WO2021203437A1 - 资源配置方法、装置、设备及存储介质 - Google Patents

资源配置方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2021203437A1
WO2021203437A1 PCT/CN2020/084306 CN2020084306W WO2021203437A1 WO 2021203437 A1 WO2021203437 A1 WO 2021203437A1 CN 2020084306 W CN2020084306 W CN 2020084306W WO 2021203437 A1 WO2021203437 A1 WO 2021203437A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
resource configuration
terminal device
resource
model
Prior art date
Application number
PCT/CN2020/084306
Other languages
English (en)
French (fr)
Inventor
沈嘉
Original Assignee
Oppo广东移动通信有限公司
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 Oppo广东移动通信有限公司 filed Critical Oppo广东移动通信有限公司
Priority to EP20930151.4A priority Critical patent/EP4135438A4/en
Priority to CN202080099659.0A priority patent/CN115380594A/zh
Priority to PCT/CN2020/084306 priority patent/WO2021203437A1/zh
Publication of WO2021203437A1 publication Critical patent/WO2021203437A1/zh
Priority to US17/961,964 priority patent/US20230038071A1/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/02—Selection of wireless resources by user or terminal
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/20—Control channels or signalling for resource management
    • H04W72/23—Control channels or signalling for resource management in the downlink direction of a wireless link, i.e. towards a terminal
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/098—Distributed learning, e.g. federated learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/94—Hardware or software architectures specially adapted for image or video understanding
    • G06V10/95—Hardware or software architectures specially adapted for image or video understanding structured as a network, e.g. client-server architectures
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/04—Wireless resource allocation
    • H04W72/044—Wireless resource allocation based on the type of the allocated resource
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W80/00—Wireless network protocols or protocol adaptations to wireless operation
    • H04W80/02—Data link layer protocols

Definitions

  • the embodiments of the present application relate to the field of communication technologies, and in particular, to a resource configuration method, device, device, and storage medium.
  • AI Artificial Intelligence
  • AR Augmented Reality
  • VR Virtual Reality
  • 3GPP (3rd Generation Partnership Project) proposes three application scenarios for the combined application of 5G (5th Generation Mobile Networks) and AI, which are: segmentation in 5G systems Splitting AI operation, downloading of AI models in 5G systems, and training of AI models in 5G systems.
  • segmentation in 5G systems Splitting AI operation
  • downloading of AI models in 5G systems downloading of AI models in 5G systems
  • training of AI models in 5G systems training of AI models in 5G systems.
  • “segmented AI operation” means that the terminal device completes the time-sensitive, privacy-sensitive and computationally small part of the AI operation, and reports the intermediate data (intermediate data) to the network device, and the network device completes the rest.
  • AI model download refers to when the terminal device is in a mobile environment, facing different AI tasks and experiencing different AI working environments, different AI needs to be used Model, if the terminal device does not have the required model, you need to download a new model from the network device to use;
  • AI model training means that the global model for training needs to be allocated to the terminal device through the network device during the model training process , And then report the local gradient after the terminal device training to the network device, and the network device merges the local models of the terminal device to form a more optimized global model.
  • the above three application scenarios all need to use the wireless resources of the network device to transmit AI resources, including: “Segmented AI operation in 5G system” requires terminal equipment to upload intermediate results; “Download of AI model in 5G system” requires The terminal device downloads the AI model; “training of the AI/ML model in the 5G system” requires the terminal device to download the global model and upload the gradient. Therefore, for the above three application scenarios, how to configure wireless resources and AI resources to ensure the normal transmission of AI resources by terminal devices requires further discussion and research.
  • the embodiments of the present application provide a resource configuration method, device, equipment, and storage medium.
  • the technical solution is as follows:
  • an embodiment of the present application provides a resource configuration method, which is applied to a terminal device, and the method includes:
  • first resource configuration information from a network device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first type resource configuration information and second type resource configuration information, where n is Positive integer
  • the first type of resource configuration information is used to indicate wireless resource configuration
  • the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration
  • an embodiment of the present application provides a resource configuration method, which is applied to a network device, and the method includes:
  • the terminal device send first resource configuration information to the terminal device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first type resource configuration information and second type resource configuration information, where n is a positive integer ;
  • the first type of resource configuration information is used to indicate wireless resource configuration
  • the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration
  • an embodiment of the present application provides a resource configuration device, which is applied to a terminal device, and the device includes:
  • the configuration information receiving module is configured to receive first resource configuration information from a network device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first-type resource configuration information and second-type resources Configuration information, where n is a positive integer;
  • the first type of resource configuration information is used to indicate wireless resource configuration
  • the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration
  • an embodiment of the present application provides a resource configuration device, which is applied to a network device, and the device includes:
  • a configuration information sending module configured to send first resource configuration information to a terminal device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first type resource configuration information and second type resource configuration information ,
  • the n is a positive integer
  • the first type of resource configuration information is used to indicate wireless resource configuration
  • the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration
  • an embodiment of the present application provides a terminal device, the terminal device including a processor and a transceiver connected to the processor; wherein:
  • the transceiver is configured to receive first resource configuration information from a network device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first-type resource configuration information and second-type resources Configuration information, where n is a positive integer;
  • the first type of resource configuration information is used to indicate wireless resource configuration
  • the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration
  • an embodiment of the present application provides a network device, the network device including a processor and a transceiver connected to the processor; wherein:
  • the transceiver is configured to send first resource configuration information to a terminal device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first type resource configuration information and second type resource configuration information ,
  • the n is a positive integer
  • the first type of resource configuration information is used to indicate wireless resource configuration
  • the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration
  • an embodiment of the present application provides a computer-readable storage medium in which a computer program is stored, and the computer program is used to be executed by a processor of a terminal device to implement the above-mentioned resources on the terminal device side. Configuration method.
  • an embodiment of the present application provides a computer-readable storage medium in which a computer program is stored, and the computer program is used to be executed by a processor of a network device to implement the resources on the network device side as described above. Configuration method.
  • an embodiment of the present application provides a chip that includes a programmable logic circuit and/or program instructions.
  • the chip runs on a terminal device, it is used to implement resource configuration on the terminal device side as described above. method.
  • an embodiment of the present application provides a chip that includes a programmable logic circuit and/or program instructions.
  • the chip runs on a network device, it is used to implement resource configuration on the network device side as described above. method.
  • the network device sends resource configuration information to the terminal device.
  • the resource configuration information includes multiple resource configuration combinations, and each resource configuration combination includes multiple resource configuration information, thereby providing multiple resource combination configuration methods.
  • the embodiment of the present application configures multiple resource combinations, so that a network device can schedule multiple resources once resource scheduling is performed. Compared with the separate scheduling of multiple resources, multiple resource scheduling is required.
  • the combination provided by the embodiment of the present application The configured solution can reduce the number of resource scheduling performed by the network device, reduce the processing overhead of the network device, and save data transmission resources.
  • each resource configuration combination can include two types of resource configuration information.
  • One type of resource configuration information can be used to indicate wireless resource configuration, and the other type of resource configuration information can be used to indicate AI resource configuration, thereby achieving
  • the combined configuration between wireless resources and AI resources is compared with the problem that wireless resources and AI resources may not be compatible with each other that may occur when wireless resources and AI resources are separately configured.
  • the resource and AI resource configuration together form a resource configuration combination, which ensures the sufficient wireless resources and the quality of AI services at the same time, improves the utilization of wireless resources, and avoids waste of wireless resources or insufficient wireless resources for data interaction when separate configuration occurs. The situation also improves the reliability of AI services, and avoids wasting AI resources or insufficient AI resources to perform AI operations during separate configuration.
  • FIG. 1 is a schematic diagram of a network architecture provided by an embodiment of the present application.
  • Figure 2 is a schematic diagram of the combination of AI services and 5G services provided by an embodiment of the present application
  • Fig. 3 is a flowchart of a resource configuration method provided by an embodiment of the present application.
  • FIG. 4 is a flowchart of a resource configuration method provided by another embodiment of the present application.
  • Figure 5 is a schematic diagram of a resource configuration combination provided by an embodiment of the present application.
  • Fig. 6 is a flowchart of a resource configuration method corresponding to Fig. 5;
  • FIG. 7 is a schematic diagram of a resource configuration combination provided by another embodiment of the present application.
  • FIG. 8 is a schematic diagram of a resource configuration method corresponding to FIG. 7;
  • Fig. 9 is a flowchart of a resource configuration method corresponding to Fig. 7;
  • FIG. 10 is a schematic diagram of a resource configuration combination provided by another embodiment of the present application.
  • FIG. 11 is a schematic diagram of a resource configuration method corresponding to FIG. 10;
  • Fig. 12 is a flowchart of a resource configuration method corresponding to Fig. 10;
  • FIG. 13 is a schematic diagram of a resource configuration method provided by an embodiment of the present application.
  • Fig. 14 is a flowchart of a resource configuration method corresponding to Fig. 13;
  • FIG. 15 is a schematic diagram of a resource configuration method provided by another embodiment of the present application.
  • FIG. 16 is a flowchart of a resource configuration method corresponding to FIG. 15;
  • FIG. 17 is a block diagram of a resource configuration device provided by an embodiment of the present application.
  • FIG. 18 is a block diagram of a resource configuration device provided by another embodiment of the present application.
  • FIG. 19 is a block diagram of a resource configuration device provided by still another embodiment of the present application.
  • FIG. 20 is a block diagram of a resource configuration device according to another embodiment of the present application.
  • FIG. 21 is a structural block diagram of a terminal device provided by an embodiment of the present application.
  • FIG. 22 is a structural block diagram of a network device provided by an embodiment of the present application.
  • FIG. 1 shows a schematic diagram of a network architecture provided by an embodiment of the present application.
  • the network architecture may include: a terminal device 10 and a network device 20.
  • the number of terminal devices 10 is usually multiple, and one or more terminal devices 10 may be distributed in a cell managed by each network device 20.
  • the terminal device 10 may include various handheld devices with wireless communication functions, vehicle-mounted devices, wearable devices, computing devices or other processing devices connected to a wireless modem, as well as various forms of user equipment (UE), and mobile stations. (Mobile Station, MS) and so on.
  • UE user equipment
  • MS Mobile Station
  • the network device 20 is a device deployed in an access network to provide a wireless communication function for the terminal device 10.
  • the network device 20 may include various forms of macro base stations, micro base stations, relay stations, access points, and so on.
  • the names of devices with network device functions may be different. For example, in 5G NR systems, they are called gNodeB or gNB. As communication technology evolves, the name "network equipment" may change.
  • network devices For ease of description, in the embodiments of the present application, the above-mentioned devices that provide wireless communication functions for the terminal device 10 are collectively referred to as network devices.
  • the "5G NR system" in the embodiments of the present disclosure may also be referred to as a 5G system or an NR system, but those skilled in the art can understand its meaning.
  • the technical solutions described in the embodiments of the present disclosure may be applicable to the 5G NR system, and may also be applicable to the subsequent evolution system of the 5G NR system.
  • AI is taking on more and more important tasks in mobile communication terminals, such as taking pictures, image recognition, video calls, AR (Augmented Reality)/VR (Virtual Reality), games, etc.
  • AI is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results. That is, AI is to study the design principles and implementation methods of various intelligent machines, so that the machines have the functions of perception, reasoning and decision-making.
  • AI technology is a comprehensive discipline, covering a wide range of fields, including both hardware-level technology and software-level technology.
  • Basic AI technologies generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, and operations.
  • AI software technologies mainly include CV (Computer Vision, computer vision technology), speech technology (Speech Technology), NLP (Nature Language processing, natural language processing technology) and ML (Machine Learning, Machine learning)/Deep learning and other major directions.
  • 3GPP has proposed three major application scenarios for the combined application of 5G and AI, namely: segmented AI operation in 5G systems, downloading of AI models in 5G systems, and training of AI models in 5G systems.
  • a feasible method is that the terminal device cooperates with the network device to complete the AI reasoning operation, that is, the "segmented AI reasoning operation", that is, the terminal device completes the time-sensitive, privacy-sensitive and less computationally expensive part of the AI operation.
  • the intermediate results are reported to the network equipment, and the network equipment completes the remaining delay-insensitive, privacy-insensitive, and computationally intensive part.
  • the global model for training needs to be allocated to the terminal device through the mobile communication network, and then the local gradient after the terminal device training is reported to the network device, and then the network device merges the local model of the terminal device to form A more optimized global model.
  • the terminal device 10 is required to upload the intermediate results to the cloud 20 in the "Segmented AI operation in the 5G system";
  • the download of the AI model in the 5G system requires the terminal device 10 to download the AI model from the cloud 20;
  • the “training of the AI/ML model in the 5G system” requires the terminal device 10 to download the global model from the cloud 20 and upload the gradient to the cloud 20 .
  • the terminal device can schedule AI resources separately, that is, the terminal device separately schedules AI resources and wireless resources.
  • the terminal device since the terminal device is in a changing wireless channel environment, and it will constantly move its position, there may be problems such as reduced transmission rate, data packet loss, and uncertain transmission delay.
  • the chip processing resources, storage resources, etc. that different terminal devices can allocate to AI computing are also different, and the chip processing resources, storage resources, etc. of the terminal device may change at any time.
  • the separate scheduling of AI resources and wireless resources there will be situations where the two cannot adapt to each other. For example, in a certain resource allocation mode, AI resources meet the requirements of the terminal equipment to use the AI model, but the wireless resources do not meet the requirements of the terminal equipment. Data interaction requirements; or, wireless resources meet the data interaction requirements of the terminal device, but the AI resources do not meet the requirements of the terminal device to use the AI model. Therefore, separate scheduling of AI resources and wireless resources may cause degradation of AI service performance, and may waste AI resources or wireless resources.
  • the embodiment of the present application provides a resource configuration method that sends resource configuration information to a terminal device through a network device.
  • the resource configuration information includes multiple resource configuration combinations, and each resource configuration combination can include two types of resource configuration information.
  • One type of resource configuration information can be used to indicate wireless resource configuration
  • the other type of resource configuration information can be used to indicate AI resource configuration, thereby realizing the combined configuration between wireless resources and AI resources, compared to wireless resources and AI resources Separate configuration may cause the problem of inability to adapt to each other between wireless resources and AI resources.
  • the embodiment of the present application configures mutually adapted wireless resources and AI resources to form a resource configuration combination, which ensures that sufficient wireless resources are ensured at the same time.
  • FIG. 3 shows a flowchart of a resource configuration method provided by an embodiment of the present application.
  • the method can be applied to the network architecture shown in FIG. 1.
  • the method may include the following steps:
  • Step 310 The network device sends first resource configuration information to the terminal device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combination includes first-type resource configuration information and second-type resource configuration information.
  • the first resource configuration information is combination information of a type of resource configuration, that is, the first resource configuration information is configuration information that combines at least one type of resource configuration.
  • the terminal device may receive the first resource configuration information from the network device, so as to use the first resource configuration information to perform subsequent data interaction, operation execution, and so on.
  • the embodiment of this application does not limit the method for determining the first resource configuration information.
  • the first resource configuration information is determined by the network device. For example, after the network device obtains the service usage requirements of the terminal device, it determines what needs to be configured for the terminal device. Resource type, which further combines multiple types of resource configuration to form multiple resource configuration combinations, that is, the first resource configuration information; or, the first resource configuration information is pre-defined by the agreement, for example, for the possible availability of terminal equipment The agreement predetermines the resource configuration of each business requirement, combines the resource configuration of multiple business requirements, and specifies the first resource configuration information including multiple resource configuration combinations.
  • the embodiment of this application also does not limit the transmission mode of the first resource configuration information.
  • the first resource configuration information is carried in the RRC (Radio Resource Control) configuration information, so that the terminal device can access the network
  • the first configuration information is obtained when the device is configured; or, the first resource configuration information is carried in a system message, so that the terminal device can obtain the first configuration information from the system message broadcast by the network device; or, the first resource configuration information can also be carried In other high-level configuration information, for example, it is carried in DCI (Downlink Control Information), MAC (Media Access Control, Media Access Control) CE (Control Element, control unit).
  • the first resource configuration information includes n resource configuration combinations, and n is a positive integer.
  • the embodiment of this application does not limit the specific number of resource configuration combinations included in the first resource configuration information. That is, the embodiment of this application does not limit the size of n. In practical applications, it can be combined with the resource configuration information in the resource configuration combination. The number of types and the number of resource configuration information of each type determine the size of n.
  • Each resource configuration combination can include multiple types of resource configuration information.
  • multiple types of resource configuration information in each resource configuration combination have an association relationship and are adapted to each other, that is, the embodiment of the present application combines and configures resource configuration information that is associated with each other , And combine the pairwise adapted resource configuration information to form a resource configuration combination.
  • the embodiment of the present application performs a combined configuration on the above-mentioned radio resources and AI resources, and combines the pairwise adapted radio resource configuration information and AI resource configuration information to form a resource configuration combination.
  • the embodiment of the present application does not limit the expression form of the pairwise adaptation between multiple resource configuration information.
  • the pairwise adaptation performance means that the multiple resource configuration information in each resource configuration combination can meet the requirements of the terminal device.
  • the wireless resource configuration information and AI resource configuration information of the pairwise adaptation are expressed as that the wireless resource configuration information in each resource configuration combination can meet the needs of the terminal device for normal data interaction, and the resource configuration combination
  • the AI resource configuration information in can meet the needs of the terminal device for normal AI operations.
  • the embodiments of this application do not limit the quantity and type of resource configuration information included in the resource configuration combination.
  • the quantity and type of the service requirements of the terminal device can be combined to determine the quantity of resource configuration information included in the resource configuration combination. And type.
  • the embodiment of this application proposes a resource configuration combination for situations where the service requirements of terminal devices include "data interaction with network devices" and "execute AI-related operations".
  • the resource configuration combination includes two types of resource configuration information, respectively Configure information for the first type of resource and the second type of resource configuration information, where the first type of resource configuration information is used to indicate the wireless resource configuration, so that the terminal device can use the wireless resource configuration for data interaction with the network device; the second type of resource The configuration information is used to indicate the AI resource configuration, so that the terminal device can use the AI resource configuration to perform AI-related operations.
  • first type of resource configuration information is used to indicate the wireless resource configuration, so that the terminal device can use the wireless resource configuration for data interaction with the network device
  • the second type of resource The configuration information is used to indicate the AI resource configuration, so that the terminal device can use the AI resource configuration to perform AI-related operations.
  • the technical solution provided by the embodiments of the present application sends resource configuration information to a terminal device through a network device.
  • the resource configuration information includes multiple resource configuration combinations, and each resource configuration combination includes multiple resource configuration information.
  • This provides a variety of resource combination configuration methods.
