一种示例或第二种示例中的实现方式指示M’个分组。可选地，第一装置可以采用动态切换的方式向第二装置指示M’个分组。例如，在指示信息#A的头部可以包括第一比特，如1bit，用于指示当前第一装置使用分组序号指示M’个第一数据分组(如比特“1”，此时指示信息#A中携带“1 10 11”)，或者，使用第一比特位图指示M’个第一数据分组(如比特“0”，此时指示信息#A中携带“0 0101”)。
可选地，当M’较大时，采用第一比特位图的方式指示M’个分组的开销更低，反之，采用分组序号的方式指示M’个分组的效果更好。例如，针对第一装置需要标记实际发送的分组的优先级顺序时(如采用不等量化的场景)，则第一装置可以采用分组序号的方式来表示。
可选地，第一装置还可以使用其他的方式指示M’个分组。例如，增加1bit用于指示M’＝M，或M’<M，当M’＝M时，指示信息#A的头部无需额外增加1bit，可以进一步节省指示开销。或者，第一装置可以隐式指示使用第一比特位图或分组序号指示的方式指示M’个第一数据分组，例如通过当前发送的第一数据的数据类型确定，当第一数据的数据类型是信道数据或CSI时，可以理解为：第一装置采用第一比特位图的方式来指示M’个分组；当第一数据的数据类型是感知数据、成像数据、或AI数据时，可以理解为：第一装置采用分组序号指示的方式来指示M’个分组。
可选地，该第一数据是由第一数据分组经过处理得到的，包括：对M个第一数据分组进行量化处理，得到第一数据，可选地，该量化处理可以在数据选择或数据丢弃方案之后执行，即第一装置对M’个第一数据分组进行量化处理得到第一数据。
在一种实现方式中，第一装置可以采用采用标量量化SQ方式或矢量量化VQ方式，其中，基于量化比特长度Q1对M’个第一数据分组中的M1个第一数据分组进行量化处理，得到第一数据，M1为大于0且小于或等于M’的整数，Q1为大于0的整数。
在另一种实现方式中，采用SQ方式或VQ方式，其中，基于量化比特长度Q2对M’个第一数据分组中的M2个第一数据分组进行量化处理，得到第一数据，M2为大于0且小于或等于M’的整数，Q2为大于或等于Q1的整数。
也就是说，假设M1等于M’，则第一装置可以使用SQ的方式，基于量化比特长度Q1对M1＝M’个第一数据分组进行量化处理得到第一数据；或者，第一装置可以使用VQ的方式，基于量化比特长度Q1对M1＝M’个第一数据分组进行量化处理得到第一数据。
也就是说，假设M1小于M’，M1+M2＝M’，则第一装置可以使用SQ的方式，基于量化比特长度Q1对M1个第一数据分组进行量化处理得到第一数据的一部分，以及使用SQ的方式，基于量化比特长度Q2对M2个第一数据分组进行量化处理得到第一数据的另一部分；或者，第一装置可以使用SQ的方式，基于量化比特长度Q1对M1个第一数据分组进行量化处理得到第一数据的一部分，以及使用VQ的方式，基于量化比特长度Q2对M2个第一数据分组进行量化处理得到第一数据的另一部分；或者，第一装置可以使用VQ的方式，基于量化比特长度Q1对M1个第一数据分组进行量化处理得到第一数据的一部分，以及使用VQ的方式，基于量化比特长度Q2对M2个第一数据分组进行量化处理得到第一数据的另一部分；或者，第一装置可以使用VQ的方式，基于量化比特长度Q1对M1个第一数据分组进行量化处理得到第一数据的一部分，以及使用SQ的方式，基于量化比特长度Q2对M2个第一数据分组进行量化处理得到第一数据的另一部分。
可选地，该量化比特长度可以是预定义或预配置的，也可以是根据资源约束自行确定的。
可选地，本申请对M1和M2的数量以及大小关系不作具体限定。例如，M2等于M1，且M1+M2＝M’，例如，第一装置可以将最重要的一半的第一数据分组(例如，M1个第一数据分组)采用量化比特长度为Q2＝Q+1进行量化处理，将剩余的一半的第一数据分组(例如，M2个第一数据分组)采用量化比特长度为Q1＝Q-1进行量化处理，以满足平均每个数据对应的量化比特长度为Q。
可选地，第一装置向第二装置发送指示信息#B，或者，第一装置接收来自第二装置的指示信息#B；其中，指示信息#B用于指示Q，Q表示用于对M’个第一数据分组进行量化的量化比特长度，Q为大于0的整数。例如，指示信息#B用于指示Q1，Q1表示用于对M’个第一数据分组中M1个第一数据分组进行量化的量化比特长度，和/或，指示信息#B用于指示Q2，Q2表示用于对M’个第一数据分组中M2个第一数据分组进行量化的量化比特长度。
在本申请中，量化比特长度可以替换为量化比特数、量化长度、量化比特、或量化分辨率，其中量化分辨率可以表示为SQ量化区间或VQ量化码字的数量，当量化比特长度越长时，对应的量化比特数、量化长度、或量化比特也越多，量化分辨率越高。
作为示例，当Q2等于Q1时，说明M’个第一数据分组采用相同的量化比特长度进行量化处理，即M1个第一数据分组和M2个第一数据分组采用相同的量化分辨率进行量化处理，即M个第一数据分组的量化精度相同，具体示例如下图11所示；当Q2大于Q1时，说明M2个第一数据分组采用的量化比特长度大于M1个第一数据分组采用的量化比特长度，可以理解的，该M2个第一数据分组对应的重要性高于M1个第一数据分组对应的重要性。针对Q2不等于Q1的情况，可以实现对M’个第一数据分组的动态非均等量化，以提升重构数据的整体量化精度。
可选地，第一装置还可以向第二装置发送指示信息#C，该指示信息#C用于指示量化配置，量化配置包括量化方式和量化参数。例如，当量化方式为SQ方式时，量化参数包括量化区间信息；再例如，当量化方式为VQ方式时，量化参数包括VQ码本矩阵信息；又例如，当量化方式为SQ方式和VQ方式时，量化参数包括量化区间信息和VQ码本矩阵信息。
也就是说，本申请对M’个第一数据分组采用的量化方式不作限定，可以是SQ方式和/或VQ方式，对应的，量化参数可以包括量化区间信息和/或VQ码本矩阵信息。