  • the embodiment of the present application configures multiple resource combinations, so that a network device can schedule multiple resources once resource scheduling is performed. Compared with the separate scheduling of multiple resources, multiple resource scheduling is required.
  • the combination provided by the embodiment of the present application The configured solution can reduce the number of resource scheduling performed by the network device, reduce the processing overhead of the network device, and save data transmission resources.
  • each resource configuration combination can include two types of resource configuration information.
  • One type of resource configuration information can be used to indicate wireless resource configuration, and the other type of resource configuration information can be used to indicate AI resource configuration, thereby achieving
  • the combined configuration between wireless resources and AI resources is compared with the problem that wireless resources and AI resources may not be compatible with each other that may occur when wireless resources and AI resources are separately configured.
  • the resource and AI resource configuration together form a resource configuration combination, which ensures the sufficient wireless resources and the quality of AI services at the same time, improves the utilization of wireless resources, and avoids waste of wireless resources or insufficient wireless resources for data interaction when separate configuration occurs. The situation also improves the reliability of AI services, and avoids wasting AI resources or insufficient AI resources to perform AI operations during separate configuration.
  • the content contained in the first type of resource configuration information and the second type of resource configuration information will be introduced below.
  • the content included in the first type of resource configuration information is introduced.
  • the first type of resource configuration information includes at least one of the following: time domain resource information, frequency domain resource information, spatial domain resource information, and code domain resource information.
  • the solution can be Stagger the time domain range or frequency domain range used when the terminal device and the network device exchange data, that is, each terminal can be configured with its available time domain range or frequency domain range, and the terminal device uses the time domain range or the frequency domain range configured for it. Data interaction in the frequency domain can avoid conflicts.
  • the time domain resource information is used to indicate the available time domain range of the terminal device
  • the frequency domain resource information is used to indicate the available frequency domain range of the terminal device.
  • the spatial domain resource information is used to indicate the available beam pair of the terminal device.
  • the code domain resource information is Used to indicate the available encoding range of the terminal device.
  • resource configuration information including time domain resource information, frequency domain resource information, spatial domain resource information, and code domain resource information as examples for illustration.
  • a type of resource configuration information may also include other information used to indicate wireless resource configuration, such as power resource information, which is not limited in the embodiment of the present application.
  • the second type of resource configuration information includes at least one of the following information: model usage information, model operation information, model operation information, model download information, data usage information, data report information, and resource usage information.
  • the model usage information is used to indicate the AI model used by the terminal device.
  • the AI model used by the terminal device is also different.
  • the embodiment of the present application proposes to include model usage information in the second type of resource configuration information, and the model usage information may indicate the AI model used by the terminal device.
  • the model usage information includes the identification of the AI model used by the terminal device, for example, the number of the AI model used by the terminal device.
  • the model operation information is used to indicate the part of the model that the terminal device is responsible for running in the AI model used by the terminal device. Due to the limited computing power and processing overhead of the terminal device, for a certain AI model used by the terminal device, the terminal device may only be responsible for running a part of the AI model, and the remaining part of the AI model needs to be completed by the network device. In order to make the terminal device clarify the part of the model that it needs to run, this embodiment of the application proposes to include model operation information in the second type of resource configuration information.
  • the model operation information is used to indicate that the terminal device is responsible for running in the AI model used by the terminal device. Part of the model.
  • the embodiment of this application does not limit the division method of the model part executed by the terminal device and the network device in a certain AI model.
  • the terminal can be divided according to the level of the AI model.
  • the part of the model executed by the device and the network device separately for example, for an AI model including 4 sub-models, the first sub-model can be divided into execution by the terminal device, and the second to fourth sub-models can be executed by the network device.
  • the embodiment of the application does not limit the specific content of the model operation information.
  • the model operation information includes the information of the model division point of the AI model used by the terminal device.
  • the AI model used by the terminal device can be divided into 4 sub-models. Then the model segmentation points of the AI model are three. If the model part that the terminal device is responsible for running is the first and second sub-models, then the model segmentation points corresponding to the AI model included in the model operation information are 2 .
  • the model operation information is used to indicate the part of the AI model used by the terminal device that the terminal device is responsible for running. Due to the limited computing power and processing overhead of the terminal device, for a certain AI model used by the terminal device, the terminal device may only be responsible for running part of the operations corresponding to the AI model, and the remaining operations in the operations corresponding to the AI model are Need to be completed by the network equipment. That is, for a certain AI model used by the terminal device, the operation performed when the AI model is run can be divided into two parts, one part is run by the terminal device, and the other is run by the network device. In order to make the terminal device clarify the operating part that it needs to run, this embodiment of the application proposes to include model operation information in the second type of resource configuration information.
  • the model operation information is used to indicate the AI model used by the terminal device, which is responsible for running the terminal device Operation part.
  • the embodiment of this application does not limit the division of the operation parts performed by the terminal device and the network device in a certain AI model.
  • the number of operations corresponding to the AI model may be used. Divide the terminal device and the operation part that the network device is responsible for running. For example, for an AI model that includes 12 operations, you can divide the first 4 operations to be run by the terminal device, and the last 8 operations to be run by the network device.
  • the embodiment of the present application does not limit the specific content of the model operation information.
  • the model operation information includes information about the operation division points of the AI model used by the terminal device.
  • the AI model used by the terminal device includes 12 operations corresponding to it. Then there are 11 operation division points corresponding to the AI model. If the operation part responsible for the operation of the terminal device is the first to fifth operations, then the operation division points corresponding to the AI model included in the model operation information are 5 .
  • the model download information is used to indicate the AI model downloaded by the terminal device.
  • the AI model used by the terminal device is different. Due to the limited storage resources of the terminal device, the AI model is usually stored on the network device side. Therefore, for different AI tasks or different AI working environments, the AI model that needs to be downloaded from the network device is also different for the terminal device.
  • the AI model distinguishes the AI model downloaded by the terminal device.
  • the embodiment of the present application proposes to include model download information in the second type of resource configuration information, and the model download information may indicate the AI model downloaded by the terminal device.
  • the model download information includes the identifier of the AI model downloaded by the terminal device, for example, the number of the AI model downloaded by the terminal device.
  • the data usage information is used to indicate the training data used by the terminal device when training the AI model. Since the terminal device needs to use training samples to train the AI model, for different AI models, the terminal device needs to use different training samples to train the AI model. For example, for image processing AI models, the terminal device needs to use image training samples To train the AI model. Moreover, for a specific AI model, different terminal devices have different computing capabilities and storage space, so the training samples that can be used to train the AI model are also different. For example, for terminal devices with weak computing capabilities, The number of training samples that can be used to train a certain AI model is relatively small.
  • this embodiment of the application proposes to include data usage information in the second type of resource configuration information, which can indicate the terminal device used when training a certain AI model Training data.
  • the embodiment of this application does not limit the specific content of the data usage information.
  • the data usage information includes the amount of training data used by the terminal device to train the AI model, for example, the number of training samples used by the terminal device to train the AI model And/or type.
  • the data report information is used to indicate how often the terminal device reports the training result of the AI model.
  • a terminal device participates in distributed learning or federated learning, for the training of a certain AI model, the terminal device needs to report the training result of the AI model to the network device. Since a single data transmission of a terminal device has a certain amount of data limitation, reporting the training result after the AI model is fully trained may result in the training result not being reported at one time, and because the terminal device participates in distributed learning or federated learning, It needs to report the training results to the network device in time for its trained model version to ensure that other training subjects of the AI model obtain the training results in time for subsequent training. Therefore, the terminal device needs to report the training results to the network device multiple times.
  • the embodiment of this application proposes to include data reporting information in the second type of resource configuration information.
  • the data reporting information may indicate the frequency of the terminal device reporting the training result of a certain AI model.
  • the embodiment of the present application does not limit the specific content of the data report information.
  • the data report information includes the report period for the terminal device to report the training result of the AI model.
  • the embodiment of the application does not limit the specific division of the reporting period.
  • the reporting period is divided by time, for example, every 5 seconds is set as the reporting period; or the reporting period is divided by the number of training rounds, for example, every 3 Round training is the reporting period.
  • the resource usage information is used to indicate the amount of resources used by the terminal device when performing AI model-related operations. Since the amount of available resources of the terminal device is different at different times, the amount of resources that can be invested in the training of the AI model or the execution of the AI task is also different. In order to make the terminal device clarify the amount of resources that it can invest in AI model-related operations, this embodiment of the application proposes to include resource usage information in the second type of resource configuration information, which can instruct the terminal device to use it when performing AI model-related operations The amount of resources. The embodiments of this application do not limit the specific content of the resource usage information.
  • the resource usage information includes the computing power used by the terminal device when performing AI model-related operations, that is, the computing power invested by the terminal device when performing AI model-related operations .
  • the technical solutions provided by the embodiments of the present application include time domain resource information, frequency domain resource information, spatial domain resource information, and code domain resource information in the first type of resource configuration information, so that the first type of resource
  • the configuration of configuration information can avoid conflicts in data interaction between different terminal devices and network devices, and ensure successful data interaction between terminal devices and network devices.
  • the technical solutions provided by the embodiments of the present application include model usage information, model operation information, model operation information, model download information, data usage information, data report information, resource usage information, etc., in the second type of resource configuration information.
  • Make the configuration of the second type of resource configuration information match the computing power, processing overhead, and storage space of the terminal device, and prevent the terminal device from not having enough computing power and storage space to perform AI tasks or run AI models, which may cause AI services to fail
  • the terminal device is configured with the second type of resource configuration information used to indicate the AI resource, which ensures the normal operation of the AI service and improves the quality of the AI service performed by the terminal device.
  • the following describes the process of the terminal device selecting the resource configuration combination from the first resource configuration information.
  • the above method further includes the following steps:
  • Step 322 The terminal device selects a first resource configuration combination from the first resource configuration information, and the first resource configuration combination matches the device operation information of the terminal device.
  • the terminal device may select the first resource configuration combination from the multiple resource configuration combinations of the first resource configuration information according to its own device operation information. That is, the main body for determining the first resource configuration combination may be the terminal device, and the basis may be the device operation information of the terminal device.
  • the equipment operation information is used to indicate the amount of resources available for the terminal equipment to perform services currently.
  • the embodiment of the application does not limit the specific content of the device operation information.
  • the device operation information of the terminal device includes: the inactive wireless resource of the terminal device and the inactive computing power of the terminal device.
  • the inactive wireless resource can be used.
  • the computing power to be used can be used to execute AI services.
  • the above-mentioned first resource configuration combination matches the device operation information of the terminal device, including: the first type of resource configuration information in the first resource configuration combination and The inactive wireless resources of the terminal device match, and the second-type resource configuration information in the first resource configuration combination matches the inactive computing power of the terminal device.
  • the above method further includes the following steps:
  • Step 32A The network device sends resource indication information to the terminal device, where the resource indication information is used to indicate a first resource configuration combination in the first resource configuration information, and the first resource configuration combination matches the device operation information of the terminal device.
  • the network device After the network device sends the first resource configuration information to the terminal device, it may continue to send resource indication information to the terminal device, where the resource indication information is used to indicate the first resource configuration combination.
  • the method further includes: the terminal device sends the device operation information of the terminal device to the network device. After the network device obtains the device operating information, it can select the first resource configuration combination from the multiple resource configuration combinations of the first resource configuration information according to the device operating information. That is, the main body for determining the first resource configuration combination may be the network device, and the basis may be the device operation information of the terminal device.
  • the embodiment of this application does not limit the transmission mode of resource indication information.
  • the resource indication information is carried in DCI (Downlink Control Information), so that when the terminal device receives the downlink control information, it can parse out the resource. Indication information; or, resource indication information is carried in MAC CE.
  • the embodiment of the present application also does not limit the encapsulation mode of the resource indication information.
  • the resource indication information is separately encapsulated as a single signaling; or, the resource indication information and other information are combined and encapsulated as a single signaling.
  • Step 32B The terminal device selects the first resource configuration combination according to the resource indication information.
  • the terminal device After receiving the resource indication information, the terminal device parses out the resource indication information, and then determines to use the first resource configuration combination to execute the service.
  • the above method further includes the following steps:
  • Step 321 The network device sends resource activation information to the terminal device, where the resource activation information is used to indicate m resource configuration combinations in the first resource configuration information.
  • the network device After the network device sends the first resource configuration information to the terminal device, it may continue to send resource activation information to the terminal device.
  • the resource activation information is used to indicate m resource configuration combinations, and m is a positive integer less than or equal to n.
  • the embodiment of the present application does not limit the method for determining m resource configuration combinations.
  • the network device determines m resource configuration combinations according to the device identifier of the terminal device, so that the network device can determine different resource configurations for different terminal devices.
  • the network device Combining to realize that multiple terminal devices share the first resource configuration information, avoid the need for network devices to configure different first resource configuration information for different terminal devices, and reduce the processing overhead of the network device; or, the network device according to the terminal device's device operating information Determine m resource configuration combinations, so that the network device can determine different resource configuration combinations based on the amount of available resources of the terminal device at different times, that is, the network device monitors that the terminal device’s device operating information changes, or the change meets certain requirements.
  • update to obtain k resource configuration combinations k is a positive integer less than or equal to n, and send updated resource activation information to the terminal device, and the updated resource activation information is used to indicate the first resource configuration information K resource configuration combinations in.
  • the embodiment of this application does not limit the transmission mode of resource activation information.
  • the resource activation information is carried in the DCI, so that the terminal device can parse out the resource activation information when receiving the downlink control information; or, the resource activation information bears In MAC CE.
  • the embodiment of the present application also does not limit the way of encapsulating the resource activation information.
  • the resource activation information is separately encapsulated as a single signaling; or, the resource activation information and other information are combined and encapsulated as a single signaling.
  • Step 323 The terminal device activates m resource configuration combinations according to the resource activation information.
  • the terminal device After receiving the resource activation information, the terminal device activates the m resource configuration combinations from the n resource configuration combinations according to the m resource configuration combinations indicated by the resource activation information.
  • step 323 it further includes:
  • Step 325 The terminal device selects a first resource configuration combination from the m resource configuration combinations, and the first resource configuration combination matches the device operation information of the terminal device.
  • the terminal device After the terminal device activates m resource configuration combinations, it can use a certain resource configuration combination of the m resource configuration combinations to perform services.
  • the terminal device can configure m resource configurations according to its own device operation information. Select the first resource configuration combination from the combination. That is, the subject of determining the first resource configuration combination from the m resource configuration combinations may be the terminal device, and the basis may be the device operation information of the terminal device.
  • the technical solution provided by the embodiment of the present application uses the terminal device to determine the resource configuration combination that matches its device operating information from the resource configuration information, thereby avoiding the occurrence of the occurrence of the amount of resources available to the terminal device that does not support the selected resource.
  • the configuration combination leads to the situation that the terminal device cannot perform the business normally, and ensures the quality and reliability of the terminal device to perform the service.
  • the terminal device is the subject of determining the resource configuration combination used, which can improve the flexibility of the terminal device in determining the resource configuration combination, and provide the terminal device with space for autonomously selecting the resource configuration combination.
  • the technical solution provided by the embodiments of the present application sends resource indication information to the terminal device through the network device, and the resource indication information may indicate a resource configuration combination that matches the device operation information of the terminal device, so that the terminal device receives the resource indication information Then, the resource configuration combination determined by the network device is selected from the multiple resource configuration combinations to execute the service. Since the determined resource configuration combination matches the device operation information of the terminal device, the quality and reliability of the service performed by the terminal device are ensured.
  • the technical solution provided by the embodiments of the present application sends resource activation information to the terminal device through the network device, and the resource activation information may indicate at least one resource configuration combination, so that the terminal device can activate at least one resource configuration combination according to the resource activation information , And select the resource configuration combination that matches its own device operation information from the activated resource configuration combination to execute the service, ensuring the quality and reliability of the service performed by the terminal device.
  • the network device configures the same resource configuration information for different terminal devices and controls different terminal devices to activate different resource configuration combinations, it is possible to share one resource configuration information without conflicts between terminal devices. It prevents network equipment from configuring different resource configuration information for different terminal equipment, and saves the processing overhead of the network equipment.
  • the first type of resource configuration information includes time-frequency resource information
  • the second type of resource configuration information includes an identifier of an AI model used by the terminal device
  • the time-frequency resource information includes time domain information and frequency domain information
  • the identification of the AI model and the time-frequency for transmitting the output data generated by the AI model allows the terminal equipment to schedule AI resources and wireless resources together, ensuring the adaptation between AI resources and wireless resources, and avoiding the possible lack of wireless resources or waste of wireless resources in the separate scheduling of AI resources and wireless resources And other issues, to ensure the quality and reliability of AI services.
  • AI resources and wireless resources can be scheduled together, the number of resource scheduling performed by the terminal device is reduced, and the processing overhead of the terminal device is reduced.
  • each resource configuration combination includes the identification of the AI model used by the terminal device and its adapted time-frequency resource information.
  • Table 1 List of resource configuration combinations used for AI model selection
  • Resource allocation combination AI model identification Time-frequency resource information Resource allocation combination 1 AI model 1 Time frequency resource 1 Resource allocation combination 2 AI model 2 Time frequency resource 2 Resource allocation combination 3 AI model 3 Time-frequency resources 3
  • the terminal device switching resource configuration combination has the following two ways.
  • Figure 6(a) shows the first method.
  • the network device configures the terminal device with multiple resource configuration combinations, and the terminal device currently uses resource configuration combination 1 to perform services.
  • the network device discovers that it should switch from AI model 1 to AI model 2, it sends resource indication information to the terminal device, and the terminal device switches from resource configuration combination 1 to resource configuration combination 2 according to the resource indication information, that is, the terminal device switches from running AI model 1 Switch to running AI model 2, and at the same time upload output data from time-frequency resource 1 to time-frequency resource 2 to upload output data.
  • Figure 6(b) shows the second way.
  • the network device configures the terminal device with multiple resource configuration combinations and activates the multiple resource configuration combinations.
  • the terminal equipment currently uses resource configuration combination 1 to perform services.
  • the terminal device finds that it should switch from AI model 1 to AI model 2, it switches from resource configuration combination 1 to resource configuration combination 2, that is, the terminal device switches from running AI model 1 to running AI model 2, and at the same time from using time-frequency resources 1 Upload the output data to the time-frequency resource. 2 Upload the output data.
  • the first type of resource configuration information includes time-frequency resource information
  • the second type of resource configuration information includes information about the model division point of the AI model used by the terminal device
  • the time-frequency resource information includes time domain information and frequency domain information. information.
  • the terminal device When using a specific AI model, due to the limited computing power of the terminal device, it may only execute part of an AI model (such as some layers of a neural network), and the rest of the AI model (such as the rest of a neural network) Those layers) need to be completed by network equipment.
  • an AI model such as some layers of a neural network
  • the rest of the AI model such as the rest of a neural network
  • Those layers need to be completed by network equipment.
  • the structure of a CNN Convolutional Neural Networks, convolutional neural network for image recognition is shown in Figure 7. Different layers in CNN have different amounts of calculation and output data.
  • split point 1 split point 1
  • split point 2 split point 2
  • split point 3 split point 3
  • the AI model split point needs to be switched. Since a certain AI model split point and its required time-frequency resources for transmitting output data need to be matched with each other, combining an AI model split point and the corresponding time-frequency resource into a resource configuration combination can make the terminal equipment together Scheduling the AI model segmentation point and the corresponding wireless resources to ensure the adaptation between the AI model segmentation point and wireless resources, avoiding the problem of insufficient wireless resources or waste of wireless resources that may exist in the separate scheduling of the AI model segmentation point and wireless resources, and ensuring AI services Quality and reliability. In addition, since AI resources and wireless resources can be scheduled together, the number of resource scheduling performed by the terminal device is reduced, and the processing overhead of the terminal device is reduced.
  • each resource configuration combination includes the model division point information and time-frequency resource information of the AI model used by the terminal device.
  • Table 2 List of resource allocation combinations used for AI model segmentation
  • Resource allocation combination Model split point of AI model Time-frequency resource information Resource allocation combination 1 Split point 1 Time frequency resource 1 Resource allocation combination 2 Split point 2 Time frequency resource 2 Resource allocation combination 3 Split point 3 Time-frequency resources 3
  • the terminal device When the available computing power of the terminal device changes, the terminal device cannot provide the computing power required by the original split point, and needs to switch to another split point and allocate new wireless resources adapted to it at the same time. As shown in Figure 8(a), in the first period, the computing power of the terminal device can support the split point 3.
  • the time-frequency resource 3 with the smallest amount of time-frequency resources in the resource configuration combination 3 is used to report the output data;
  • the computing power of the terminal device decreases, and resource configuration combination 1 needs to be used, that is, switch to the split point 1 and use the time-frequency resource 1 with the largest amount of time-frequency resources to report output data; enter the third period, the terminal device's computing power is As a result, resource allocation combination 2 can be used instead, that is, switching to split point 2 and using time-frequency resource 2 with a moderate amount of time-frequency resources to report output data.
  • the wireless resources that can be scheduled to the terminal device are affected by various factors, such as changes in wireless channel conditions, interference conditions, and the number of terminal devices.
  • the wireless resource that the network device can schedule to the terminal device changes, the wireless resource cannot support the upload of the output data of the original split point, and it needs to be switched to another split point.
  • the network device in the first period, can schedule the largest time-frequency resource 1 for the terminal device.
  • the terminal's achievable output data transmission volume is the highest, which can support the split point 1.
  • the terminal The computing power resource that the device needs to provide is the smallest; in the second period, the time-frequency resource that the network device can schedule to the terminal device drops to the smallest time-frequency resource 3, and it needs to switch to the resource configuration combination 3, that is, switch to the division point 3.
  • dividing point 3 requires the terminal device to provide the most computing resources; in the third period, the network is hungry and the largest time-frequency resource that can be dispatched to the terminal device rises to time-frequency resource 2, and resource configuration combination 2 can be used instead. That is, switch to split point 2 to reduce the computing power resources invested by the terminal.
  • the terminal device switching resource configuration combination has the following two ways.
  • Figure 9(a) shows the first method.
  • the network device configures the terminal device with multiple resource configuration combinations, and the terminal device currently uses the resource configuration combination 1 to perform services.
  • the network device finds that it should switch from split point 1 to split point 2, it sends resource indication information to the terminal device, and the terminal device switches from resource configuration combination 1 to resource configuration combination 2 according to the resource indication information, that is, the terminal device switches from using the split point 1.
  • Fig. 9(b) shows the second way.
  • the network device configures multiple resource configuration combinations for the terminal device and activates multiple resource configuration combinations.
  • the terminal device currently uses resource configuration combination 1 to perform services.
  • the terminal device discovers that it should switch from split point 1 to split point 2, it switches from resource configuration combination 1 to resource configuration combination 2, that is, the terminal device switches from using split point 1 to run the AI model to split point 2 to run the AI model, and at the same time From using time-frequency resource 1 to upload output data to using time-frequency resource 2 to upload output data.
  • the first type of resource configuration information includes time-frequency resource information
  • the second type of resource configuration information includes information about the operation division point of the AI model used by the terminal device
  • the time-frequency resource information includes time domain information and frequency domain information. information.
  • the operation corresponding to the AI model consists of multiple steps or multiple parts
  • the limited computing resources of the terminal device due to the limited computing resources of the terminal device, only some steps or parts of the operation corresponding to an AI model may be executed, while the rest of the operation corresponding to the AI model
  • the steps or parts need to be completed by network equipment.
  • the available computing resources of the terminal device change or the wireless resource that can be used to transmit and output data changes
  • the corresponding steps or parts of the AI model that the terminal device is responsible for need to be adjusted.
  • the operation segmentation point corresponding to a certain AI model and the time-frequency resource required to transmit output data need to be adapted to each other, and the operation segmentation point corresponding to an AI model and its corresponding time-frequency resource are combined into one resource.
  • the configuration combination enables the terminal equipment to schedule the operation segmentation point of the AI model and the corresponding wireless resources at the same time, to ensure the adaptation between the operation segmentation point of the AI model and the wireless resource, and avoid the possibility of separate scheduling of the operation segmentation point of the AI model and wireless resources
  • the existing problems of insufficient wireless resources or waste of wireless resources ensure the quality and reliability of AI services.
  • AI resources and wireless resources can be scheduled together, the number of resource scheduling performed by the terminal device is reduced, and the processing overhead of the terminal device is reduced.
  • each resource configuration combination includes information on the operation division point of the AI model used by the terminal device and time-frequency resource information.
  • Table 3 List of resource configuration combinations used for AI operation segmentation
  • Resource allocation combination Operational split point of AI model Time-frequency resource information Resource allocation combination 1 Operation split point 1 (the terminal device is responsible for steps 1 to 3) Time frequency resource 1 Resource allocation combination 2 Operation split point 2 (the terminal device is responsible for steps 1 to 5) Time frequency resource 2 Resource allocation combination 3 Operation split point 3 (the terminal device is responsible for steps 1 to 8) Time-frequency resources 3
  • the terminal device When the available computing power of the terminal device changes, the terminal device cannot provide the computing power required by the original operation segmentation point, and needs to switch to another operation segmentation point, and at the same time allocate new wireless resources adapted to it. As shown in Figure 11(a), in the first period, the computing power of the terminal device can complete steps 1 to 8 (operation segmentation point 3).
  • the time-frequency resource 3 with the smallest amount of resources in the resource configuration combination 3 is used to report Output data; entering the second period, the computing power of the terminal device drops, only steps 1 to 3 can be completed, resource configuration combination 1 is required, that is, switch to operation division point 1 and use the time-frequency resource 1 with the largest amount of time-frequency resources to report Output data; entering the third period, the computing power of the terminal device has rebounded, and steps 1 to 5 can be completed, and resource configuration combination 2 can be used instead, that is, switch to operation division point 2 and use time-frequency resource 2 with a moderate amount of resources Report output data.
  • the wireless resources that can be scheduled to the terminal device are affected by various factors, such as changes in wireless channel conditions, interference conditions, and the number of terminal devices.
  • the wireless resource that the network device can schedule to the terminal device changes, the wireless resource cannot support the upload of the output data of the original operation division point, and it needs to switch to other operation division points.
  • the network device in the first period, can schedule the largest time-frequency resource 1 for the terminal device, and the terminal device has the highest achievable output data transmission volume, which can support the operation division point 1 (terminal device Responsible for steps 1 to 3).
  • the computing power resources that the terminal device needs to provide are the smallest; in the second period, the time-frequency resource that can be scheduled to the terminal device drops to the smallest time-frequency resource 3, and it needs to switch to resource configuration combination 3. , That is, switch to operation division point 3 (the terminal device is responsible for steps 1 to 8).
  • the terminal device is required to provide the most computing resources; entering the third period, the network device can schedule the maximum time-frequency resource for the terminal to rise to the time For frequency resource 2, resource configuration combination 2 can be used instead, that is, switching to operation division point 2 (the terminal device is responsible for steps 1 to 5) to reduce the computing resources invested by the terminal device.
  • Fig. 12(a) shows the first method.
  • the network device configures multiple resource configuration combinations for the terminal device, and the terminal device currently uses resource configuration combination 1 to perform services.
  • the network device discovers that it should switch from operation segmentation point 1 (the terminal device is responsible for steps 1 to 3) to operation segmentation point 2 (the terminal device is responsible for steps 1 to 5), it sends resource indication information to the terminal device, and the terminal device starts from Resource configuration combination 1 is switched to resource configuration combination 2, that is, the terminal device switches from being responsible for steps 1 to 3 to being responsible for steps 1 to 5, and at the same time, uploading output data from using time-frequency resource 1 to using time-frequency resource 2 to upload output data.
  • Figure 12(b) shows the second way.
  • the network device configures the terminal device with multiple resource configuration combinations and activates the multiple resource configuration combinations.
  • the terminal device currently uses resource configuration combination 1 to perform services.
  • switch from resource configuration combination 1 to resource configuration combination 2 that is, the terminal The device switches from being responsible for steps 1 to 3 to being responsible for steps 1 to 5, and at the same time, from using time-frequency resource 1 to upload output data to using time-frequency resource 2 to upload output data.
  • the first type of resource configuration information includes time-frequency resource information
  • the second type of resource configuration information includes the reporting period for the terminal device to report the training result of the AI model
  • the time-frequency resource information includes time domain information and frequency domain information .
  • the number of training rounds that the terminal device can complete in a unit time varies according to the available computing power of the terminal device and the time-frequency resources available for data transmission.
  • the smaller the training data reporting period the higher the required computing power and time-frequency resources.
  • the interval period for the terminal device to report training data also needs to be adjusted.
  • the reporting period of a certain training data and the time-frequency resources required to transmit the output data need to be adapted to each other.
  • the terminal equipment also schedules the training data reporting period and the corresponding wireless resources to ensure the adaptation between the training data reporting period and the wireless resources, avoiding the possible lack of wireless resources or waste of wireless resources when the training data reporting period and wireless resources are separately scheduled To ensure the quality and efficiency of AI model training.
  • AI resources and wireless resources can be scheduled together, the number of resource scheduling performed by the terminal device is reduced, and the processing overhead of the terminal device is reduced.
  • each resource configuration combination includes the reporting period of training data and time-frequency resource information.
  • the terminal equipment has the longest interval to report training data, the calculation amount per unit time is the smallest, and the time-frequency resources required to transmit the output data are also the smallest;
  • the second type of training data reporting cycle is 2
  • the interval for the terminal equipment to report training data is shortened, the amount of calculation per unit time increases, and the time-frequency resources required to transmit the output data are also increased;
  • the third type of training data reporting cycle is 3 times, the terminal equipment reports the training data.
  • the interval is the shortest, the calculation amount per unit time is the largest, and the time-frequency resources required to transmit the output data are also the largest.
  • Table 4 List of resource configuration combinations used to adjust the reporting period of AI model training results
  • Resource allocation combination The reporting period of the training results of the AI model Time-frequency resource information Resource allocation combination 1 Reporting cycle 1 (reported every 200 milliseconds) Time frequency resource 1 Resource allocation combination 2 Reporting cycle 2 (reported every 100 milliseconds) Time frequency resource 2 Resource allocation combination 3 Reporting cycle 3 (reported once in 50 milliseconds) Time-frequency resources 3
  • the terminal device When the available computing power of the terminal device changes, the terminal device cannot support the reporting period of the training results of the original AI model, and needs to switch to the reporting period of other training results, and at the same time allocate new wireless resources adapted to it.
  • the computing power of the terminal device in the first period, can complete a round of training and report the training results every 50 milliseconds (reporting cycle 3).
  • the resource configuration combination 3 with the largest amount of resources is used.
  • Time-frequency resource 3 reports the training results; entering the second period, the computing power of the terminal device decreases, and only one round of training can be completed every 200 milliseconds and the training results are reported (reporting cycle 1).
  • Resource configuration combination 1 is required, that is, switching to reporting Cycle 1 uses the time-frequency resource 1 with the smallest amount of time-frequency resources to report output data; entering the third period, the computing power of the terminal device has rebounded, and can complete a round of training every 100 milliseconds and report the training results (reporting cycle 2), Instead, use resource configuration combination 2, that is, switch to reporting period 2 and use time-frequency resource 2 with a moderate amount of resources to report output data.
  • the wireless resources that can be scheduled to the terminal device are affected by various factors, such as changes in wireless channel conditions, interference conditions, and the number of terminal devices.
  • the wireless resource that the network device can schedule to the terminal device changes, the wireless resource cannot support the original reporting period of the training result, and it needs to be switched to another reporting period.
  • the network device in the first period, can schedule the largest time-frequency resource 3 for the terminal device, and the terminal device has the highest achievable output data transmission volume, which can support the reporting of training results every 50 milliseconds (Reporting cycle 3).
  • the computing power resources that the terminal device needs to provide are the largest; in the second period, the time-frequency resource that can be scheduled to the terminal device drops to the smallest time-frequency resource 1, and it needs to switch to resource configuration combination 1. That is, switch to reporting training results every 200 milliseconds (reporting cycle 1). At this time, the terminal device is required to provide the least computing power resources; in the third period, the network device can schedule the largest time-frequency resource for the terminal back to the time-frequency resource 2. You can switch to resource configuration combination 2, that is, switch to reporting training results every 100 milliseconds (reporting cycle 1), and at the same time, the computing resources invested by terminal equipment have also rebounded.
  • the terminal device switching resource configuration combination has the following two ways.
  • Figure 14(a) shows the first method.
  • the network device configures the terminal device with multiple resource configuration combinations, and the terminal device currently uses resource configuration combination 1 to perform services.
  • the network device discovers that it should switch from reporting period 1 (reporting once in 200 milliseconds) to reporting period 2 (reporting once in 100 milliseconds)
  • it sends resource indication information to the terminal device
  • the terminal device switches from resource configuration combination 1 to resource configuration according to the resource indication information Combination 2, that is, the terminal device switches from reporting a training result once in 200 milliseconds to reporting a training result once in 100 milliseconds, and at the same time, uploading the training result from using the time-frequency resource 1 to uploading the training result using the time-frequency resource 2.
  • Figure 14(b) shows the second method.
  • the network device configures the terminal device with multiple resource configuration combinations and activates the multiple resource configuration combinations.
  • the terminal device currently uses resource configuration combination 1 to perform services.
  • the terminal device discovers that it should switch from reporting period 1 (reported once in 200 milliseconds) to reporting period 2 (reported once in 100 milliseconds)
  • it switches from resource configuration combination 1 to resource configuration combination 2 that is, the terminal device reports the training result once from 200 milliseconds Switch to 100 milliseconds to report the training result once, and at the same time upload the training result from the time-frequency resource 1 to the time-frequency resource 2 to upload the training result.
  • the first type of resource configuration information includes time-frequency resource information
  • the second type of resource configuration information includes the computing power used by the terminal device to perform AI model-related operations
  • the time-frequency resource information includes time-domain information and frequency Domain information.
  • the computing resources invested by the terminal equipment and the time-frequency resources needed to transmit and output data need to be adapted to each other, combining a computing power resource and the corresponding time-frequency resource into a resource configuration combination , Can make the terminal equipment dispatch the computing power resources and the corresponding wireless resources together, ensure the adaptation between the computing power resources of the terminal equipment and the wireless resources, and avoid the possible lack of wireless resources in the separate scheduling of the computing resources and wireless resources of the terminal equipment Or the waste of wireless resources to ensure the performance and efficiency of AI operations.
  • AI resources and wireless resources can be scheduled together, the number of resource scheduling performed by the terminal device is reduced, and the processing overhead of the terminal device is reduced.
  • each resource configuration combination includes the computing power resource and time-frequency resource information of the terminal device.
  • the terminal equipment invests the largest computing power resources, but the time-frequency resources required to transmit output data is the smallest; under the second computing power level, the computing power resources invested by the terminal equipment are reduced, but the required The time-frequency resources for transmitting output data are improved; in the third level of computing power, the computing power resources invested by the terminal equipment are the smallest, but the time-frequency resources required to transmit the output data are the largest.
  • Table 5 A list of resource allocation combinations for computing power used when performing AI model-related operations
  • Resource allocation combination The computing power used when performing AI model-related operations Time-frequency resource information Resource allocation combination 1 Computing power 1 (high computing power) Time frequency resource 1 Resource allocation combination 2 Hash Power 2 (Medium Hash Power) Time frequency resource 2 Resource allocation combination 3 Hashrate 3 (low hashrate) Time-frequency resources 3
  • the terminal device When the available computing power of the terminal device changes, the terminal device cannot support the original AI operation, and needs to switch to a lower computing power level, and at the same time allocate new wireless resources adapted to it. As shown in Figure 15(a), in the first period, the terminal device can only invest in a lower level of computing power (computing power 3).
  • the time-frequency resource 3 with the largest amount of resources in the resource configuration combination 3 is used to report the training result ;
  • the computing power level of the terminal equipment rises, you can switch to resource configuration combination 1, using the time-frequency resource 1 with the smallest amount of time-frequency resources to report output data, saving wireless resources;
  • entering the third period the terminal computing power is If it drops to computing power level 2, you can switch to resource configuration combination 2, and use the time-frequency resource 2 with a moderate amount of resources to report the output data.
  • the wireless resources that can be scheduled to the terminal device are affected by various factors, such as changes in wireless channel conditions, interference conditions, and the number of terminal devices.
  • the wireless resource that the network device can schedule to the terminal device changes, the wireless resource cannot support the terminal device to maintain the original computing power level, and it needs to be switched to another computing power level.
  • the network device in the first period, can schedule the largest time-frequency resource 3 for the terminal device, and the terminal device can invest the minimum computing power (calculation power 3); in the second period, it can be scheduled to The time-frequency resource of the terminal equipment drops to the minimum time-frequency resource 1, and the terminal equipment must increase the computing power level to the computing power level 1.
  • the network equipment can schedule the maximum time-frequency resource for the terminal back to time-frequency resource 2 , The terminal device can reduce the computing power level to computing power level 2.
  • the terminal device switching resource configuration combination has the following two ways.
  • Figure 16(a) shows the first method.
  • the network device configures the terminal device with multiple resource configuration combinations, and the terminal device currently uses resource configuration combination 1 to perform services.
  • the network device finds that the computing power level of the terminal device can be reduced (from computing power 1 to computing power 2), it sends resource indication information to the terminal device, and the terminal device switches from resource configuration combination 1 to resource configuration combination 2 according to the resource indication information, That is, the terminal device is switched from high computing power to medium computing power, and at the same time, from using time-frequency resource 1 to upload output data to using time-frequency resource 2 to upload output data.