图11是本申请实施例提供的SQ及VQ码本的结构示意图。如图11的(a)所示，采用SQ方式对M’个第一数据分组进行量化处理。示例性地，通过量化区间信息bi或量化重构恢复值yi＝(bi-1+bi)/2表示量化参数，以Q＝3为例，对应的有23+1＝9个量化区间信息(如b0,b1,b2,…,b8)或23＝8个量化恢复值参数(如y1,y2,…,y8)；此时，通过量化获得的二进制数据码流(即，第二数据)表示为Q bit一组(每组对应一个数据)，共计组、比特。如图11的(b)所示，采用VQ方式对M’个第一数据分组进行量化处理。定义矢量长度L，以Q＝3，L＝3为例，对应的量化参数可以表示为L*2QL＝3*512的码本矩阵的取值。此时，通过量化获得的二进制数据码流(即，第二数据)表示为QL bit一组(每组对应L个数据，长度为L)，共计组、比特。
可以理解的，上述图11的(a)和图11的(b)对应的量化方式采用相同的量化精度，仅是为便于描述给出的示例，不排除其他方案，例如采用不同的量化精度对M’个第一数据分组进行量化处理。
可选地，在固定量化比特长度Q的情况下，第一装置可以动态选择合适的量化参数，如通过均方误差(mean-square error，MSE)、或平均绝对值误差(mean absolute error，MAE)等作为评价指标，选择MSE或MAE最小的量化参数用于量化。
可选地，上述量化方式和/或量化参数可以是预定义或预配置的，也可以是通过信令指示的，本申请对此不作限定。
可选地，上述图11的(a)所示的SQ方式，以及图11的(b)所示的VQ方式的示例可以独立实现，也可以组合实现，例如假设存在M’＝4，M1＝2，M2＝2，则第一装置可以采用上述图11的(a)所示的方案对M1＝1个第一数据分组进行量化处理，以及采用上述图11的(a)所示的方案对M2＝3个第一数据分组进行量化处理，具体实现方式可参考上述描述，这里不重复说明。
下面，结合图12和图13针对调度资源固定，以及量化比特长度Q固定或不固定的情况下，确定M’个第一数据分组(和量化比特长度Q)的具体实现方式进行说明。
图12是本申请实施例提供的固定资源约束和量化比特长度的情况下的分组筛选方案的示意图。如图12所示，假设提前固定调度的资源和量化比特长度Q，可以是浮点数(如四分之一精度浮点数fp8、半精度浮点数fp16、或单精度浮点数fp32等)、或定点数等，第一装置可以根据调度的传输资源和量化比特长度，确定实际发送的第一数据分组的个数M’。
示例性地，第一装置根据调度的传输资源(例如，时频资源个数、信道码率、或调制阶数等)，估算出当前能够发送的总比特数B，进一步可以确定当前能够发送的总数据的个数K＝B/Q。其中，N为数据的长度，M为第一数据分组的个数。首先，初始测算实际发送的第一数据分组个数M’，即判断数据的长度N是否小于或等于K，如果是，则确定M’＝M，即第一装置确定M个分组(或者说，长度为N的数据)可以全部发送；反之，第一装置可以根据K计算M’＝floor(K*M/N)，floor()表示下取整函数；然后，第一装置判断第一信息占用的总比特长度是否小于或等于B，如果是，则表示M’个第一数据分组对应的数据全部可以输出，即第一装置向第二装置发送分组筛选信息(即，指示信息#A)，指示M’个第一数据分组；反之，第一装置可以执行微调过程，即尝试减小M’的取值，例如M’＝M’-1，再判断传输的信息包括第一数据和指示信息#A(可选地，还包括指示信息#B和/或指示信息#C)占用的总比特长度是否小于或等于B，依次执行，直到在满足传输资源的约束的情况下，第一装置向第二装置发送分组筛选信息(即，指示信息#A)，指示M’个第一数据分组。
图13是本申请实施例提供的固定资源约束情况下的分组筛选和量化方案的示意图。如图13所示，假设提前固定调度的资源，第一装置可以根据调度的传输资源，确定实际发送的第一数据分组的个数M’和量化比特长度Q。
示例性地，第一装置可以根据调度的传输资源(例如，时频资源个数、信道码率、或调制阶数等)，估算出当前能够发送的总比特数B。假设预定义或预配置最低的量化比特长度Qmin，这是考虑到低于Qmin时AI编解码的性能较差，可不考虑。其中，N为数据的长度，M为第一数据分组的个数。首先，初始测算实际发送的第一数据分组的个数M’和量化比特长度Q，即判断采用最低量化比特长度Qmin时，比特长度N×Qmin是否小于或等于B，如果是，则确定M’＝M，即第一装置确定M个分组(或者说，长度为N的数据)可以全部发送，且实际使用的量化比特长度Q＝floor(B/N)，floor()表示下取整函数；反之，第一装置可以根据B计算M’＝floor(B/Qmin*M/N)，且实际使用的量化比特长度Q＝Qmin；然后，第一装置判断传输的信息包括第一数据和指示信息#A(可选地，还包括指示信息#B和/或指示信息#C)占用的总比特长度是否小于或等于B，如果是，则表示M’个第一数据分组对应的数据全部可以输出，即第一装置向第二装置发送分组筛选信息(即，指示信息#A)和指示信息#B，指示M’个第一数据分组和量化比特长度Q；反之，第一装置可以执行微调过程，即尝试减小M’的取值，例如M’＝M’-1，再判断第一信息占用的总比特长度是否小于或等于B，依次执行，直到在满足传输资源的约束的情况下，第一装置向第二装置发送分组筛选信息(即，指示信息#A)和指示信息#A，指示M’个第一数据分组和量化比特长度Q。
针对上述第一装置对M个第一数据分组进行筛选(或压缩)得到M’个第一数据分组，以及对M’个第一数据分组进行量化处理得到第一数据的方案，对应地，第二装置可以对获取的第一数据进行反量化处理、数据分组恢复以及数据重构操作，得到重构的第一数据类型的数据，具体实现方式如下所示。
可选地，针对上述步骤S340中的第二装置对第一数据进行处理得到第一数据分组，包括：第二装置对第一数据进行反量化处理得到重构的M’个第一数据分组，重构的M’个第一数据分组属于重构的M个第一数据分组，M’为小于或等于M的正整数。其中，本申请针对第一数据进行反量化得到重构的M’个特征分组的实现方式不作限定，可以参考现有方案的相关描述，为了简洁，这里不再说明。
可选地，针对上述步骤S340中的第二装置对第一数据进行处理得到第一数据分组，包括：第二装置对M-M’个第一数据分组进行补零操作得到重构的M个第一数据分组，M为大于1的整数。
作为示例，下面结合图14和图15，分别从编码侧针对数据分组方案、数据选择或压缩方案，以及量化操作等，从解码侧针对解量化、数据分组恢复、数据重构等操作等进行示例说明。
图14是本申请实施例提供的编码单元的处理流程的示意图。如图14所示，该编码单元的处理流程包括但不限于如下步骤。其中，以第一数据类型为AI中间特征数据、第一重组模式为基于顺序拆分的重组模式为例进行说明。
S1：重组AI中间特征数据。
示例性地，编码单元接收长度为N的特征数据之后，重组特征数据。即编码单元将长度为N的特征进行分组，例如根据第一重组模式将N个特征等间隔地分为M个组，分别是数据分组1、数据分组2、数据分组3、…、数据分组M，则每个数据分组的长度为N/M，或者说每个数据分组包含N/M个特征数据。
可选地，针对该特征数据的重组(或分组)方式可以有一种或多种，不同的分组图案表征不同的分组映射关系。
S2：筛选M’个组。
应理解，该M’个组属于M个组，该M’个组的重要程度大于M-M’个组的重要程度，在降低传输开销的同时保证数据传输精度。
示例性地，编码单元根据M个组的重要程度，对步骤S1中的M个组进行排序，并从中选择M’个组。可选地，每个分组的重要程度可以通过计算该分组对应的能量或L1范数进行确定。
S3：量化操作。
示例性地，针对分组筛选得到的M’个分组，编码单元可以根据调度的资源确定量化比特长度Q，进而基于Q对M’个分组进行量化处理得到第一数据，Q为大于或等于1的整数。
S4：输出待发送的信号。
示例性地，待发送的信号包括数据码流(即第一数据)和指示信息#A，其中，指示信息#A表示分组筛选信息(对应S2，即第二指示信息)，可选地，待发送的信号还包括用于指示第一重组模式的重组指示信息#1(对应S1，可参见上述图4)。可选地，待发送的信号还可以包括码本指示信息(对应S3，即指示信息#C)，指示量化配置。
图15是本申请实施例提供的解码单元的处理流程的示意图。如图15所示，该解码单元的处理流程包括如下多个步骤。其中，以第一数据类型为AI中间特征数据、第一重组模式为基于顺序拆分的重组模式为了进行示例说明。
S5：反量化。
示例性地，解码单元根据指示信息#C所指示的量化配置，对数据码流进行反量化操作，得到M’个分组的数据特征。
S6：分组筛选恢复。
示例性地，解码单元对第一装置筛选掉的M-M’个分组进行补零操作，然后根据指示信息#A中携带的分组筛选信息对M个分组恢复顺序，得到重构数据分组，例如重构数据分组1、重构数据分组2、重构数据分组3、…、重构数据分组M。
S7：特征数据恢复。
示例性地，解码单元根据重组指示信息#1，确定针对AI中间特征数据的分组方式，即基于顺序拆分的重组模式，然后对重构数据分组进行反映射操作(即，恢复特征数据顺序)，得到重构的AI中间特征数据(例如，长度为N的矢量)。
基于上述方案，第一装置根据确定的第一重组模式，对第一数据类型的数据进行分组得到M个第一数据分组，并且从M个第一数据分组中筛选M’个第一数据分组(或者说，丢弃M-M’个第一数据分组)后量化处理并输出第一数据，不仅使得第一装置可以根据不同的业务或任务传输需求，灵活选择满足传输需求的第一重组模式进行数据分组，甚至可以为满足不同的业务需求灵活切换重组模式，即支持多样化的数据传输需求，保证数据传输性能。同时，第一装置还可以将M个第一数据分组进行处理，例如数据选择和量化处理等操作，适应不同的AI模型和/或传输场景，支持动态匹配传输资源约束，满足对AI压缩场景下的细粒度的码率调节需求，将有限的传输资源用于发送更重要的特征数据，达到较好的数据丢弃效果，以充分发挥AI模型的端到端性能，用于支持在物理层实现灵活、高效地数据压缩和传输。另外，采用动态优化量化配置的方案，可以进一步提升量化的精度，保障数据传输性能。
应理解，上述各过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
还应理解，本申请将围绕可包括多个设备、组件、模块等的系统来呈现各个方面、实施例或特征。应当理解和明白的是，各个系统可包括另外的设备、组件、模块等，并且/或者可以并不包括结合附图讨论的所有设备、组件、模块等。此外，还可以使用这些方案的组合。
还应理解，在上述一些实施例中，主要以现有的网络架构中的设备为例进行了示例性说明(例如，第一装置或第二装置等)，应理解，对于设备的具体形式本申请实施例不作限定。例如，在未来可以实现同样功能的设备都适用于本申请实施例。
可以理解的是，上述各个方法实施例中，由设备(例如，第一装置或第二装置)实现的方法和操作，也可以由设备的部件(例如，芯片或者电路)实现。
以上，结合图1至图16详细说明了本申请实施例提供的数据传输方法。上述数据传输方法主要从第一装置和第二装置之间交互的角度进行了介绍。可以理解的是，第一装置和第二装置为了实现上述功能，其包含了执行各个功能相应的硬件结构和/或软件模块。
本领域技术人员应该可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，本申请能够以硬件或硬件和计算机软件的结合形式来实现。某个功能究竟以硬件还是计算机软件驱动硬件的方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
以下，结合图16至图19详细说明本申请实施例提供的通信装置。装置实施例的描述与方法实施例的描述相互对应，因此，未详细描述的内容可以参见上文方法实施例，为了简洁，部分内容不作说明。
本申请实施例可以根据上述方法示例对通信装置进行功能模块的划分，例如，可以对应各个功能划分各个功能模块，也可以将两个或两个以上的功能集成在一个处理模块中。上述集成的模块既可以采用硬件的形式实现，也可以采用软件功能模块的形式实现，或者软件和硬件结合。本申请实施例中对模块的划分是示意性的，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式。下面以采用对应各个功能划分各个功能模块为例进行说明。
图16是本申请实施例提供的通信装置的示例性框图。如图16所示，该通信装置1000可以包括芯片系统1100、存储器1200、总线(bus)1300、电源管理模块1400、或收发器1500等。
其中，芯片系统1100可以是集成电路芯片，具有信号的处理能力。在实现过程中，上述方法的各步骤可以通过芯片系统1100中的硬件的集成逻辑电路或者软件形式的指令完成。
作为示例而非限定，芯片系统1100可以包括负责信号的处理的电路或芯片(如调制解调(modem)芯片，又称基带(baseband)芯片，或包含modem核的片上系统SoC芯片或SIP芯片)。