  • Figure 16(b) shows the second method.
  • the network device configures the terminal device with multiple resource configuration combinations and activates the multiple resource configuration combinations.
  • the terminal device currently uses resource configuration combination 1 to perform services.
  • the terminal device can reduce the computing power level (from computing power 1 to computing power 2), switch from resource configuration combination 1 to resource configuration combination 2, that is, the terminal device switches from high computing power to medium computing power, and from the time of adoption
  • the frequency resource 1 uploads the output data to the time-frequency resource 2 uploads the output data.
  • the technical solution of the present application is introduced and explained mainly from the perspective of interaction between the terminal device and the network device.
  • the above-mentioned steps related to the terminal device can be separately implemented as a resource configuration method on the terminal device side; the above-mentioned steps related to the network device can be separately implemented as a resource configuration method on the network device side.
  • FIG. 17 shows a block diagram of a resource configuration device provided by an embodiment of the present application.
  • the device has the function of realizing the above-mentioned method example on the terminal device side, and the function can be realized by hardware, or by hardware executing corresponding software.
  • the device can be the terminal device described above, or it can be set in the terminal device.
  • the apparatus 1700 may include: a configuration information receiving module 1710.
  • the configuration information receiving module 1710 is configured to receive first resource configuration information from a network device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first-type resource configuration information and second-type resource configuration information.
  • Resource configuration information where n is a positive integer; wherein the first type of resource configuration information is used to indicate wireless resource configuration, and the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration.
  • the second type of resource configuration information includes at least one of the following information: model usage information, used to indicate the AI model used by the terminal device; model operation information, used to indicate the AI model used in the terminal device In the AI model, the part of the model that the terminal device is responsible for running; model operation information is used to indicate the part of the AI model used by the terminal device that the terminal device is responsible for running; model download information is used to indicate all The AI model downloaded by the terminal device; data usage information, used to indicate the training data used when the terminal device trains the AI model; data reporting information, used to indicate the frequency at which the terminal device reports the training result of the AI model; resources The usage information is used to indicate the amount of resources used by the terminal device when performing AI model-related operations.
  • the model usage information includes: the identifier of the AI model used by the terminal device.
  • the model operation information includes: model division point information of the AI model used by the terminal device.
  • the model operation information includes: information of operation division points of the AI model used by the terminal device.
  • the model download information includes: the identifier of the AI model downloaded by the terminal device.
  • the data usage information includes: the amount of training data used when the terminal device trains the AI model.
  • the data report information includes: a report period for the terminal device to report the training result of the AI model.
  • the resource usage information includes: the computing power used by the terminal device when performing AI model related operations.
  • the first type of resource configuration information includes at least one of the following: time domain resource information, frequency domain resource information, spatial domain resource information, and code domain resource information.
  • the first resource configuration information is carried in radio resource control RRC configuration information; or, the first resource configuration information is carried in system information.
  • the apparatus 1700 further includes: a configuration combination selection module 1720, configured to select a first resource configuration combination from the first resource configuration information, the first resource configuration combination and The equipment operation information of the terminal equipment matches.
  • the apparatus 1700 further includes: an indication information receiving module 1730, configured to receive resource indication information from a network device, where the resource indication information is used to indicate the first resource configuration The first resource configuration combination in the information, the first resource configuration combination matches the device operation information of the terminal device; the configuration combination selection module 1720 is configured to select the first resource configuration according to the resource indication information combination.
  • the resource indication information is carried in the downlink control information DCI; or, the resource indication information is carried in the control unit MAC CE of medium access control.
  • the apparatus 1700 further includes: an activation information receiving module 1740, configured to receive resource activation information from a network device, where the resource activation information is used to indicate the first resource configuration The m resource configuration combinations in the information, where m is a positive integer less than or equal to the n; a configuration combination activation module 1750, configured to activate the m resource configuration combinations according to the resource activation information.
  • the apparatus 1700 further includes: a configuration combination selection module 1720, configured to select a first resource configuration combination from the m resource configuration combinations, the first resource configuration combination and The equipment operation information of the terminal equipment matches.
  • the resource activation information is carried in DCI; or, the resource activation information is carried in MAC CE.
  • the device operation information of the terminal device includes: the inactive wireless resources of the terminal device and the inactive computing power of the terminal device; the first resource configuration combination and the device operation of the terminal device
  • the information matching includes: the first-type resource configuration information in the first resource configuration combination matches the inactive radio resource of the terminal device, and the second-type resource configuration in the first resource configuration combination The information matches the standby computing power of the terminal device.
  • the technical solution provided by the embodiments of the present application sends resource configuration information to a terminal device through a network device.
  • the resource configuration information includes multiple resource configuration combinations, and each resource configuration combination includes multiple resource configuration information.
  • This provides a variety of resource combination configuration methods.
  • the embodiment of the present application configures multiple resource combinations, so that a network device can schedule multiple resources once resource scheduling is performed. Compared with the separate scheduling of multiple resources, multiple resource scheduling is required.
  • the combination provided by the embodiment of the present application The configured solution can reduce the number of resource scheduling performed by the network device, reduce the processing overhead of the network device, and save data transmission resources.
  • each resource configuration combination can include two types of resource configuration information.
  • One type of resource configuration information can be used to indicate wireless resource configuration, and the other type of resource configuration information can be used to indicate AI resource configuration, thereby achieving
  • the combined configuration between wireless resources and AI resources is compared with the problem that wireless resources and AI resources may not be compatible with each other that may occur when wireless resources and AI resources are separately configured.
  • the resource and AI resource configuration together form a resource configuration combination, which ensures the sufficient wireless resources and the quality of AI services at the same time, improves the utilization of wireless resources, and avoids waste of wireless resources or insufficient wireless resources for data interaction when separate configuration occurs. The situation also improves the reliability of AI services, and avoids wasting AI resources or insufficient AI resources to perform AI operations during separate configuration.
  • FIG. 19 shows a block diagram of a resource configuration device provided by an embodiment of the present application.
  • the device has the function of realizing the example of the method on the network device side, and the function can be realized by hardware, or by hardware executing corresponding software.
  • the device can be the network device described above, or it can be set in the network device.
  • the apparatus 1900 may include: a configuration information sending module 1910.
  • the configuration information sending module 1910 is configured to send first resource configuration information to a terminal device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first-type resource configuration information and second-type resource configuration Information, the n is a positive integer; wherein the first type of resource configuration information is used to indicate wireless resource configuration, and the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration.
  • the second type of resource configuration information includes at least one of the following information: model usage information, used to indicate the AI model used by the terminal device; model operation information, used to indicate the AI model used in the terminal device In the AI model, the part of the model that the terminal device is responsible for running; model operation information is used to indicate the part of the AI model used by the terminal device that the terminal device is responsible for running; model download information is used to indicate all The AI model downloaded by the terminal device; data usage information, used to indicate the training data used when the terminal device trains the AI model; data reporting information, used to indicate the frequency at which the terminal device reports the training result of the AI model; resources The usage information is used to indicate the amount of resources used by the terminal device when performing AI model-related operations.
  • the model usage information includes: the identifier of the AI model used by the terminal device.
  • the model operation information includes: model division point information of the AI model used by the terminal device.
  • the model operation information includes: information of operation division points of the AI model used by the terminal device.
  • the model download information includes: the identifier of the AI model downloaded by the terminal device.
  • the data usage information includes: the amount of training data used when the terminal device trains the AI model.
  • the data report information includes: a report period for the terminal device to report the training result of the AI model.
  • the resource usage information includes: the computing power used by the terminal device when performing AI model related operations.
  • the first type of resource configuration information includes at least one of the following: time domain resource information, frequency domain resource information, spatial domain resource information, and code domain resource information.
  • the first resource configuration information is carried in radio resource control RRC configuration information; or, the first resource configuration information is carried in system information.
  • the apparatus 1900 further includes: an indication information sending module 1920, configured to send resource indication information to the terminal device, where the resource indication information is used to indicate the first resource configuration The first resource configuration combination in the information, and the first resource configuration combination matches the device operation information of the terminal device.
  • an indication information sending module 1920 configured to send resource indication information to the terminal device, where the resource indication information is used to indicate the first resource configuration The first resource configuration combination in the information, and the first resource configuration combination matches the device operation information of the terminal device.
  • the resource indication information is carried in the downlink control information DCI; or, the resource indication information is carried in the control unit MAC CE of medium access control.
  • the device operation information of the terminal device includes: the inactive wireless resources of the terminal device and the inactive computing power of the terminal device; the first resource configuration combination and the device operation of the terminal device
  • the information matching includes: the first-type resource configuration information in the first resource configuration combination matches the inactive radio resource of the terminal device, and the second-type resource configuration in the first resource configuration combination The information matches the standby computing power of the terminal device.
  • the apparatus 1900 further includes: an activation information sending module 1930, configured to send resource activation information to the terminal device, where the resource activation information is used to indicate the first resource configuration For m resource configuration combinations in the information, the m is a positive integer less than or equal to the n.
  • the resource activation information is carried in DCI; or, the resource activation information is carried in MAC CE.
  • the technical solution provided by the embodiments of the present application sends resource configuration information to a terminal device through a network device.
  • the resource configuration information includes multiple resource configuration combinations, and each resource configuration combination includes multiple resource configuration information.
  • This provides a variety of resource combination configuration methods.
  • the embodiment of the present application configures multiple resource combinations, so that a network device can schedule multiple resources once resource scheduling is performed. Compared with the separate scheduling of multiple resources, multiple resource scheduling is required.
  • the combination provided by the embodiment of the present application The configured solution can reduce the number of resource scheduling performed by the network device, reduce the processing overhead of the network device, and save data transmission resources.
  • each resource configuration combination can include two types of resource configuration information.
  • One type of resource configuration information can be used to indicate wireless resource configuration, and the other type of resource configuration information can be used to indicate AI resource configuration, thereby achieving
  • the combined configuration between wireless resources and AI resources is compared with the problem that wireless resources and AI resources may not be compatible with each other that may occur when wireless resources and AI resources are separately configured.
  • the resource and AI resource configuration together form a resource configuration combination, which ensures the sufficient wireless resources and the quality of AI services at the same time, improves the utilization of wireless resources, and avoids waste of wireless resources or insufficient wireless resources for data interaction when separate configuration occurs. The situation also improves the reliability of AI services, and avoids wasting AI resources or insufficient AI resources to perform AI operations during separate configuration.
  • the device provided in the above embodiment realizes its functions, only the division of the above-mentioned functional modules is used as an example for illustration. In actual applications, the above-mentioned functions can be allocated by different functional modules according to actual needs. That is, the content structure of the device is divided into different functional modules to complete all or part of the functions described above.
  • FIG. 21 shows a schematic structural diagram of a terminal device 210 provided by an embodiment of the present application.
  • the terminal device may be the above-mentioned terminal device for executing the above-mentioned resource configuration method on the terminal device side.
  • the terminal device 210 may include: a processor 211, a receiver 212, a transmitter 213, a memory 214, and a bus 215.
  • the processor 211 includes one or more processing cores, and the processor 211 executes various functional applications and information processing by running software programs and modules.
  • the receiver 212 and the transmitter 213 may be implemented as a transceiver 216, and the transceiver 216 may be a communication chip.
  • the memory 214 is connected to the processor 211 through the bus 215.
  • the memory 214 may be used to store a computer program, and the processor 211 is used to execute the computer program to implement each step executed by the terminal device in the foregoing method embodiment.
  • the memory 214 can be implemented by any type of volatile or non-volatile storage device or a combination thereof.
  • the volatile or non-volatile storage device includes but is not limited to: RAM (Random-Access Memory, random access memory) And ROM (Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory, Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory, Electrically Erasable Programmable Read-Only Memory) Memory), flash memory or other solid-state storage technology, CD-ROM (Compact Disc Read-Only Memory), DVD (Digital Video Disc, high-density digital video disc) or other optical storage, tape cartridges, magnetic tapes, disks Storage or other magnetic storage devices. in:
  • the transceiver 216 is configured to receive first resource configuration information from a network device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first-type resource configuration information and second-type resource configuration information.
  • Resource configuration information where n is a positive integer; wherein, the first type of resource configuration information is used to indicate wireless resource configuration, and the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration.
  • the second type of resource configuration information includes at least one of the following information: model usage information, used to indicate the AI model used by the terminal device; model operation information, used to indicate the AI model used in the terminal device In the AI model, the part of the model that the terminal device is responsible for running; model operation information is used to indicate the part of the AI model used by the terminal device that the terminal device is responsible for running; model download information is used to indicate all The AI model downloaded by the terminal device; data usage information, used to indicate the training data used when the terminal device trains the AI model; data reporting information, used to indicate the frequency at which the terminal device reports the training result of the AI model; resources The usage information is used to indicate the amount of resources used by the terminal device when performing AI model-related operations.
  • the model usage information includes: the identifier of the AI model used by the terminal device.
  • the model operation information includes: model division point information of the AI model used by the terminal device.
  • the model operation information includes: information of operation division points of the AI model used by the terminal device.
  • the model download information includes: the identifier of the AI model downloaded by the terminal device.
  • the data usage information includes: the amount of training data used when the terminal device trains the AI model.
  • the data report information includes: a report period for the terminal device to report the training result of the AI model.
  • the resource usage information includes: the computing power used by the terminal device when performing AI model related operations.
  • the first type of resource configuration information includes at least one of the following: time domain resource information, frequency domain resource information, spatial domain resource information, and code domain resource information.
  • the first resource configuration information is carried in radio resource control RRC configuration information; or, the first resource configuration information is carried in system information.
  • the processor 211 is configured to select a first resource configuration combination from the first resource configuration information, and the first resource configuration combination matches the device operation information of the terminal device.
  • the transceiver 216 is configured to receive resource indication information from a network device, where the resource indication information is used to indicate a first resource configuration combination in the first resource configuration information, and the first The resource configuration combination matches the device operation information of the terminal device; the processor 211 is configured to select the first resource configuration combination according to the resource indication information.
  • the resource indication information is carried in the downlink control information DCI; or, the resource indication information is carried in the control unit MAC CE of medium access control.
  • the transceiver 216 is used to receive resource activation information from a network device, where the resource activation information is used to indicate m resource configuration combinations in the first resource configuration information, where m is A positive integer less than or equal to the n; the processor 211 is configured to activate the m resource configuration combinations according to the resource activation information.
  • the processor 211 is configured to select a first resource configuration combination from the m resource configuration combinations, and the first resource configuration combination matches the device operation information of the terminal device.
  • the resource activation information is carried in DCI; or, the resource activation information is carried in MAC CE.
  • the device operation information of the terminal device includes: the inactive wireless resources of the terminal device and the inactive computing power of the terminal device; the first resource configuration combination and the device operation of the terminal device
  • the information matching includes: the first-type resource configuration information in the first resource configuration combination matches the inactive radio resource of the terminal device, and the second-type resource configuration in the first resource configuration combination The information matches the standby computing power of the terminal device.
  • FIG. 22 shows a schematic structural diagram of a network device 220 provided by an embodiment of the present application.
  • the network device may be the above-mentioned network device for executing the above-mentioned resource configuration method on the network device side.
  • the network device 220 may include: a processor 221, a receiver 222, a transmitter 223, a memory 224, and a bus 225.
  • the processor 221 includes one or more processing cores, and the processor 221 executes various functional applications and information processing by running software programs and modules.
  • the receiver 222 and the transmitter 223 may be implemented as a transceiver 226, and the transceiver 226 may be a communication chip.
  • the memory 224 is connected to the processor 221 through the bus 225.
  • the memory 224 may be used to store a computer program, and the processor 221 is used to execute the computer program to implement each step executed by the network device in the foregoing method embodiment.
  • the memory 224 may be implemented by any type of volatile or non-volatile storage device or a combination thereof.
  • the volatile or non-volatile storage device includes but is not limited to: RAM (Random-Access Memory, random access memory) And ROM (Read-Only Memory), EPROM (Erasable Programmable Read-Only Memory, Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory, Electrically Erasable Programmable Read-Only Memory) Memory), flash memory or other solid-state storage technology, CD-ROM (Compact Disc Read-Only Memory), DVD (Digital Video Disc, high-density digital video disc) or other optical storage, tape cartridges, magnetic tapes, disks Storage or other magnetic storage devices. in:
  • the transceiver 226 is configured to send first resource configuration information to a terminal device, where the first resource configuration information includes n resource configuration combinations, and the resource configuration combinations include first-type resource configuration information and second-type resource configuration Information, the n is a positive integer; wherein the first type of resource configuration information is used to indicate wireless resource configuration, and the second type of resource configuration information is used to indicate artificial intelligence AI resource configuration.
  • the second type of resource configuration information includes at least one of the following information: model usage information, used to indicate the AI model used by the terminal device; model operation information, used to indicate the AI model used in the terminal device In the AI model, the part of the model that the terminal device is responsible for running; model operation information is used to indicate the part of the AI model used by the terminal device that the terminal device is responsible for running; model download information is used to indicate all The AI model downloaded by the terminal device; data usage information, used to indicate the training data used when the terminal device trains the AI model; data reporting information, used to indicate the frequency at which the terminal device reports the training result of the AI model; resources The usage information is used to indicate the amount of resources used by the terminal device when performing AI model-related operations.
  • the model usage information includes: the identifier of the AI model used by the terminal device.
  • the model operation information includes: model division point information of the AI model used by the terminal device.
  • the model operation information includes: information of operation division points of the AI model used by the terminal device.
  • the model download information includes: the identifier of the AI model downloaded by the terminal device.
  • the data usage information includes: the amount of training data used when the terminal device trains the AI model.
  • the data report information includes: a report period for the terminal device to report the training result of the AI model.
  • the resource usage information includes: the computing power used by the terminal device when performing AI model related operations.
  • the first type of resource configuration information includes at least one of the following: time domain resource information, frequency domain resource information, spatial domain resource information, and code domain resource information.
  • the first resource configuration information is carried in radio resource control RRC configuration information; or, the first resource configuration information is carried in system information.
  • the transceiver 226 is further configured to send resource indication information to the terminal device, where the resource indication information is used to indicate a first resource configuration combination in the first resource configuration information, and the first resource configuration information A resource configuration combination matches the device operation information of the terminal device.
  • the resource indication information is carried in the downlink control information DCI; or, the resource indication information is carried in the control unit MAC CE of medium access control.
  • the device operation information of the terminal device includes: the inactive wireless resources of the terminal device and the inactive computing power of the terminal device; the first resource configuration combination and the device operation of the terminal device
  • the information matching includes: the first-type resource configuration information in the first resource configuration combination matches the inactive radio resource of the terminal device, and the second-type resource configuration in the first resource configuration combination The information matches the standby computing power of the terminal device.