可选地，芯片系统1100中还可以设置存储器(如cache)，用于存储指令和数据。在一些实施例中，芯片系统1100中的存储器为高速缓冲存储器。该存储器可以保存芯片系统1100刚用过或循环使用的指令或数据。如果芯片系统1100需要再次使用该指令或数据，可从所述存储器中直接调用。避免了重复存取，减少了芯片系统1100的等待时间，因而提高了系统的效率。
在一些实施例中，芯片系统1100可以包括一个或多个接口。接口可以包括集成电路(inter-integrated circuit，I2C)接口，集成电路内置音频(inter-integrated circuitsound，I2S)接口，脉冲编码调制(pulse code modulation，PCM)接口，通用异步收发传输器(universal asynchronous receiver/transmitter，UART)接口，移动产业处理器接口(mobile industry processor interface，MIPI)，通用输入输出(general-purposeinput/output，GPIO)接口，用户身份模块(subscriber identity moudle，SIM)接口，和/或通用串行总线(universal serial bus，USB)接口等。
存储器1200可以包括随机存取存储器(random access memory，RAM)和只读存储器(read-only memory，ROM)。存储器1200可以存储计算机可读的计算机可执行的代码，该代码包括指令，这些指令在被执行时使处理器执行本申请所述的各种功能。
可选地，代码可以包括用于实现本申请是实施例的各方面的指令，其中包括用于支持生成第一信息或解析第一信息的指令。代码可以存储在诸如系统存储器或其他类型的存储器之类的非暂时性计算机可读介质中。在一些情况下，代码可能不能直接由芯片系统1100执行，而是可以使计算机(例如，在编译和执行时)执行本申请所述的功能。在一些情况下，存储器1200尤其可以包含基本输入/出(input/output，I/O)系统，其可以控制基本硬件或软件操作，例如与外围组件或设备的交互。
示例性地，芯片系统1100通过运行存储在存储器1200的指令，从而执行通信装置1000的各种功能应用以及数据处理。例如，通信装置1000与其他设备(例如，终端设备、或者网络设备、或者核心网设备)进行文件传输时，通信装置1000的芯片系统1100可以调用存储器1200存储的计算机可执行程序代码，实现本申请实施例提供的数据和/或信令传输方法。
另外，存储器1200可以集成在上述芯片系统1100中，或者独立于芯片系统1100。
总线1300可以是USB，用于支持通信装置1000中各个部分之间的相互通信。
电源管理模块1400用于从充电器接收充电输入。可选地，电源管理模块1400为通信装置1000(如，通信装置1000的电池模块)充电的同时，还可以为通信装置1000供电。作为示例而非限定，该电源管理模块1400还可以为除该通信装置1000之外的其他装置供电。
收发器1500可以经由一个或多个天线、有线链路或无线链路双向地通信，例如，收发器1500可以表示无线收发器，并且可以与另一个无线收发器双向通信。收发器1500还可以包括调制解调器，用于调制分组并将调制后的分组提供给天线以进行传输，并且还用于解调从天线接收的分组。其中，收发器1500可以包括接收器和发送器，接收器实现接收信息的功能，发送器实现发送信息的功能。
在一些情况下，无线设备可以包括单个天线。然而，在一些情况下，设备可以具有一个以上的天线，如图16中所示的天线1和天线2，其可能能够同时发送或接收多个无线传输。示例性地，天线1和天线2用于发射和接收电磁波信号。通信装置1000中的每个天线可用于覆盖单个或多个通信频带。不同的天线还可以复用，以提高天线的利用率。例如：可以将天线1复用为无线局域网的分集天线。在另外一些实施例中，天线可以和调谐开关结合使用。通信装置1000可以通过无线通信功能向其他设备传输文件。
在一种设计中，该通信装置1000可对应于上文方法实施例中的第一装置。
该装置1000可实现对应于上文方法实施例中的第一装置执行的步骤或者流程，其中，收发器1500可用于执行上文方法实施例中第一装置的收发相关的操作；芯片系统1100可用于执行上文方法实施例中第一装置的处理相关的操作。
在另一种设计中，该通信装置1000可对应于上文方法实施例中的第二装置。
该装置1000可实现对应于上文方法实施例中的第二装置执行的步骤或者流程，其中，收发器1500可用于执行上文方法实施例中第二装置的收发相关的操作；芯片系统1100可用于执行上文方法实施例中第二装置的处理相关的操作。
在该设计下，通信装置1000可能包括如图16中所示短距通信模块1640、传感器1610、显示器1620、或摄像头1630等模块。
其中，短距通信模块1640可以包括无线网络(WI-FI，或WIFI)，或者蓝牙等支持短距通信的模块。
传感器1610可以包括压力传感器、陀螺仪传感器、气压传感器、磁传感器、加速度传感器、距离传感器、接近光传感器、指纹传感器、温度传感器、触摸传感器、环境光传感器、骨传导传感器等。
显示器1620用于显示图像、视频等。显示器包括显示面板。显示面板可以采用液晶显示屏(liquid crystal display，LCD)，有机发光二极管(organic light-emittingdiode，OLED)，有源矩阵有机发光二极体或主动矩阵有机发光二极体(active-matrixorganic light emitting diode，AMOLED)，柔性发光二极管(flex light-emittingdiode，FLED)，Miniled，MicroLed，Micro-oLed，量子点发光二极管(quantum dot lightemitting diodes，QLED)等。例如，本申请实施例中，显示器可以用于显示通信装置1000所需显示的界面。示例性地，通信装置1000通过图形处理器(graphic processing unit，GPU)，显示器，以及应用处理器等实现显示功能。GPU为图像处理的微处理器，连接显示器和应用处理器。GPU用于执行数学和几何计算，用于图形渲染。芯片系统1100可包括一个或多个GPU，其执行程序指令以生成或改变显示信息。
摄像头1630用于获取图像、视频等。
可以理解的是，图16所示的结构并不构成对通信装置1000的具体限定，终端设备和/或网络设备的具体结构可以参考图16所示。在一些实施例中，通信装置1000也可以包括比图16所示更多或更少的部件，或者组合某些部件，或者拆分某些部件，或者不同的部件布置等。又或者，图16所示的一些部件可以以硬件，软件，或软件和硬件的组合实现，终端设备和/或网络设备可以是在图16给出的结构的基础上增加或减少了组件。
图17是本申请实施例提供的通信装置的示意性框图。如图17所示，通信装置2000可以包括基带单元2100，该基带单元2100可通过蜂窝RF收发机2200与外部设备通信(如，通信装置2000为终端设备，则基带单元2100可通过蜂窝RF收发机2200与网络设备通信；还如，通信装置200为网络设备，则基带单元2100可通过蜂窝RF收发机2200与终端设备和/或核心网设备通信)。
基带单元2100可包括计算机可读介质/存储器。基带单元2100负责一般性处理，包括对存储在计算机可读介质/存储器上的软件的执行。该软件在由基带单元2100执行时使该基带单元2100执行上文所描述的各种功能。计算机可读介质/存储器还可被用于存储由基带单元2100在执行软件时操纵的数据。
基带单元2100进一步包括接收单元2010、管理单元2020和发送单元2030。管理单元2020包括该一个或多个图17中所示的子单元(如，编码单元和/或解码单元，其中，编码单元可以参考上述图12的相关描述，例如用于执行上述方法实施例400中的步骤S410至步骤S430，解码单元可以参考上述图13的相关描述，例如用于执行上述方法实施例400中的步骤S430至步骤S470)。管理单元2020内的单元可被存储在计算机可读介质/存储器中和/或配置为基带单元2100内的硬件。其中，接收单元2010和发送单元2030可以称为收发单元。
当通信装置2000用于实现上述各方法实施例中第一装置的功能时，接收单元2010用于执行第一装置的接收步骤，发送单元2030用于执行第一装置的发送步骤，管理单元2020用于执行第一装置的处理步骤。
例如，当该装置2000用于执行图3中的方法时，接收单元2010可用于执行方法中的接收信息的步骤；管理单元2020可用于执行方法中的处理步骤；发送单元2030可用于执行方法中的发送信息的步骤。
当通信装置2000用于实现上述各方法实施例中第二装置的功能时，接收单元2010用于执行第二装置的接收步骤，发送单元2030用于执行第二装置的发送步骤，管理单元2020用于执行第二装置的处理步骤。
例如，当该装置2000用于执行图3中的方法时，接收单元2010可用于执行方法中的接收信息的步骤；管理单元2020可用于执行方法中的处理步骤；发送单元2030可用于执行方法中的发送信息的步骤。
关于上述接收单元2010、管理单元2020和发送单元2030更详细的描述，可参考上述方法实施例中的相关描述，在此不再说明。
图18是本申请实施例提供的一种芯片系统3000的示意性框图。示例性地，该芯片系统包括但不限于：调制解调(modem)芯片，又称基带(baseband)芯片，或包含modem核的片上系统SoC芯片或系统级封装SIP芯片。
如图18所示，该芯片系统(或者也可以称为处理系统)包括处理器3100、存储器3200以及输入/输出接口(input/output interface)3300。
其中，处理器3100可以为芯片系统中的处理电路(包括至少一个处理器，如图18中所示的处理器1和处理器2等)。处理器3100可以耦合连接存储器3200，调用存储器3200中的指令，使得芯片系统可以实现本申请各实施例的方法和功能。输入/输出接口3300，可以为芯片系统中的输入输出电路，将芯片系统处理好的信息输出，或将待处理的数据或信令信息输入芯片系统进行处理。
作为一种方案，该芯片系统用于实现上文各个方法实施例中由第一装置或第二装置执行的操作。
例如，处理器3100用于实现上文方法实施例中由第一装置或第二装置执行的处理相关的操作，具体可以参考前述实施例中的描述；输入/输出接口3300用于实现上文方法实施例中由第一装置或第二装置执行的发送和/或接收相关的操作，具体可以参考前述实施例中的描述。
图19是本申请实施例提供的另一种芯片系统4000的示意性框图。如图19所示，该芯片系统(或者也可以称为处理系统)包括输入/输出接口4100和逻辑电路4200。其中，输入/输出接口4100，可以为芯片系统中的输入输出电路，将芯片系统处理好的信息输出，或将待处理的数据或信令信息输入芯片系统进行处理，具体可以参考前述实施例中的描述；逻辑电路4200用于执行上述的数据传输方法，具体可以参考前述实施例中的描述。
作为一种方案，该芯片系统用于实现上文各个方法实施例中由第一装置或第二装置执行的操作。
例如，逻辑电路4200用于实现上文方法实施例中由第一装置或第二装置执行的处理相关的操作；输入/输出接口4100用于实现上文方法实施例中由第一装置或第二装置执行的发送和/或接收相关的操作。
本申请实施例还提供一种计算机可读存储介质，其上存储有用于实现上述各方法实施例中由装置执行的方法的计算机指令。
例如，该计算机程序被计算机执行时，使得该计算机可以实现上述方法各实施例中由第一装置或第二装置执行的方法。
本申请实施例还提供一种计算机程序产品，包含指令，该指令被计算机执行时以实现上述各方法实施例中由第一装置或第二装置执行的方法。
本申请实施例还提供了一种通信系统，包括前述的第一装置和/或第二装置。
上述提供的任一种装置中相关内容的解释及有益效果均可参考上文提供的对应的方法实施例，此处不作说明。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再说明。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、ROM、RAM、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以所述权利要求的保护范围为准。