  • the transceiver 226 is further configured to send resource activation information to the terminal device, where the resource activation information is used to indicate m resource configuration combinations in the first resource configuration information, and the m Is a positive integer less than or equal to the n.
  • the resource activation information is carried in DCI; or, the resource activation information is carried in MAC CE.
  • the embodiment of the present application also provides a computer-readable storage medium in which a computer program is stored, and the computer program is used to be executed by a processor of a terminal device to implement the above-mentioned resource configuration method on the terminal device side.
  • the embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored, and the computer program is used to be executed by a processor of a network device to implement the above-mentioned resource configuration method on the network device side.
  • An embodiment of the present application also provides a chip, which includes a programmable logic circuit and/or program instructions, and when the chip runs on a terminal device, it is used to implement the resource configuration method on the terminal device side as described above.
  • An embodiment of the present application also provides a chip, which includes a programmable logic circuit and/or program instructions, and when the chip runs on a network device, it is used to implement the resource configuration method on the network device side as described above.
  • This application also provides a computer program product, which when the computer program product runs on a terminal device, causes the computer to execute the above-mentioned resource configuration method on the terminal device side.
  • This application also provides a computer program product, which when the computer program product runs on a network device, causes the computer to execute the above-mentioned resource configuration method on the network device side.
  • the functions described in the embodiments of the present application may be implemented by hardware, software, firmware, or any combination thereof. When implemented by software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on the computer-readable medium.
  • the computer-readable medium includes a computer storage medium and a communication medium, where the communication medium includes any medium that facilitates the transfer of a computer program from one place to another.
  • the storage medium may be any available medium that can be accessed by a general-purpose or special-purpose computer.

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请公开了一种资源配置方法、装置、设备及存储介质，属于通信技术领域。所述方法包括：接收来自于网络设备的第一资源配置信息，第一资源配置信息包括n个资源配置组合，资源配置组合包括第一类资源配置信息和第二类资源配置信息，n为正整数；其中，第一类资源配置信息用于指示无线资源配置，第二类资源配置信息用于指示人工智能AI资源配置。本申请实施例提供的组合配置的方案可以减少网络设备进行资源调度的次数，降低网络设备的处理开销。并且，本申请实施例将彼此适配的无线资源与AI资源配置在一起形成资源配置组合，实现了同时确保无线资源的充足与AI业务的质量，提升了无线资源的利用率，也提升了AI业务的可靠性。

Description

资源配置方法、装置、设备及存储介质 技术领域
本申请实施例涉及通信技术领域，特别涉及一种资源配置方法、装置、设备及存储介质。
背景技术
AI(Artificial Intelligence，人工智能)正在移动通信终端中承担越来越多的重要任务，如拍照、图像识别、视频通话、AR(Augmented Reality，增强现实)/VR(Virtual Reality，虚拟现实)、游戏等。
3GPP(3rd Generation Partnership Project，第三代合作伙伴计划)对5G(5th Generation Mobile Networks，第五代移动通信技术)与AI的结合应用提出了三大应用场景，分别为：在5G系统中的分割式AI操作(splitting AI operation)、在5G系统中的AI模型的下载、在5G系统中的AI模型的训练。其中，“分割式AI操作”是指终端设备完成AI操作中时延敏感、隐私敏感而计算量较小的一部分，将中间结果(intermediate data)上报给网络设备，由网络设备完成剩下的时延不敏感、隐私不敏感而计算量较大的部分；“AI模型的下载”是指当终端设备在移动环境中，面对不同的AI任务，经历不同的AI工作环境，需要使用不同的AI模型，如果终端设备不拥有所需模型，则需要从网络设备下载新的模型使用；“AI模型的训练”是指在模型训练过程中，需要通过网络设备将供训练的全局模型分配给终端设备，再将终端设备训练后的局部梯度(gradient)上报给网络设备，再由网络设备对终端设备的局部模型进行合并，形成更优化的全局模型。
上述三种应用场景都需要使用网络设备的无线资源传输AI资源，包括：“在5G系统中的分割式AI操作”中需要终端设备上传中间结果；“在5G系统中的AI模型的下载”需要终端设备下载AI模型；“在5G系统中的AI/ML模型的训练”需要终端设备下载全局模型并上传梯度。因此，针对上述三种应用场景，如何配置无线资源和AI资源，以确保终端设备正常传输AI资源，还需要进一步地讨论研究。
发明内容
本申请实施例提供了一种资源配置方法、装置、设备及存储介质。所述技术方案如下：
一方面，本申请实施例提供了一种资源配置方法，应用于终端设备中，所述方法包括：
接收来自于网络设备的第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
另一方面，本申请实施例提供了一种资源配置方法，应用于网络设备中，所述方法包括：
向终端设备发送第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
又一方面，本申请实施例提供了一种资源配置装置，应用于终端设备中，所述装置包括：
配置信息接收模块，用于接收来自于网络设备的第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
再一方面，本申请实施例提供了一种资源配置装置，应用于网络设备中，所述装置包括：
配置信息发送模块，用于向终端设备发送第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
还一方面，本申请实施例提供了一种终端设备，所述终端设备包括处理器和与所述处理器相连的收发器；其中：
所述收发器，用于接收来自于网络设备的第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
还一方面，本申请实施例提供了一种网络设备，所述网络设备包括处理器和与所述处理器相连的收发器；其中：
所述收发器，用于向终端设备发送第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
还一方面，本申请实施例提供了一种计算机可读存储介质，所述存储介质中存储有计算机程序，所述计算机程序用于被终端设备的处理器执行，以实现上述终端设备侧的资源配置方法。
还一方面，本申请实施例提供了一种计算机可读存储介质，所述存储介质中存储有计算机程序，所述计算机程序用于被网络设备的处理器执行，以实现上述网络设备侧的资源配置方法。
还一方面，本申请实施例提供了一种芯片，所述芯片包括可编程逻辑电路和/或程序指令，当所述芯片在终端设备上运行时，用于实现如上述终端设备侧的资源配置方法。
还一方面，本申请实施例提供了一种芯片，所述芯片包括可编程逻辑电路和/或程序指令，当所述芯片在网络设备上运行时，用于实现如上述网络设备侧的资源配置方法。
本申请实施例提供的技术方案可以包括如下有益效果：
通过网络设备向终端设备发送资源配置信息，该资源配置信息中包括多个资源配置组合，每个资源配置组合中包括多种资源配置信息，从而提供了多种资源组合配置的方法。并且，本申请实施例将多种资源组合配置，从而网络设备进行一次资源调度即可调度多种资源，相比于将多种资源分开调度需要进行多次资源调度，本申请实施例提供的组合配置的方案可以减少网络设备进行资源调度的次数，降低网络设备的处理开销，节约数据传输资源。
另外，本申请实施例中，每个资源配置组合可以包括两类资源配置信息，一类资源配置信息可以用于指示无线资源配置，另一类资源配置信息可以用于指示AI资源配置，从而实现了无线资源与AI资源之间的组合配置，相比于无线资源与AI资源分开配置可能出现的无线资源与AI资源之间的不能互相适配的问题，本申请实施例将彼此适配的无线资源与AI资源配置在一起形成资源配置组合，实现了同时确保无线资源的充足与AI业务的质量，提升了无线资源的利用率，避免出现分开配置时无线资源浪费或无线资源不足无法进行数据交互的情况，也提升了AI业务的可靠性，避免出现分开配置时AI资源浪费或AI资源不足无法进行AI操作的情况。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是本申请一个实施例提供的网络架构的示意图；
图2是本申请一个实施例提供的AI业务与5G业务结合的示意图；
图3是本申请一个实施例提供的资源配置方法的流程图；
图4是本申请另一个实施例提供的资源配置方法的流程图；
图5是本申请一个实施例提供的资源配置组合的示意图；
图6是图5对应的资源配置方法的流程图；
图7是本申请另一个实施例提供的资源配置组合的示意图；
图8是图7对应的资源配置方法的示意图；
图9是图7对应的资源配置方法的流程图；
图10是本申请另一个实施例提供的资源配置组合的示意图；
图11是图10对应的资源配置方法的示意图；
图12是图10对应的资源配置方法的流程图；
图13是本申请一个实施例提供的资源配置方法的示意图；
图14是图13对应的资源配置方法的流程图；
图15是本申请另一个实施例提供的资源配置方法的示意图；
图16是图15对应的资源配置方法的流程图；
图17是本申请一个实施例提供的资源配置装置的框图；
图18是本申请另一个实施例提供的资源配置装置的框图；
图19是本申请再一个实施例提供的资源配置装置的框图；
图20是本申请还一个实施例提供的资源配置装置的框图；
图21是本申请一个实施例提供的终端设备的结构框图；
图22是本申请一个实施例提供的网络设备的结构框图。
具体实施方式
为使本申请的目的、技术方案和优点更加清楚，下面将结合附图对本申请实施方式作进一步地详细描述。
本申请实施例描述的网络架构以及业务场景是为了更加清楚地说明本申请实施例的技术方案，并不构成对本申请实施例提供的技术方案的限定，本领域普通技术人员可知，随着网络架构的演变和新业务场景的出现，本申请实施例提供的技术方案对于类似的技术问题，同样适用。
请参考图1，其示出了本申请一个实施例提供的网络架构的示意图。该网络架构可以包括：终端设备10和网络设备20。
终端设备10的数量通常为多个，每一个网络设备20所管理的小区内可以分布一个或多个终端设备10。终端设备10可以包括各种具有无线通信功能的手持设备、车载设备、可穿戴设备、计算设备或连接到无线调制解调器的其它处理设备，以及各种形式的用户设备(User Equipment，UE)，移动台(Mobile Station，MS)等等。为方便描述，本申请实施例中，上面提到的设备统称为终端设备。
网络设备20是一种部署在接入网中用以为终端设备10提供无线通信功能的装置。网络设备20可以包括各种形式的宏基站，微基站，中继站，接入点等等。在采用不同的无线接入技术的系统中，具备网络设备功能的设备的名称可能会有所不同，例如在5G NR系统中，称为gNodeB或者gNB。随着通信技术的演进，“网络设备”这一名称可能会变化。为方便描述，本申请实施例中，上述为终端设备10提供无线通信功能的装置统称为网络设备。
本公开实施例中的“5G NR系统”也可以称为5G系统或者NR系统，但本领域技术人员可以理解其含义。本公开实施例描述的技术方案可以适用于5G NR系统，也可以适用于5G NR系统后续的演进系统。
目前，AI正在移动通信终端中承担越来越多的重要任务，如拍照、图像识别、视频通话、AR(Augmented Reality，增强现实)/VR(Virtual Reality，虚拟现实)、游戏等。
AI是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能，感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。也即，AI是研究各种智能机器的设计原理与实现方法，使机器具有感知、推理与决策的功能。AI技术是一门综合学科，涉及领域广泛，既有硬件层面的技术也有软件层面的技术，AI基础技术一般包括如传感器、专用人工智能芯片、云计算、分布式存储、大数据处理技术、操作/交互系统、机电一体化等技术；AI软件技术主要包括CV(Computer Vision，计算机视觉技术)、语音处理技术(Speech Technology)、NLP(Nature Language processing，自然语言处理技术)以及ML(Machine Learning，机器学习)/深度学习等几大方向。
3GPP对5G与AI的结合应用提出了三大应用场景，分别为：在5G系统中的分割式AI操作、在5G系统中的AI模型的下载、在5G系统中的AI模型的训练。
首先，介绍说明“在5G系统中的分割式AI操作”。
传统“将AI推理(inference)操作卸载到云端”的方法需要依赖极低的“传感—推理—控制”端到端还回时延，而毫秒(ms)级别的还回时延不仅要求终端设备和网络设备支持URLLC(Ultra Reliable Low Latency Communications，极可靠低时延通信)，还需要无所不在的MEC(Mobile Edge Computing，移动边缘计算)部署，这在未来的5G网络部署中是极具挑战的。99.9999％的时延则需要完全的网络覆盖，这在5G毫米波频段是无法实现的。另外，“AI操作卸载”还有可能带来隐私保护的风险，将很多终端设备本地数据上传到网络设备可能违反隐私保护法规和用户的意愿，因此终端设备本地的AI操作是必须的。一种可行的方法是终端设备与网络设备合作完成AI推理操作，即“分割式的AI推理操作”，也即，终端设备完成AI操作中时延敏感、隐私敏感而计算量较小的一部分，将中间结果上报给网络设备，由网络设备完成剩下的时延不敏感、隐私不敏感而计算量较大的部分。
其次，介绍说明“在5G系统中的AI模型的下载”。
对于传统“在终端设备进行AI推理操作”的方法，由于终端设备的计算能力、电池资源有限，终端设备上只能运行很低复杂度的AI模型，但低复杂度的AI模型“泛化能力”较差，只能适用于特定的应用 场景和工作环境。当终端设备在移动环境中，面对不同的AI任务，经历不同的AI工作环境，需要使用不同的AI模型。如果终端不拥有所需模型，则需要从网络设备下载新的模型使用。
最后，介绍说明“在5G系统中的AI模型的训练”。
为了利用终端设备采集的宝贵的“小样本数据”进行AI模型的训练，同时保护终端设备对应的用户的数据隐私，需要采用基于移动通信网络的分布式学习(Distributed Learning)和联邦学习(Federated Learning)来实现。在训练过程中，需要通过移动动通信网络将供训练的全局模型分配给终端设备，再将终端设备训练后的局部梯度上报给网络设备，再由网络设备对终端设备的局部模型进行合并，形成更优化的全局模型。
针对上述三种应用场景，都需要使用网络设备的无线资源传输AI资源，如图2所示，“在5G系统中的分割式AI操作”中需要终端设备10上传中间结果到云端20；“在5G系统中的AI模型的下载”需要终端设备10从云端20下载AI模型；“在5G系统中的AI/ML模型的训练”需要终端设备10从云端20下载全局模型，并上传梯度到云端20。在一个示例中，由于“在5G系统中的分割式AI操作”采用的是静态分割，即“哪一部分由终端设备计算、哪一部分由网络设备计算”是固定的，并且，“在5G系统中的AI模型的下载”是按照AI推理任务的需要选择需要下载的AI模型，另外，“在5G系统中的AI模型的训练”是按照AI训练任务的需要确定训练的参数，如训练数据集的大小、上报局部训练结果的频度，因此，终端设备可以单独调度AI资源，也即，终端设备分开调度AI资源和无线资源。
然而，由于终端设备处于变化的无线信道环境中，且其本身也会不断地移动位置，可能会存在传输速率降低、数据丢包、传输时延不确定等问题。另外，不同的终端设备能够分配给AI计算的芯片处理资源、存储资源等也不同，且终端设备的芯片处理资源、存储资源等随时可能发生变化。按照AI资源和无线资源分开调度的话，将出现两者不能互相适配的情况，例如，在某种资源分配模式下，AI资源满足终端设备使用AI模型的要求，但是无线资源不满足终端设备进行数据交互的要求；或者，无线资源满足终端设备进行数据交互的要求，但是AI资源不满足终端设备使用AI模型的要求。因此，AI资源和无线资源分开调度可能会造成AI业务性能的下降，且可能会浪费AI资源或无线资源。
基于此，本申请实施例提供了一种资源配置方法，通过网络设备向终端设备发送资源配置信息，该资源配置信息中包括多个资源配置组合，每个资源配置组合可以包括两类资源配置信息，一类资源配置信息可以用于指示无线资源配置，另一类资源配置信息可以用于指示AI资源配置，从而实现了无线资源与AI资源之间的组合配置，相比于无线资源与AI资源分开配置可能出现的无线资源与AI资源之间的不能互相适配的问题，本申请实施例将彼此适配的无线资源与AI资源配置在一起形成资源配置组合，实现了同时确保无线资源的充足与AI业务的质量，提升了无线资源的利用率，避免出现分开配置时无线资源浪费或无线资源不足无法进行数据交互的情况，也提升了AI业务的可靠性，避免出现分开配置时AI资源浪费或AI资源不足无法进行AI操作的情况。
下面，将结合几个示例性实施例，对本申请技术方案进行介绍说明。
请参考图3，其示出了本申请一个实施例提供的资源配置方法的流程图，该方法可应用于图1所示的网络架构中，该方法可以包括如下步骤：
步骤310，网络设备向终端设备发送第一资源配置信息，该第一资源配置信息包括n个资源配置组合，该资源配置组合包括第一类资源配置信息和第二类资源配置信息。
第一资源配置信息是一种资源配置的组合信息，也即，第一资源配置信息是将至少一种资源配置组合起来的配置信息。本申请实施例中，终端设备可以接收来自于网络设备的第一资源配置信息，以利用该第一资源配置信息进行后续的数据交互、操作执行等。
本申请实施例对第一资源配置信息的确定方式不作限定，可选地，第一资源配置信息由网络设备确定，例如，网络设备获取终端设备的业务使用需求后，确定需要给终端设备配置的资源类型，从而进一步将多种类型的资源配置组合起来，形成多个资源配置组合，即形成第一资源配置信息；或者，第一资源配置信息由协议预先规定，例如，针对终端设备的可能具备的业务需求，协议中预先确定了每种业务需求的资源配置，并将多种业务需求的资源配置组合起来，规定了包括多个资源配置组合的第一资源配置信息。
本申请实施例对第一资源配置信息的传输方式也不作限定，可选地，第一资源配置信息承载于RRC(Radio Resource Control，无线资源控制)配置信息中，从而终端设备可以在接入网络设备时，获取第一配置信息；或者，第一资源配置信息承载于系统消息中，从而终端设备可以从网络设备广播的系统消息中获取第一配置信息；或者，第一资源配置信息还可以承载于其它高层配置信息中，如承载于DCI(Downlink Control Information，下行控制信息)、MAC(Media Access Control，媒体接入控制)CE(Control Element，控制单元)中。
第一资源配置信息包括n个资源配置组合，n为正整数。本申请实施例对第一资源配置信息中包括的资源配置组合的具体数量不作限定，也即，本申请实施例对n的大小不作限定，实际应用中，可以结合资 源配置组合中资源配置信息的类型的数量，以及每种类型的资源配置信息的数量，来确定n的大小。
每个资源配置组合可以包括多种类型的资源配置信息。可选地，每个资源配置组合中的多种类型的资源配置信息之间存在关联关系，并且彼此适配，也即，本申请实施例是对彼此之间存在关联的资源配置信息进行组合配置，且将两两适配的资源配置信息组合在一起形成资源配置组合。例如，本申请实施例对上文所述的无线资源和AI资源进行组合配置，并且将两两适配的无线资源配置信息和AI资源配置信息组合在一起形成资源配置组合。本申请实施例对多种资源配置信息之间两两适配的表现形式不作限定，可选地，两两适配表现为每个资源配置组合中的多种资源配置信息均能够满足终端设备的业务需求，例如，两两适配的无线资源配置信息与AI资源配置信息表现为，每个资源配置组合中的无线资源配置信息能够满足终端设备正常进行数据交互的需求，并且，该资源配置组合中的AI资源配置信息能够满足终端设备正常进行AI操作的需求。
本申请实施例对资源配置组合中包括的资源配置信息的数量和类型均不作限定，实际应用中，可以结合终端设备业务需求的数量和类型，来确定资源配置组合中包括的资源配置信息的数量和类型。本申请实施例针对终端设备的业务需求包括“与网络设备进行数据交互”和“执行AI相关操作”的情况，提出了一种资源配置组合，该资源配置组合中包括两种资源配置信息，分别为第一类资源配置信息和第二类资源配置信息，其中，第一类资源配置信息用于指示无线资源配置，从而终端设备可以使用该无线资源配置与网络设备进行数据交互；第二类资源配置信息用于指示AI资源配置，从而终端设备可以使用该AI资源配置执行AI相关操作。有关第一类资源配置信息与第二类资源配置信息包含的内容的介绍说明，请参见下述方法实施例，此处不再赘述。
综上所述，本申请实施例提供的技术方案，通过网络设备向终端设备发送资源配置信息，该资源配置信息中包括多个资源配置组合，每个资源配置组合中包括多种资源配置信息，从而提供了多种资源组合配置的方法。并且，本申请实施例将多种资源组合配置，从而网络设备进行一次资源调度即可调度多种资源，相比于将多种资源分开调度需要进行多次资源调度，本申请实施例提供的组合配置的方案可以减少网络设备进行资源调度的次数，降低网络设备的处理开销，节约数据传输资源。
另外，本申请实施例中，每个资源配置组合可以包括两类资源配置信息，一类资源配置信息可以用于指示无线资源配置，另一类资源配置信息可以用于指示AI资源配置，从而实现了无线资源与AI资源之间的组合配置，相比于无线资源与AI资源分开配置可能出现的无线资源与AI资源之间的不能互相适配的问题，本申请实施例将彼此适配的无线资源与AI资源配置在一起形成资源配置组合，实现了同时确保无线资源的充足与AI业务的质量，提升了无线资源的利用率，避免出现分开配置时无线资源浪费或无线资源不足无法进行数据交互的情况，也提升了AI业务的可靠性，避免出现分开配置时AI资源浪费或AI资源不足无法进行AI操作的情况。
下面对第一类资源配置信息和第二类资源配置信息包含的内容进行介绍说明。
首先，对第一类资源配置信息包含的内容进行介绍说明。
在一个示例中，第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
由于在同一频段范围内，或者在同一时刻，可能有多个终端设备需要与网络设备之间进行数据交互，为了避免多个终端设备与网络设备之间的数据交互发生冲突，解决的方法可以是错开终端设备与网络设备进行数据交互时使用的时域范围或频域范围，也即，可以为各个终端配置其可用的时域范围或频域范围，终端设备使用为其配置的时域范围或频域范围进行数据交互即可避免冲突。本申请实施例中，时域资源信息即用于指示终端设备的可用时域范围，频域资源信息即用于指示终端设备的可用频域范围。
另外，由于终端设备与网络设备在进行数据交互时，可能有多个候选波束对可供选择，但每个候选波束对对应的参考信号质量各不相同，使用与终端设备适配的波束对进行数据交互可以在一定程度上避免波束失败或数据交互失败。本申请实施例中，空间域资源信息，或称为空域资源信息，即用于指示终端设备的可用波束对。此外，由于终端设备与网络设备之间有多个数据进行交互，为了区分不同的数据，并且，为了降低交互的数据量，常常会对数据进行编码，本申请实施例中，码域资源信息即用于指示终端设备的可用编码范围。
应理解，上述示例性实施例仅以第一类资源配置信息包括时域资源信息、频域资源信息、空间域资源信息和码域资源信息为例进行举例说明，在可选实施例中，第一类资源配置信息还可以包括其它用于指示无线资源配置的信息，例如功率资源信息等，本申请实施例对此不作限定。
接着，对第二类资源配置信息包含的内容进行介绍说明。
在一个示例中，第二类资源配置信息包括以下至少一项信息：模型使用信息、模型运行信息、模型操作信息、模型下载信息、数据使用信息、数据上报信息、资源使用信息。
模型使用信息用于指示终端设备使用的AI模型。针对终端设备需要执行的不同AI任务，或者终端设 备所处的不同AI工作环境，终端设备所使用的AI模型也是不一样的，为了从多个AI模型中区分出终端设备可使用的AI模型，本申请实施例提出在第二类资源配置信息中包括模型使用信息，该模型使用信息可以指示终端设备使用的AI模型。本申请实施例对模型使用信息的具体内容不作限定，可选地，模型使用信息包括终端设备使用的AI模型的标识，例如，终端设备使用的AI模型的编号等。
模型运行信息用于指示在终端设备使用的AI模型中，终端设备负责运行的模型部分。由于终端设备的运算能力和处理开销有限，对于终端设备使用的某一AI模型中，终端设备可能只负责运行该AI模型中的一部分，该AI模型中的剩余部分则需要由网络设备运行完成。为了使得终端设备明确自身需要运行的模型部分，本申请实施例提出在第二类资源配置信息中包括模型运行信息，该模型运行信息用于指示在终端设备使用的AI模型中，终端设备负责运行的模型部分。本申请实施例对某一AI模型中终端设备与网络设备分别执行的模型部分的划分方式不作限定，可选地，针对终端设备使用的某一AI模型，可以根据该AI模型的层级来划分终端设备与网络设备分别执行的模型部分，例如，对于包括4个子模型的AI模型而言，可以划分第1个子模型由终端设备执行，第2个至第4个子模型由网络设备来执行。本申请实施例对模型运行信息的具体内容不作限定，可选地，模型运行信息包括终端设备使用的AI模型的模型分割点的信息，例如，终端设备使用的AI模型可以划分为4个子模型，那么该AI模型的模型分割点为3个，若终端设备负责运行的模型部分是第1个和第2个子模型，那么，在模型运行信息中包括的对应于该AI模型的模型分割点为2。
模型操作信息用于指示在终端设备使用的AI模型中，终端设备负责运行的操作部分。由于终端设备的运算能力和处理开销有限，对于终端设备使用的某一AI模型中，终端设备可能只负责运行该AI模型对应的操作中的一部分操作，该AI模型对应的操作中的剩余操作则需要由网络设备运行完成。也即，对于终端设备使用的某一AI模型来说，可以将该运行该AI模型时所进行的操作划分为两部分，一部分由终端设备运行，一部分由网络设备运行。为了使得终端设备明确自身需要运行的操作部分，本申请实施例提出在第二类资源配置信息中包括模型操作信息，该模型操作信息用于指示终端设备使用的AI模型中，终端设备负责运行的操作部分。本申请实施例对某一AI模型中终端设备与网络设备分别执行的操作部分的划分方式不作限定，可选地，针对终端设备使用的某一AI模型，可以根据该AI模型对应的操作的数量划分终端设备与网络设备负责运行的操作部分，例如，对于包括12个操作的AI模型，可以划分前4个操作由终端设备负责运行，后8个操作由网络设备负责运行。本申请实施例对模型操作信息的具体内容不作限定，可选地，模型操作信息包括终端设备使用的AI模型的操作划分点的信息，例如，终端设备使用的AI模型对应的操作包括12个，那么对应于该AI模型的操作划分点包括11个，若终端设备负责运行的操作部分是第1至5个操作，那么，在模型操作信息中包括的对应于该AI模型的操作划分点为5。
模型下载信息用于指示终端设备下载的AI模型。针对终端设备需要执行的不同AI任务，或者终端设备所处的不同AI工作环境，终端设备所使用的AI模型是不一样的。由于终端设备的存储资源有限，AI模型通常存储于网络设备侧，从而终端设备针对不同的AI任务或者不同的AI工作环境等，需要从网络设备下载的AI模型也是不一样的，为了从多个AI模型中区分出终端设备下载的AI模型，本申请实施例提出在第二类资源配置信息中包括模型下载信息，该模型下载信息可以指示终端设备下载的AI模型。本申请实施例对模型下载信息的具体内容不作限定，可选地，模型下载信息包括终端设备下载的AI模型的标识，例如，终端设备下载的AI模型的编号等。
数据使用信息用于指示终端设备训练AI模型时所使用的训练数据。由于终端设备需要使用训练样本对AI模型进行训练，对于不同的AI模型，终端设备需要采用不同的训练样本来训练该AI模型，例如，对于图像处理类AI模型，终端设备需要使用图像类训练样本来训练该AI模型。并且，对于某一具体的AI模型，不同的终端设备由于运算能力和存储空间不同，其所能采用的训练该AI模型的训练样本也是不同的，例如，对于运算能力较弱的终端设备，其训练某一AI模型所能使用的训练样本的数量较少。为了使得终端设备明确其训练AI模型时所使用的训练样本，本申请实施例提出在第二类资源配置信息中包括数据使用信息，该数据使用信息可以指示终端设备训练某一AI模型时所使用的训练数据。本申请实施例对数据使用信息的具体内容不作限定，可选地，数据使用信息包括终端设备训练AI模型时所使用的训练数据量，例如，终端设备训练AI模型时所使用的训练样本的数量和/或类型。
数据上报信息用于指示终端设备上报AI模型的训练结果的频率。在终端设备参与分布式学习或联邦学习的情况下，针对某一AI模型的训练，终端设备需要向网络设备上报该AI模型的训练结果。由于终端设备单次数据传输具有一定的数据量的限制，将该AI模型全部训练完成后上报训练结果可能导致该训练结果不能一次上报完成，并且，由于终端设备参与了分布式学习或者联邦学习，其需要针对其训练的模型版本及时将训练结果上报给网络设备，以确保该AI模型的其它训练主体及时获取训练结果进行后续的训练，因此，终端设备需要多次向网络设备上报训练结果。为了使得终端设备明确其上报训练结果的频率，本申请实施例提出在第二类资源配置信息中包括数据上报信息，该数据上报信息可以指示终端设备上报某一AI模型的训练结果的频率。本申请实施例对数据上报信息的具体内容不作限定，可选地，数据上报信 息包括终端设备上报AI模型的训练结果的上报周期。本申请实施例对上报周期的具体划分方式不作限定，可选地，上报周期以时间划分，例如，设置每隔5秒为上报周期；或者，上报周期以训练轮数划分，例如，设置每3轮训练为上报周期。
资源使用信息用于指示终端设备执行AI模型相关操作时所使用的资源量。由于在不同时刻下，终端设备的可用资源量是不相同的，其可以投入到AI模型的训练或者执行AI任务中的资源量也是不相同的。为了使得终端设备明确其可以投入AI模型相关操作时的资源量，本申请实施例提出在第二类资源配置信息中包括资源使用信息，该资源使用信息可以指示终端设备执行AI模型相关操作时使用的资源量。本申请实施例对资源使用信息的具体内容不作限定，可选地，资源使用信息包括终端设备执行AI模型相关操作时所使用的算力，也即终端设备执行AI模型相关操作时投入的运算能力。
综上所述，本申请实施例提供的技术方案，通过在第一类资源配置信息中包括时域资源信息、频域资源信息、空间域资源信息和码域资源信息等，使得第一类资源配置信息的配置可以避免不同终端设备与网络设备之间的数据交互不发生冲突，确保终端设备与网络设备之间成功进行数据交互。
并且，本申请实施例提供的技术方案，通过在第二类资源配置信息中包括模型使用信息、模型运行信息、模型操作信息、模型下载信息、数据使用信息、数据上报信息和资源使用信息等，使得第二类资源配置信息的配置与终端设备的运算能力、处理开销和存储空间等匹配，避免终端设备因为没有足够的算力和存储空间来执行AI任务或者运行AI模型等，导致AI业务失败，本申请实施例给终端设备配置了用于指示AI资源的第二类资源配置信息，确保了AI业务的正常运行，提升了终端设备执行AI业务的质量。
下面对终端设备从第一资源配置信息中选择资源配置组合的过程进行介绍说明。
在一个示例中，如图4所示，上述方法还包括如下步骤：
步骤322，终端设备从第一资源配置信息中选择第一资源配置组合，该第一资源配置组合与终端设备的设备运行信息相匹配。
终端设备在接收到第一资源配置信息后，可以根据自身的设备运行信息，从第一资源配置信息的多个资源配置组合中选择第一资源配置组合。也即，确定第一资源配置组合的主体可以是终端设备，依据可以是终端设备的设备运行信息。
设备运行信息用于指示终端设备当前执行业务可用的资源量。本申请实施例对设备运行信息的具体内容不作限定，可选地，终端设备的设备运行信息包括：终端设备的待用无线资源和终端设备的待用算力，其中，待用无线资源可以用于数据交互，待用算力可以用于执行AI业务。
在设备运行信息包括待用无线资源和待用算力的情况下，上述第一资源配置组合与终端设备的设备运行信息相匹配，包括：第一资源配置组合中的第一类资源配置信息与终端设备的待用无线资源相匹配，且，第一资源配置组合中的第二类资源配置信息与终端设备的待用算力相匹配。
在另一个示例中，上述方法还包括如下步骤：
步骤32A，网络设备向终端设备发送资源指示信息，该资源指示信息用于指示第一资源配置信息中的第一资源配置组合，该第一资源配置组合与终端设备的设备运行信息相匹配。
网络设备在向终端设备发送第一资源配置信息后，可以继续向终端设备发送资源指示信息，该资源指示信息用于指示第一资源配置组合。可选地，在步骤32A之前，还包括：终端设备向网络设备发送终端设备的设备运行信息。网络设备获取到设备运行信息后，可以根据该设备运行信息，从第一资源配置信息的多个资源配置组合中选择第一资源配置组合。也即，确定第一资源配置组合的主体可以是网络设备，依据可以是终端设备的设备运行信息。
有关设备运行信息的介绍说明，请参见上述第一个示例，此处不再赘述。
本申请实施例对资源指示信息的传输方式不作限定，可选地，资源指示信息承载于DCI(Downlink Control Information，下行控制信息)中，从而终端设备在接收到下行控制信息时，可以解析出资源指示信息；或者，资源指示信息承载于MAC CE中。本申请实施例对资源指示信息的封装方式也不作限定，可选地，资源指示信息单独封装为一个信令；或者，资源指示信息与其它信息组合封装为一个信令。
步骤32B，终端设备根据资源指示信息，选择第一资源配置组合。
终端设备在接收到资源指示信息后，解析出该资源指示信息，即可确定使用第一资源配置组合执行业务。
在又一个示例中，上述方法还包括如下步骤：
步骤321，网络设备向终端设备发送资源激活信息，该资源激活信息用于指示第一资源配置信息中的m个资源配置组合。
网络设备在向终端设备发送第一资源配置信息后，可以继续向终端设备发送资源激活信息，该资源激活信息用于指示m个资源配置组合，m为小于或等于n的正整数。本申请实施例对m个资源配置组合的确定方式不作限定，可选地，网络设备根据终端设备的设备标识确定m个资源配置组合，从而网络设备可 以针对不同的终端设备，确定不同的资源配置组合，实现多个终端设备共用第一资源配置信息，避免网络设备需要针对不同的终端设备配置不同的第一资源配置信息，降低网络设备的处理开销；或者，网络设备根据终端设备的设备运行信息确定m个资源配置组合，从而网络设备可以针对终端设备在不同时刻的可用资源量，确定不同的资源配置组合，也即，网络设备监测到终端设备的设备运行信息发生变化，或者该变化满足一定条件的情况下，更新得到k个资源配置组合，k为小于或等于n的正整数，并向终端设备发送更新后的资源激活信息，该更新后的资源激活信息用于指示第一资源配置信息中的k个资源配置组合。
本申请实施例对资源激活信息的传输方式不作限定，可选地，资源激活信息承载于DCI中，从而终端设备在接收到下行控制信息时，可以解析出资源激活信息；或者，资源激活信息承载于MAC CE中。本申请实施例对资源激活信息的封装方式也不作限定，可选地，资源激活信息单独封装为一个信令；或者，资源激活信息与其它信息组合封装为一个信令。
步骤323，终端设备根据资源激活信息，激活m个资源配置组合。
终端设备在接收到资源激活信息后，根据该资源激活信息指示的m个资源配置组合，从n个资源配置组合中激活出该m个资源配置组合。
可选地，上述步骤323之后，还包括：
步骤325，终端设备从m个资源配置组合中选择第一资源配置组合，该第一资源配置组合与终端设备的设备运行信息相匹配。
终端设备激活出m个资源配置组合后，可以使用该m个资源配置组合中的某一资源配置组合执行业务，本申请实施例中，终端设备可以根据自身的设备运行信息，从m个资源配置组合中选择第一资源配置组合。也即，从m个资源配置组合中确定第一资源配置组合的主体可以是终端设备，依据可以是终端设备的设备运行信息。
有关设备运行信息的介绍说明，请参见上述第一个示例，此处不再赘述。
综上所述，本申请实施例提供的技术方案，通过终端设备从资源配置信息中，确定与其设备运行信息相匹配的资源配置组合，从而避免出现终端设备可用的资源量不支持选取出的资源配置组合，导致终端设备无法正常执行业务的情况，确保了终端设备执行业务的质量和可靠性。并且，本申请实施例中，确定使用的资源配置组合的主体是终端设备，可以提升终端设备确定资源配置组合的灵活性，提供给终端设备自主选择资源配置组合的空间。
并且，本申请实施例提供的技术方案，通过网络设备向终端设备发送资源指示信息，该资源指示信息可以指示与终端设备的设备运行信息相匹配的资源配置组合，从而终端设备接收到资源指示信息后，从多个资源配置组合中选取网络设备确定的资源配置组合执行业务，由于确定出的资源配置组合与终端设备的设备运行信息相匹配，确保了终端设备执行业务的质量和可靠性。