Claims (31)

  1. 一种数据传输方法，其特征在于，所述方法应用于第一装置，包括：
    确定第一重组模式，所述第一重组模式用于指示针对第一数据类型的数据的分组方式，所述第一重组模式属于P个重组模式，所述P个重组模式与所述第一数据类型对应，P为大于1的整数；
    根据所述第一重组模式对所述第一数据类型的数据进行分组，得到第一数据分组；
    输出第一数据，所述第一数据是由所述第一数据分组经过处理得到的。
  2. 根据权利要求1所述的方法，其特征在于，所述方法还包括：
    获取或输出第一指示信息，所述第一指示信息用于指示第一映射关系，所述第一映射关系用于指示K个重组模式与L个数据类型之间的对应关系，所述K个重组模式包括所述第一重组模式，所述L个数据类型包括所述第一数据类型，所述K个重组模式包含所述P个重组模式，P为小于或等于K的整数，L为大于或等于1的整数。
  3. 根据权利要求2所述的方法，其特征在于，所述确定第一重组模式，包括：
    根据所述第一数据类型和所述第一映射关系，确定所述第一重组模式。
  4. 根据权利要求1至3中任一项所述的方法，其特征在于，所述方法还包括：
    获取或输出第二指示信息，所述第二指示信息用于指示所述第一重组模式。
  5. 根据权利要求4所述的方法，其特征在于，在获取第二指示信息之前，所述方法还包括：
    发送第三指示信息，所述第三指示信息用于指示所述第一装置的能力信息，或者，所述第一装置支持的一个或多个重组模式，其中，所述一个或多个重组模式包括所述第一重组模式。
  6. 根据权利要求1至5中任一项所述的方法，其特征在于，所述重组模式包括以下一项或多项：
    基于顺序拆分的重组模式；
    基于比特位置的重组模式；
    基于空间位置的重组模式；
    基于语义信息的重组模式；或者，
    基于重要性排序的重组模式。
  7. 根据权利要求1至6中任一项所述的方法，其特征在于，所述第一数据类型包括以下一项：
    环境反射点、环境面片、环境重构地图、射频地图、人工智能AI训练数据、AI推理数据、AI模型参数、AI梯度参数、AI中间特征数据、AI性能数据、信道矩阵、信道状态信息CSI矩阵、或信道指示信息。
  8. 根据权利要求1至7中任一项所述的方法，其特征在于，当所述第一数据类型的数据为感知数据时，
    若所述第一数据类型为环境反射点，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式；
    若所述第一数据类型为环境面片，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，或者，基于语义信息的重组模式；
    若所述第一数据类型为环境重构地图，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式；
    若所述第一数据类型为射频地图，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于重要性排序的重组模式。
  9. 根据权利要求1至8中任一项所述的方法，其特征在于，当所述第一数据类型的数据为AI数据时，
    若所述第一数据类型为AI训练数据或AI推理数据，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于语义信息的重组模式，或者，基于重要性排序的重组模式；
    若所述第一数据类型为AI模型参数或AI梯度参数，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；
    若所述第一数据类型为AI中间特征数据，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于语义信息的重组模式；
    若所述第一数据类型为AI性能数据，则所述第一重组模式包括基于顺序拆分的重组模式。
  10. 根据权利要求1至9中任一项所述的方法，其特征在于，当所述第一数据类型的数据为信道数据时，
    若所述第一数据类型为信道矩阵或CSI矩阵，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；
    若所述第一数据类型为信道指示信息，则所述第一重组模式包括基于顺序拆分的重组模式。
  11. 根据权利要求1至10中任一项所述的方法，其特征在于，所述方法还包括：
    获取第四指示信息，所述第四指示信息用于指示第二重组模式，所述第二重组模式用于指示针对所述第一数据类型的数据的分组方式；
    根据所述第二重组模式对所述第一数据类型的数据进行分组，得到第二数据分组，所述第二数据分组与所述第一数据分组不同；
    输出第二数据，所述第二数据是由所述第二数据分组经过处理得到的。
  12. 根据权利要求11所述的方法，其特征在于，在所述获取第四指示信息之前，所述方法还包括：
    获取第五指示信息，所述第五指示信息用于指示第二映射关系，所述第二映射关系用于指示K’个重组模式与L’个数据类型之间的对应关系，所述K’个重组模式包括所述第二重组模式，所述L’个数据类型包括所述第一数据类型，K’为大于1的整数，L’为大于或等于1的整数。
  13. 根据权利要求11所述的方法，其特征在于，在所述获取第四指示信息之前，所述方法还包括：
    获取第六指示信息，所述第六指示信息用于指示所述第一数据类型对应的Q个重组模式，所述Q个重组模式包括所述第二重组模式，Q为大于或等于2的整数。
  14. 一种数据传输方法，其特征在于，所述方法应用于第二装置，包括：
    获取第一数据；
    对所述第一数据进行处理得到第一数据分组；
    根据第一重组模式对所述第一数据分组进行数据恢复，得到针对第一数据类型的重构数据；
    其中，所述第一重组模式用于指示针对所述第一数据类型的数据的分组方式，所述第一重组模式属于P个重组模式，所述P个重组模式与所述第一数据类型对应，P为大于1的整数。
  15. 根据权利要求14所述的方法，其特征在于，所述方法还包括：
    获取或输出第一指示信息，所述第一指示信息用于指示第一映射关系，所述第一映射关系用于指示K个重组模式与L个数据类型之间的对应关系，所述K个重组模式包括所述第一重组模式，所述L个数据类型包括所述第一数据类型，所述K个重组模式包含所述P个重组模式，P为小于或等于K的整数，L为大于或等于1的整数。