另外，本申请实施例提供的技术方案，通过网络设备向终端设备发送资源激活信息，该资源激活信息可以指示至少一个资源配置组合，从而终端设备可以根据该资源激活信息，激活至少一个资源配置组合，并从激活的资源配置组合中选择与自身的设备运行信息相匹配的资源配置组合来执行业务，确保了终端设备执行业务的质量和可靠性。此外，由于网络设备针对不同的终端设备配置相同的资源配置信息后，控制不同的终端设备激活不同的资源配置组合，可以在确保终端设备之间不发生冲突的情况下，共用一个资源配置信息，避免网络设备针对不同的终端设备配置不同的资源配置信息，节约网络设备的处理开销。
下面结合几个具体的示例对本申请的技术方案进行介绍说明。
在一个示例中，第一类资源配置信息包括时频资源信息，第二类资源配置信息包括终端设备使用的AI模型的标识，该时频资源信息包括时域信息和频域信息。
由于终端设备的运算能力、处理开销等资源较为有限，终端设备上只能运行较低复杂度的AI模型，但较低复杂度的AI模型的“泛化能力”较差，也即，较低复杂度的AI模型只能适用于特定的应用场景和工作环境。从而，在终端设备处于移动环境的情况下，针对不同的AI任务，或者处于不同的AI工作环境，终端设备需要使用不同的AI模型。另外，不同的AI模型所产生的输出数据量也不同，该输出数据量需要上传给网络设备。因此，某个AI模型及传输其所产生的输出数据的时频资源之间需要相互适配，针对某一AI模型，将该AI模型的标识和传输该AI模型所产生的输出数据的时频资源组合成一个资源配置组合，可以使得终端设备一起调度AI资源和无线资源，确保了AI资源和无线资源之间适配，避免AI资源和无线资源分开调度可能存在的无线资源不足或无线资源浪费等问题，确保了AI业务质量和可靠性。并且，由于AI资源和无线资源可以一起调度，减少了终端设备进行资源调度的次数，降低了终端设备的处理开销。
如图5所示，AI模型2所产生的输出数据量高于AI模型1，AI模型3所产生的输出数据量又高于AI模型2。随着终端需要使用的AI模型的变化，终端使用包含相应的时频资源的资源配置组合。如下述表1所示，每个资源配置组合包含终端设备使用的AI模型的标识及其适配的时频资源信息。
表1：用于AI模型选择的资源配置组合列表
资源配置组合 AI模型的标识 时频资源信息
资源配置组合1 AI模型1 时频资源1
资源配置组合2 AI模型2 时频资源2
资源配置组合3 AI模型3 时频资源3
如图6所示，终端设备切换资源配置组合有以下两种方式。
图6(a)示出了第一种方式，网络设备给终端设备配置了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。当网络设备发现应从AI模型1切换到AI模型2时，向终端设备发送资源指示信息，终端设备根据资源指示信息从资源配置组合1切换到资源配置组合2，也即，终端设备从运行AI模型1切换到运行AI模型2，同时从采用时频资源1上传输出数据到采用时频资源2上传输出数据。
图6(b)示出了第二种方式，网络设备给终端设备配置了多种资源配置组合，并激活了多种资源配置组合。终端设备当前采用的是资源配置组合1执行业务。当终端设备发现应从AI模型1切换到AI模型2时，从资源配置组合1切换到资源配置组合2，也即，终端设备从运行AI模型1切换到运行AI模型2，同时从采用时频资源1上传输出数据到采用时频资源2上传输出数据。
在另一个示例中，第一类资源配置信息包括时频资源信息，第二类资源配置信息包括终端设备使用的AI模型的模型分割点的信息，该时频资源信息包括时域信息和频域信息。
在采用某个特定的AI模型时，由于终端设备的运算能力有限，可能只执行一个AI模型的一部分(如一个神经网络的一些层)，而AI模型的剩余部分(如一个神经网络的剩下那些层)需要由网络设备完成。例如，一个用于图像识别的CNN(Convolutional Neural Networks，卷积神经网络)的结构如图7所示，CNN中的不同层具有不同的计算量和输出数据量，例如，分割点1(split point 1)具有最大的输出数据量，但需要终端设备完成的计算量最小；分割点2(split point 2)的输出数据量小于分割点1，但需要终端设备完成的计算量比分割点1大；分割点3(split point 3)具有最小的输出数据量，但需要终端设备完成的计算量最大。
当终端设备的可用算力发生变化或可用无线资源发生变化时，需要切换AI模型分割点。由于某个AI模型分割点和其所需的传输输出数据的时频资源之间需要相互匹配，将一个AI模型分割点和对应的时频资源组合成一个资源配置组合，可以使得终端设备一并调度AI模型分割点和对应的无线资源，确保AI模型分割点和无线资源之间适配，避免AI模型分割点和无线资源分开调度可能存在的无线资源不足或无线资源浪费的问题，确保AI业务质量和可靠性。并且，由于AI资源和无线资源可以一起调度，减少了终端设备进行资源调度的次数，降低了终端设备的处理开销。
如图7所示，分割点1下，终端设备需要计算的层数最少，计算量最小，但需要上报的输出数据量最高；分割点2下，终端设备需要计算的层数增加，计算量随之增加，但需要上报的输出数据量随之降低；分割点3下，终端设备需要计算的层数最多，计算量最大，但需要上报的输出数据量最低。如下述表2所示，每个资源配置组合包含终端设备使用的AI模型的模型分割点的信息和时频资源信息。
表2：用于AI模型分割的资源配置组合列表
资源配置组合 AI模型的模型分割点 时频资源信息
资源配置组合1 分割点1 时频资源1
资源配置组合2 分割点2 时频资源2
资源配置组合3 分割点3 时频资源3
如图8所示，有两种情况可能导致终端设备使用的资源配置组合发生变化。
当终端设备的可用算力发生变化时，终端设备不能提供原有的分割点所需的算力，需要切换到其他分割点，同时分配与之适配的新的无线资源。如图8(a)所示，在第一时段，终端设备的算力可以支撑分割点3，此时采用资源配置组合3中的时频资源量最小的时频资源3上报输出数据；进入第二时段，终端设备的算力下降，需要采用资源配置组合1，即切换到分割点1并采用时频资源量最大的时频资源1上报输出数据；进入第三时段，终端设备的算力有所回升，可以转而采用资源配置组合2，即切换到分割点2并采用时频资源量适中的时频资源2上报输出数据。
在网络设备调度终端设备资源时，能够调度给终端设备的无线资源受到各种因素影响，例如无线信道条件、干扰情况、终端设备数量等变化。当网络设备能够调度给终端设备的无线资源发生变化时，无线资源不能支持原有的分割点的输出数据上传，需要切换到其他分割点。如图8(b)所示，在第一时段，网络设备能够调度给终端设备最大的时频资源1，终端的可实现的输出数据的传输量最高，可以支持分割点1，此时，终端设备需要提供的算力资源最小；进入第二时段，网络设备能够调度给终端设备的时频资源下降到最小的时频资源3，需要切换到资源配置组合3，即切换到分割点3，此时，分割点3需要终端设备提供最多的算力资源；进入第三时段，网络是被饿能够调度给终端设备最大的时频资源回升到时频资源2，可 以转而采用资源配置组合2，即切换到分割点2，减少终端投入的算力资源。
如图9所示，终端设备切换资源配置组合有以下两种方式。
图9(a)示出了第一种方式，网络设备给终端设备配置了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。当网络设备发现应从分割点1切换到分割点2时，向终端设备发送资源指示信息，终端设备根据资源指示信息从资源配置组合1切换到资源配置组合2，也即，终端设备从采用分割点1运行AI模型切换到采用分割点2运行AI模型，同时从采用时频资源1上传输出数据到采用时频资源2上传输出数据。
图9(b)示出了第二种方式，网络设备给终端设备配置了多种资源配置组合，并激活了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。终端设备发现应从分割点1切换到分割点2时，从资源配置组合1切换到资源配置组合2，也即，终端设备从采用分割点1运行AI模型切换到采用分割点2运行AI模型，同时从采用时频资源1上传输出数据到采用时频资源2上传输出数据。
在又一个示例中，第一类资源配置信息包括时频资源信息，第二类资源配置信息包括终端设备使用的AI模型的操作分割点的信息，该时频资源信息包括时域信息和频域信息。
当AI模型对应的操作由多个步骤或多个部分构成时，由于终端设备的算力资源有限，可能只执行一个AI模型对应的操作的一些步骤或一些部分，而AI模型对应的操作的剩余步骤或部分需要由网络设备完成。当终端设备的可用计算资源发生变化或者能够实现的传输输出数据的无线资源发生变化时，需要调整终端设备负责的AI模型对应的步骤或部分。而某种AI模型对应的操作分割点和其所需的传输输出数据的时频资源之间需要相互适配，将一种AI模型对应的操作分割点和其对应的时频资源组合成一个资源配置组合，可以使得终端设备一并调度AI模型的操作分割点和对应的无线资源，保证AI模型的操作分割点和无线资源之间适配，避免AI模型的操作分割点和无线资源分开调度可能存在的无线资源不足或无线资源浪费的问题，确保AI业务质量和可靠性。并且，由于AI资源和无线资源可以一起调度，减少了终端设备进行资源调度的次数，降低了终端设备的处理开销。
如图10所示，操作分割点1下，终端设备负责执行的步骤最少，计算量最小，但需要上报的输出数据量也最高；操作分割点2下，终端设备负责执行的步骤增加，计算量随之增加，但需要上报的输出数据量随之降低；操作分割点3下，终端设备负责执行的步骤最多，计算量最大，但需要上报的输出数据量最低。如下述表3所示，每个资源配置组合包含终端设备使用的AI模型的操作分割点的信息和时频资源信息。
表3：用于AI操作分割的资源配置组合列表
资源配置组合 AI模型的操作分割点 时频资源信息
资源配置组合1 操作分割点1(终端设备负责步骤1至3) 时频资源1
资源配置组合2 操作分割点2(终端设备负责步骤1至5) 时频资源2
资源配置组合3 操作分割点3(终端设备负责步骤1至8) 时频资源3
如图11所示，有两种情况可能导致终端设备使用的资源配置组合发生变化。
当终端设备的可用算力发生变化时，终端设备不能提供原有的操作分割点所需的算力，需要切换到其他操作分割点，同时分配与之适配的新的无线资源。如图11(a)所示，在第一时段，终端设备的算力可以完成步骤1至8(操作分割点3)，此时采用资源配置组合3中的资源量最小的时频资源3上报输出数据；进入第二时段，终端设备的算力下降，只能完成步骤1至3，需要采用资源配置组合1，即切换到操作分割点1并采用时频资源量最大的时频资源1上报输出数据；进入第三时段，终端设备的算力有所回升，能够完成步骤1至5，可以转而采用资源配置组合2，即切换到操作分割点2并采用资源量适中的时频资源2上报输出数据。
在网络设备调度终端设备资源时，能够调度给终端设备的无线资源受到各种因素影响，例如无线信道条件、干扰情况、终端设备数量等变化。当网络设备能够调度给终端设备的无线资源发生变化时，无线资源不能支持原有的操作分割点的输出数据上传，需要切换到其它操作分割点。如图11(b)所示，在第一时段，网络设备能够调度给终端设备最大的时频资源1，终端设备的可实现的输出数据的传输量最高，可以支持操作分割点1(终端设备负责步骤1至3)，此时，终端设备需要提供的算力资源最小；进入第二时段，能够调度给终端设备的时频资源下降到最小的时频资源3，需要切换到资源配置组合3，即切换到操作分割点3(终端设备负责步骤1至8)，此时，需要终端设备提供最多的算力资源；进入第三时段，网络设备能够调度给终端最大的时频资源回升到时频资源2，可以转而采用资源配置组合2，即切换到操作分割点2(终端设备负责步骤1至5)，减少终端设备投入的算力资源。
如图12所示，终端设备切换资源配置组合有以下两种方式。
图12(a)示出了第一种方式，网络设备给终端设备配置了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。当网络设备发现应从操作分割点1(终端设备负责步骤1至3)切换到操作 分割点2(终端设备负责步骤1至5)时，向终端设备发送资源指示信息，终端设备根据资源指示信息从资源配置组合1切换到资源配置组合2，也即，终端设备从负责步骤1至3切换到负责步骤1至5，同时从采用时频资源1上传输出数据到采用时频资源2上传输出数据。
图12(b)示出了第二种方式，网络设备给终端设备配置了多种资源配置组合，并激活了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。终端设备发现应从操作分割点1(终端设备负责步骤1至3)切换到操作分割点2(终端设备负责步骤1至5)时，从资源配置组合1切换到资源配置组合2，也即，终端设备从负责步骤1至3切换到负责步骤1至5，同时从采用时频资源1上传输出数据到采用时频资源2上传输出数据。
在再一个示例中，第一类资源配置信息包括时频资源信息，第二类资源配置信息包括终端设备上报AI模型的训练结果的上报周期，该时频资源信息包括时域信息和频域信息。
当终端设备参与分布式学习或联邦学习时，根据终端设备的可用算力和可用于的数据传输的时频资源，终端设备在单位时间内能够完成的训练轮次有所不同。另外，训练数据上报周期越小，所需的算力资源和时频资源也越高。当终端设备的可用计算资源发生变化或可用于的数据传输的时频资源发生变化时，也需要调整终端设备上报训练数据的间隔周期。而某种训练数据的上报周期和其所需的传输输出数据的时频资源之间需要相互适配，将一种训练数据的上报周期和对应的时频资源组合成一个资源配置组合，可以使得终端设备一并调度训练数据的上报周期和对应的无线资源，确保训练数据上报周期和无线资源之间适配，避免训练数据的上报周期和无线资源分开调度可能存在的无线资源不足或无线资源浪费的问题，确保AI模型训练的质量和效率。并且，由于AI资源和无线资源可以一起调度，减少了终端设备进行资源调度的次数，降低了终端设备的处理开销。
如下述表4所示，每个资源配置组合包含训练数据的上报周期和时频资源信息。第一种训练数据的上报周期1下，终端设备上报训练数据的间隔最长，单位时间内计算量最小，所需的传输输出数据的时频资源也最小；第二种训练数据的上报周期2下，终端设备上报训练数据的间隔缩短，单位时间内计算量随之增加，所需的传输输出数据的时频资源也提高；第三种训练数据的上报周期3下，终端设备上报训练数据的间隔最短，单位时间内计算量最大，所需的传输输出数据的时频资源也最大。
表4：用于调整AI模型的训练结果的上报周期的资源配置组合列表
资源配置组合 AI模型的训练结果的上报周期 时频资源信息
资源配置组合1 上报周期1(200毫秒上报一次) 时频资源1
资源配置组合2 上报周期2(100毫秒上报一次) 时频资源2
资源配置组合3 上报周期3(50毫秒上报一次) 时频资源3
如图13所示，有两种情况可能导致终端设备使用的资源配置组合发生变化。
当终端设备的可用算力发生变化时，终端设备不能支持原有的AI模型的训练结果的上报周期，需要切换到其他训练结果的上报周期，同时分配与之适配的新的无线资源。如图13(a)所示，在第一时段，终端设备的算力可以每50毫秒完成一轮训练并上报训练结果(上报周期3)，此时采用资源配置组合3中的资源量最大的时频资源3上报训练结果；进入第二时段，终端设备的算力下降，只能每200毫秒完成一轮训练并上报训练结果(上报周期1)，需要采用资源配置组合1，即切换到上报周期1并采用时频资源量最小的时频资源1上报输出数据；进入第三时段，终端设备的算力有所回升，能够每100毫秒完成一轮训练并上报训练结果(上报周期2)，可以转而采用资源配置组合2，即切换到上报周期2并采用资源量适中的时频资源2上报输出数据。
在网络设备调度终端设备资源时，能够调度给终端设备的无线资源受到各种因素影响，例如无线信道条件、干扰情况、终端设备数量等变化。当网络设备能够调度给终端设备的无线资源发生变化时，无线资源不能支持原有的训练结果的上报周期，需要切换到其它上报周期。如图13(b)所示，在第一时段，网络设备能够调度给终端设备最大的时频资源3，终端设备的可实现的输出数据的传输量最高，可以支持每50毫秒上报一次训练结果(上报周期3)，此时，终端设备需要提供的算力资源最大；进入第二时段，能够调度给终端设备的时频资源下降到最小的时频资源1，需要切换到资源配置组合1，即切换到每200毫秒上报一次训练结果(上报周期1)，此时，需要终端设备提供最少的算力资源；进入第三时段，网络设备能够调度给终端最大的时频资源回升到时频资源2，可以转而采用资源配置组合2，即切换到每100毫秒上报一次训练结果(上报周期1)，同时，终端设备投入的算力资源也有所回升。
如图14所示，终端设备切换资源配置组合有以下两种方式。
图14(a)示出了第一种方式，网络设备给终端设备配置了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。当网络设备发现应从上报周期1(200毫秒上报一次)切换到上报周期2(100毫秒上报一次)时，向终端设备发送资源指示信息，终端设备根据资源指示信息从资源配置组合1切换到资源配置组合2，也即，终端设备从200毫秒上报一次训练结果切换到100毫秒上报一次训练结果，同时 从采用时频资源1上传训练结果到采用时频资源2上传训练结果。
图14(b)示出了第二种方式，网络设备给终端设备配置了多种资源配置组合，并激活了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。终端设备发现应从上报周期1(200毫秒上报一次)切换到上报周期2(100毫秒上报一次)时，从资源配置组合1切换到资源配置组合2，也即，终端设备从200毫秒上报一次训练结果切换到100毫秒上报一次训练结果，同时从采用时频资源1上传训练结果到采用时频资源2上传训练结果。
在还一个示例中，第一类资源配置信息包括时频资源信息，第二类资源配置信息包括终端设备执行AI模型相关操作时所使用的算力，该时频资源信息包括时域信息和频域信息。
针对某种AI任务，终端设备投入的算力资源和其所需的传输输出数据的时频资源之间需要相互适配，将一种算力资源和对应的时频资源组合成一个资源配置组合，可以使得终端设备一并调度算力资源和对应的无线资源，确保终端设备的算力资源和无线资源之间适配，避免终端设备的算力资源和无线资源分开调度可能存在的无线资源不足或无线资源浪费的问题，确保AI操作的性能和效率。并且，由于AI资源和无线资源可以一起调度，减少了终端设备进行资源调度的次数，降低了终端设备的处理开销。
如下述表5所示，每个资源配置组合包含终端设备的算力资源和时频资源信息。第一种算力水平下，终端设备投入的算力资源最大，但所需的传输输出数据的时频资源最小；第二种算力水平下，终端设备投入的算力资源降低，但所需的传输输出数据的时频资源提高；第三种算力水平下，终端设备投入的算力资源最小，但所需的传输输出数据的时频资源最大。
表5：用于执行AI模型相关操作时所使用的算力的资源配置组合列表
资源配置组合 执行AI模型相关操作时所使用的算力 时频资源信息
资源配置组合1 算力1(高算力) 时频资源1
资源配置组合2 算力2(中算力) 时频资源2
资源配置组合3 算力3(低算力) 时频资源3
如图15所示，有两种情况可能导致终端设备使用的资源配置组合发生变化。
当终端设备的可用算力发生变化时，终端设备不能支持原有的AI操作，需要切换到较低算力水平，同时分配与之适配的新的无线资源。如图15(a)所示，在第一时段，终端设备只能投入较低算力水平(算力3)，此时采用资源配置组合3中的资源量最大的时频资源3上报训练结果；进入第二时段，终端设备的算力水平上升，可以切换到资源配置组合1，采用时频资源量最小的时频资源1上报输出数据，节省无线资源；进入第三时段，终端算力有所下降，降低到算力水平2，可以切换到资源配置组合2，并采用资源量适中的时频资源2上报输出数据。
在网络设备调度终端设备资源时，能够调度给终端设备的无线资源受到各种因素影响，例如无线信道条件、干扰情况、终端设备数量等变化。当网络设备能够调度给终端设备的无线资源发生变化时，无线资源不能支持终端设备维持在原有的算力水平，需要切换到其它算力水平。如图15(b)所示，在第一时段，网络设备能够调度给终端设备最大的时频资源3，终端设备投入最低算力即可(算力3)；进入第二时段，能够调度给终端设备的时频资源下降到最小的时频资源1，终端设备必须提高算力水平至算力水平1；进入第三时段，网络设备能够调度给终端最大的时频资源回升到时频资源2，终端设备可以降低算力水平至算力水平2。
如图16所示，终端设备切换资源配置组合有以下两种方式。