  16. 根据权利要求15所述的方法，其特征在于，所述第一重组模式是根据所述第一数据类型和所述第一映射关系确定的。
  17. 根据权利要求14至16中任一项所述的方法，其特征在于，所述方法还包括：
    获取或输出第二指示信息，所述第二指示信息用于指示所述第一重组模式。
  18. 根据权利要求17所述的方法，其特征在于，在获取第二指示信息之前，所述方法还包括：
    接收第三指示信息，所述第三指示信息用于指示所述第一装置的能力信息，或者，所述第一装置支持的一个或多个重组模式，其中，所述一个或多个重组模式包括所述第一重组模式。
  19. 根据权利要求14至18中任一项所述的方法，其特征在于，所述重组模式包括以下一项或多项：
    基于顺序拆分的重组模式；
    基于比特位置的重组模式；
    基于空间位置的重组模式；
    基于语义信息的重组模式；或者，
    基于重要性排序的重组模式。
  20. 根据权利要求14至19中任一项所述的方法，其特征在于，所述第一数据类型包括以下一项：
    环境反射点、环境面片、环境重构地图、射频地图、人工智能AI训练数据、AI推理数据、AI模型参数、AI梯度参数、AI中间特征数据、AI性能数据、信道矩阵、信道状态信息CSI矩阵、或信道指示信息。
  21. 根据权利要求14至20中任一项所述的方法，其特征在于，当所述第一数据类型的数据为感知数据时，
    若所述第一数据类型为环境反射点，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式；
    若所述第一数据类型为环境面片，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，或者，基于语义信息的重组模式；
    若所述第一数据类型为环境重构地图，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于语义信息的重组模式；
    若所述第一数据类型为射频地图，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于空间位置的重组模式，或者，基于重要性排序的重组模式。
  22. 根据权利要求14至21中任一项所述的方法，其特征在于，当所述第一数据类型的数据为AI数据时，
    若所述第一数据类型为AI训练数据或AI推理数据，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，基于语义信息的重组模式，或者，基于重要性排序的重组模式；
    若所述第一数据类型为AI模型参数或AI梯度参数，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；
    若所述第一数据类型为AI中间特征数据，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于语义信息的重组模式；
    若所述第一数据类型为AI性能数据，则所述第一重组模式包括基于顺序拆分的重组模式。
  23. 根据权利要求14至22中任一项所述的方法，其特征在于，当所述第一数据类型的数据为信道数据时，
    若所述第一数据类型为信道矩阵或CSI矩阵，则所述第一重组模式包括以下一项：基于顺序拆分的重组模式，基于比特位置的重组模式，或者，基于重要性排序的重组模式；
    若所述第一数据类型为信道指示信息，则所述第一重组模式包括基于顺序拆分的重组模式。
  24. 根据权利要求14至23中任一项所述的方法，其特征在于，所述方法还包括：
    获取第二数据；
    对所述第二数据进行处理得到第二数据分组；
    根据第二重组模式对所述第二数据分组进行数据恢复，得到针对所述第一数据类型的重构数据；
    其中，所述第二重组模式用于指示针对所述第一数据类型的数据的分组方式。
  25. 根据权利要求24所述的方法，其特征在于，在所述获取第二数据之前，所述方法还包括：
    发送第五指示信息，所述第五指示信息用于指示第二映射关系，所述第二映射关系用于指示K’个重组模式与L’个数据类型之间的对应关系，所述K’个重组模式包括所述第二重组模式，所述L’个数据类型包括所述第一数据类型，K’为大于1的整数，L’为大于或等于1的整数。
  26. 根据权利要求24所述的方法，其特征在于，在所述获取第二数据之前，所述方法还包括：
    发送第六指示信息，所述第六指示信息用于指示所述第一数据类型对应的Q个重组模式，所述Q个重组模式包括所述第二重组模式，Q为大于或等于2的整数。
  27. 一种通信装置，其特征在于，包括用于执行权利要求1至13中任一项的方法的模块或单元，或者用于执行权利要求14至26中任一项的方法的模块或单元。
  28. 一种通信装置，其特征在于，包括至少一个处理器，所述至少一个处理器用于执行计算机程序或指令，以使得如执行权利要求1至13中任一项所述的方法被执行，或者，以使得如执行权利要求14至26中任一项所述的方法被执行。
  29. 根据权利要求28所述的通信装置，其特征在于，
    所述通信装置还包括存储器，所述存储器用于存储所述计算机程序或指令；和/或，
    所述通信装置还包括通信接口，所述通信接口与所述至少一个处理器耦合，所述通信接口用于输入和/或输出信息。
  30. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质用于存储计算机程序，当所述计算机程序在计算机上运行时，以使得如权利要求1至26中任一项所述的方法被执行。
  31. 一种计算机程序产品，其特征在于，包括计算机程序或指令，当所述计算机程序或指令被处理器执行时，以使得如权利要求1至26中任一项所述的方法被执行。
PCT/CN2025/103299 2024-06-29 2025-06-25 数据传输方法、装置和系统 Pending WO2026002004A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202410874141.X 2024-06-29
CN202410874141.XA CN121239658A (zh) 2024-06-29 2024-06-29 数据传输方法、装置和系统

Publications (1)

Publication Number Publication Date
WO2026002004A1 true WO2026002004A1 (zh) 2026-01-02

Family

ID=98143837

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2025/103299 Pending WO2026002004A1 (zh) 2024-06-29 2025-06-25 数据传输方法、装置和系统

Country Status (2)

Country Link
CN (1) CN121239658A (zh)
WO (1) WO2026002004A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1398094A (zh) * 2001-07-12 2003-02-19 Lg电子株式会社 数据通信方法
US6853619B1 (en) * 1999-02-26 2005-02-08 Ipanema Technologies System and method for measuring the transfer durations and loss rates in high volume telecommunication networks
CN111050353A (zh) * 2018-10-11 2020-04-21 华为技术有限公司 用于无线回传网络的数据传输方法和装置
US20220244862A1 (en) * 2021-01-29 2022-08-04 EMC IP Holding Company LLC Method, device, and program product for data rebuilding

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6853619B1 (en) * 1999-02-26 2005-02-08 Ipanema Technologies System and method for measuring the transfer durations and loss rates in high volume telecommunication networks
CN1398094A (zh) * 2001-07-12 2003-02-19 Lg电子株式会社 数据通信方法
CN111050353A (zh) * 2018-10-11 2020-04-21 华为技术有限公司 用于无线回传网络的数据传输方法和装置
US20220244862A1 (en) * 2021-01-29 2022-08-04 EMC IP Holding Company LLC Method, device, and program product for data rebuilding

Also Published As

Publication number Publication date
CN121239658A (zh) 2025-12-30

Similar Documents

Publication Publication Date Title
TWI692261B (zh) 資料傳輸方法、基地台和終端
CN110535509B (zh) 一种信息反馈和接收方法、装置和存储介质
WO2026002004A1 (zh) 数据传输方法、装置和系统
WO2023236803A1 (zh) 信道信息的传输方法、通信装置及通信设备
WO2025227755A1 (zh) 数据传输方法、装置和系统
WO2026016896A1 (zh) 数据传输方法、装置和系统
WO2025241963A1 (zh) 通信方法和通信装置
WO2026036725A1 (zh) 一种数据传输方法及装置
WO2026016904A1 (zh) 数据传输方法、装置和系统
WO2026085767A1 (en) System and methods for fine-tuning of foundational models in communication networks
US20250330841A1 (en) Wireless communication methods and communication devices
WO2026077401A1 (zh) 通信方法和通信装置
CN121924524A (zh) 通信方法、终端设备和网络设备
WO2025130582A1 (zh) 一种传输数据的方法和通信装置
WO2026026573A1 (zh) 通信方法及装置
CN121056062A (zh) 通信方法和通信装置
WO2025232661A1 (zh) 一种编码方法、译码方法及通信装置
WO2025103013A1 (zh) 通信方法及装置
WO2025200940A1 (zh) 通信方法、装置和系统
WO2025201026A1 (zh) 一种通信方法及装置
WO2026012168A1 (zh) 通信方法和通信装置
WO2026007602A1 (zh) 一种通信方法及装置
WO2026067252A1 (zh) 通信方法和装置
WO2025247079A1 (zh) 通信方法和装置
CN121924523A (zh) 通信方法、终端设备和网络设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 25825936

Country of ref document: EP

Kind code of ref document: A1