图16(a)示出了第一种方式，网络设备给终端设备配置了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。当网络设备发现可以降低终端设备的算力水平(从算力1到算力2)时，向终端设备发送资源指示信息，终端设备根据资源指示信息从资源配置组合1切换到资源配置组合2，也即，终端设备高算力切换至中算力，同时从采用时频资源1上传输出数据到采用时频资源2上传输出数据。
图16(b)示出了第二种方式，网络设备给终端设备配置了多种资源配置组合，并激活了多种资源配置组合，终端设备当前采用的是资源配置组合1执行业务。终端设备可以降低算力水平(从算力1到算力2)时，从资源配置组合1切换到资源配置组合2，也即，终端设备从高算力切换到中算力，同时从采用时频资源1上传输出数据到采用时频资源2上传输出数据。
需要说明的一点是，在上述方法实施例中，主要从终端设备和网络设备之间交互的角度，对本申请技术方案进行了介绍说明。上述有关终端设备执行的步骤，可以单独实现成为终端设备侧的资源配置方法；上述有关网络设备执行的步骤，可以单独实现成为网络设备侧的资源配置方法。
下述为本申请装置实施例，可以用于执行本申请方法实施例。对于本申请装置实施例中未披露的细节，请参照本申请方法实施例。
请参考图17，其示出了本申请一个实施例提供的资源配置装置的框图。该装置具有实现上述终端设备侧的方法示例的功能，所述功能可以通过硬件实现，也可以通过硬件执行相应的软件实现。该装置可以是上文所述的终端设备，也可以设置在终端设备中。如图17所示，该装置1700可以包括：配置信息接收模块1710。
配置信息接收模块1710，用于接收来自于网络设备的第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
在一个示例中，所述第二类资源配置信息包括以下至少一项信息：模型使用信息，用于指示所述终端设备使用的AI模型；模型运行信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的模型部分；模型操作信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的操作部分；模型下载信息，用于指示所述终端设备下载的AI模型；数据使用信息，用于指示所述终端设备训练AI模型时所使用的训练数据；数据上报信息，用于指示所述终端设备上报AI模型的训练结果的频率；资源使用信息，用于指示所述终端设备执行AI模型相关操作时所使用的资源量。
在一个示例中，所述模型使用信息包括：所述终端设备使用的AI模型的标识。
在一个示例中，所述模型运行信息包括：所述终端设备使用的AI模型的模型分割点的信息。
在一个示例中，所述模型操作信息包括：所述终端设备使用的AI模型的操作划分点的信息。
在一个示例中，所述模型下载信息包括：所述终端设备下载的AI模型的标识。
在一个示例中，所述数据使用信息包括：所述终端设备训练AI模型时所使用的训练数据量。
在一个示例中，所述数据上报信息包括：所述终端设备上报AI模型的训练结果的上报周期。
在一个示例中，所述资源使用信息包括：所述终端设备执行AI模型相关操作时所使用的算力。
在一个示例中，所述第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
在一个示例中，所述第一资源配置信息承载于无线资源控制RRC配置信息中；或者，所述第一资源配置信息承载于系统信息中。
在一个示例中，如图18所示，所述装置1700还包括：配置组合选择模块1720，用于从所述第一资源配置信息中选择第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
在一个示例中，如图18所示，所述装置1700还包括：指示信息接收模块1730，用于接收来自于网络设备的资源指示信息，所述资源指示信息用于指示所述第一资源配置信息中的第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配；配置组合选择模块1720，用于根据所述资源指示信息，选择所述第一资源配置组合。
在一个示例中，所述资源指示信息承载于下行控制信息DCI中；或者，所述资源指示信息承载于媒体接入控制的控制单元MAC CE中。
在一个示例中，如图18所示，所述装置1700还包括：激活信息接收模块1740，用于接收来自于网络设备的资源激活信息，所述资源激活信息用于指示所述第一资源配置信息中的m个资源配置组合，所述m为小于或等于所述n的正整数；配置组合激活模块1750，用于根据所述资源激活信息，激活所述m个资源配置组合。
在一个示例中，如图18所示，所述装置1700还包括：配置组合选择模块1720，用于从所述m个资源配置组合中选择第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
在一个示例中，所述资源激活信息承载于DCI中；或者，所述资源激活信息承载于MAC CE中。
在一个示例中，所述终端设备的设备运行信息包括：所述终端设备的待用无线资源和所述终端设备的待用算力；所述第一资源配置组合与所述终端设备的设备运行信息相匹配，包括：所述第一资源配置组合中的第一类资源配置信息与所述终端设备的待用无线资源相匹配，且，所述第一资源配置组合中的第二类资源配置信息与所述终端设备的待用算力相匹配。
综上所述，本申请实施例提供的技术方案，通过网络设备向终端设备发送资源配置信息，该资源配置信息中包括多个资源配置组合，每个资源配置组合中包括多种资源配置信息，从而提供了多种资源组合配置的方法。并且，本申请实施例将多种资源组合配置，从而网络设备进行一次资源调度即可调度多种资源，相比于将多种资源分开调度需要进行多次资源调度，本申请实施例提供的组合配置的方案可以减少网络设备进行资源调度的次数，降低网络设备的处理开销，节约数据传输资源。
另外，本申请实施例中，每个资源配置组合可以包括两类资源配置信息，一类资源配置信息可以用于指示无线资源配置，另一类资源配置信息可以用于指示AI资源配置，从而实现了无线资源与AI资源之间的组合配置，相比于无线资源与AI资源分开配置可能出现的无线资源与AI资源之间的不能互相适配的问题，本申请实施例将彼此适配的无线资源与AI资源配置在一起形成资源配置组合，实现了同时确保无线 资源的充足与AI业务的质量，提升了无线资源的利用率，避免出现分开配置时无线资源浪费或无线资源不足无法进行数据交互的情况，也提升了AI业务的可靠性，避免出现分开配置时AI资源浪费或AI资源不足无法进行AI操作的情况。
请参考图19，其示出了本申请一个实施例提供的资源配置装置的框图。该装置具有实现上述网络设备侧的方法示例的功能，所述功能可以通过硬件实现，也可以通过硬件执行相应的软件实现。该装置可以是上文所述的网络设备，也可以设置在网络设备中。如图19所示，该装置1900可以包括：配置信息发送模块1910。
配置信息发送模块1910，用于向终端设备发送第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
在一个示例中，所述第二类资源配置信息包括以下至少一项信息：模型使用信息，用于指示所述终端设备使用的AI模型；模型运行信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的模型部分；模型操作信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的操作部分；模型下载信息，用于指示所述终端设备下载的AI模型；数据使用信息，用于指示所述终端设备训练AI模型时所使用的训练数据；数据上报信息，用于指示所述终端设备上报AI模型的训练结果的频率；资源使用信息，用于指示所述终端设备执行AI模型相关操作时所使用的资源量。
在一个示例中，所述模型使用信息包括：所述终端设备使用的AI模型的标识。
在一个示例中，所述模型运行信息包括：所述终端设备使用的AI模型的模型分割点的信息。
在一个示例中，所述模型操作信息包括：所述终端设备使用的AI模型的操作划分点的信息。
在一个示例中，所述模型下载信息包括：所述终端设备下载的AI模型的标识。
在一个示例中，所述数据使用信息包括：所述终端设备训练AI模型时所使用的训练数据量。
在一个示例中，所述数据上报信息包括：所述终端设备上报AI模型的训练结果的上报周期。
在一个示例中，所述资源使用信息包括：所述终端设备执行AI模型相关操作时所使用的算力。
在一个示例中，所述第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
在一个示例中，所述第一资源配置信息承载于无线资源控制RRC配置信息中；或者，所述第一资源配置信息承载于系统信息中。
在一个示例中，如图20所示，所述装置1900还包括：指示信息发送模块1920，用于向所述终端设备发送资源指示信息，所述资源指示信息用于指示所述第一资源配置信息中的第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
在一个示例中，所述资源指示信息承载于下行控制信息DCI中；或者，所述资源指示信息承载于媒体接入控制的控制单元MAC CE中。
在一个示例中，所述终端设备的设备运行信息包括：所述终端设备的待用无线资源和所述终端设备的待用算力；所述第一资源配置组合与所述终端设备的设备运行信息相匹配，包括：所述第一资源配置组合中的第一类资源配置信息与所述终端设备的待用无线资源相匹配，且，所述第一资源配置组合中的第二类资源配置信息与所述终端设备的待用算力相匹配。
在一个示例中，如图20所示，所述装置1900还包括：激活信息发送模块1930，用于向所述终端设备发送资源激活信息，所述资源激活信息用于指示所述第一资源配置信息中的m个资源配置组合，所述m为小于或等于所述n的正整数。
在一个示例中，所述资源激活信息承载于DCI中；或者，所述资源激活信息承载于MAC CE中。
综上所述，本申请实施例提供的技术方案，通过网络设备向终端设备发送资源配置信息，该资源配置信息中包括多个资源配置组合，每个资源配置组合中包括多种资源配置信息，从而提供了多种资源组合配置的方法。并且，本申请实施例将多种资源组合配置，从而网络设备进行一次资源调度即可调度多种资源，相比于将多种资源分开调度需要进行多次资源调度，本申请实施例提供的组合配置的方案可以减少网络设备进行资源调度的次数，降低网络设备的处理开销，节约数据传输资源。
另外，本申请实施例中，每个资源配置组合可以包括两类资源配置信息，一类资源配置信息可以用于指示无线资源配置，另一类资源配置信息可以用于指示AI资源配置，从而实现了无线资源与AI资源之间的组合配置，相比于无线资源与AI资源分开配置可能出现的无线资源与AI资源之间的不能互相适配的问题，本申请实施例将彼此适配的无线资源与AI资源配置在一起形成资源配置组合，实现了同时确保无线资源的充足与AI业务的质量，提升了无线资源的利用率，避免出现分开配置时无线资源浪费或无线资源不足无法进行数据交互的情况，也提升了AI业务的可靠性，避免出现分开配置时AI资源浪费或AI资源 不足无法进行AI操作的情况。
需要说明的一点是，上述实施例提供的装置在实现其功能时，仅以上述各个功能模块的划分进行举例说明，实际应用中，可以根据实际需要而将上述功能分配由不同的功能模块完成，即将设备的内容结构划分成不同的功能模块，以完成以上描述的全部或者部分功能。
关于上述实施例中的装置，其中各个模块执行操作的具体方式已经在有关该方法的实施例中进行了详细描述，此处将不做详细阐述说明。
请参考图21，其示出了本申请一个实施例提供的终端设备210的结构示意图，例如，该终端设备可以是上文所述终端设备，用于执行上述终端设备侧的资源配置方法。具体来讲：该终端设备210可以包括：处理器211、接收器212、发射器213、存储器214和总线215。
处理器211包括一个或者一个以上处理核心，处理器211通过运行软件程序以及模块，从而执行各种功能应用以及信息处理。
接收器212和发射器213可以实现为一个收发器216，该收发器216可以是一块通信芯片。
存储器214通过总线215与处理器211相连。
存储器214可用于存储计算机程序，处理器211用于执行该计算机程序，以实现上述方法实施例中的终端设备执行的各个步骤。
此外，存储器214可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，易失性或非易失性存储设备包括但不限于：RAM(Random-Access Memory，随机存储器)和ROM(Read-Only Memory，只读存储器)、EPROM(Erasable Programmable Read-Only Memory，可擦写可编程只读存储器)、EEPROM(Electrically Erasable Programmable Read-Only Memory，电可擦写可编程只读存储器)、闪存或其他固态存储其技术，CD-ROM(Compact Disc Read-Only Memory，只读光盘)、DVD(Digital Video Disc，高密度数字视频光盘)或其他光学存储、磁带盒、磁带、磁盘存储或其他磁性存储设备。其中：
所述收发器216，用于接收来自于网络设备的第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
在一个示例中，所述第二类资源配置信息包括以下至少一项信息：模型使用信息，用于指示所述终端设备使用的AI模型；模型运行信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的模型部分；模型操作信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的操作部分；模型下载信息，用于指示所述终端设备下载的AI模型；数据使用信息，用于指示所述终端设备训练AI模型时所使用的训练数据；数据上报信息，用于指示所述终端设备上报AI模型的训练结果的频率；资源使用信息，用于指示所述终端设备执行AI模型相关操作时所使用的资源量。
在一个示例中，所述模型使用信息包括：所述终端设备使用的AI模型的标识。
在一个示例中，所述模型运行信息包括：所述终端设备使用的AI模型的模型分割点的信息。
在一个示例中，所述模型操作信息包括：所述终端设备使用的AI模型的操作划分点的信息。
在一个示例中，所述模型下载信息包括：所述终端设备下载的AI模型的标识。
在一个示例中，所述数据使用信息包括：所述终端设备训练AI模型时所使用的训练数据量。
在一个示例中，所述数据上报信息包括：所述终端设备上报AI模型的训练结果的上报周期。
在一个示例中，所述资源使用信息包括：所述终端设备执行AI模型相关操作时所使用的算力。
在一个示例中，所述第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
在一个示例中，所述第一资源配置信息承载于无线资源控制RRC配置信息中；或者，所述第一资源配置信息承载于系统信息中。
在一个示例中，所述处理器211，用于从所述第一资源配置信息中选择第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
在一个示例中，所述收发器216，用于接收来自于网络设备的资源指示信息，所述资源指示信息用于指示所述第一资源配置信息中的第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配；所述处理器211，用于根据所述资源指示信息，选择所述第一资源配置组合。
在一个示例中，所述资源指示信息承载于下行控制信息DCI中；或者，所述资源指示信息承载于媒体接入控制的控制单元MAC CE中。
在一个示例中，所处收发器216，用于接收来自于网络设备的资源激活信息，所述资源激活信息用于指示所述第一资源配置信息中的m个资源配置组合，所述m为小于或等于所述n的正整数；所述处理器 211，用于根据所述资源激活信息，激活所述m个资源配置组合。
在一个示例中，所述处理器211，用于从所述m个资源配置组合中选择第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
在一个示例中，所述资源激活信息承载于DCI中；或者，所述资源激活信息承载于MAC CE中。
在一个示例中，所述终端设备的设备运行信息包括：所述终端设备的待用无线资源和所述终端设备的待用算力；所述第一资源配置组合与所述终端设备的设备运行信息相匹配，包括：所述第一资源配置组合中的第一类资源配置信息与所述终端设备的待用无线资源相匹配，且，所述第一资源配置组合中的第二类资源配置信息与所述终端设备的待用算力相匹配。
请参考图22，其示出了本申请一个实施例提供的网络设备220的结构示意图，例如，该网络设备可以是上文所述网络设备，用于执行上述网络设备侧的资源配置方法。具体来讲：该网络设备220可以包括：处理器221、接收器222、发射器223、存储器224和总线225。
处理器221包括一个或者一个以上处理核心，处理器221通过运行软件程序以及模块，从而执行各种功能应用以及信息处理。
接收器222和发射器223可以实现为一个收发器226，该收发器226可以是一块通信芯片。
存储器224通过总线225与处理器221相连。
存储器224可用于存储计算机程序，处理器221用于执行该计算机程序，以实现上述方法实施例中的网络设备执行的各个步骤。
此外，存储器224可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，易失性或非易失性存储设备包括但不限于：RAM(Random-Access Memory，随机存储器)和ROM(Read-Only Memory，只读存储器)、EPROM(Erasable Programmable Read-Only Memory，可擦写可编程只读存储器)、EEPROM(Electrically Erasable Programmable Read-Only Memory，电可擦写可编程只读存储器)、闪存或其他固态存储其技术，CD-ROM(Compact Disc Read-Only Memory，只读光盘)、DVD(Digital Video Disc，高密度数字视频光盘)或其他光学存储、磁带盒、磁带、磁盘存储或其他磁性存储设备。其中：
所述收发器226，用于向终端设备发送第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
在一个示例中，所述第二类资源配置信息包括以下至少一项信息：模型使用信息，用于指示所述终端设备使用的AI模型；模型运行信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的模型部分；模型操作信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的操作部分；模型下载信息，用于指示所述终端设备下载的AI模型；数据使用信息，用于指示所述终端设备训练AI模型时所使用的训练数据；数据上报信息，用于指示所述终端设备上报AI模型的训练结果的频率；资源使用信息，用于指示所述终端设备执行AI模型相关操作时所使用的资源量。
在一个示例中，所述模型使用信息包括：所述终端设备使用的AI模型的标识。
在一个示例中，所述模型运行信息包括：所述终端设备使用的AI模型的模型分割点的信息。
在一个示例中，所述模型操作信息包括：所述终端设备使用的AI模型的操作划分点的信息。
在一个示例中，所述模型下载信息包括：所述终端设备下载的AI模型的标识。
在一个示例中，所述数据使用信息包括：所述终端设备训练AI模型时所使用的训练数据量。
在一个示例中，所述数据上报信息包括：所述终端设备上报AI模型的训练结果的上报周期。
在一个示例中，所述资源使用信息包括：所述终端设备执行AI模型相关操作时所使用的算力。
在一个示例中，所述第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
在一个示例中，所述第一资源配置信息承载于无线资源控制RRC配置信息中；或者，所述第一资源配置信息承载于系统信息中。
在一个示例中，所述收发器226，还用于向所述终端设备发送资源指示信息，所述资源指示信息用于指示所述第一资源配置信息中的第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
在一个示例中，所述资源指示信息承载于下行控制信息DCI中；或者，所述资源指示信息承载于媒体接入控制的控制单元MAC CE中。
在一个示例中，所述终端设备的设备运行信息包括：所述终端设备的待用无线资源和所述终端设备的待用算力；所述第一资源配置组合与所述终端设备的设备运行信息相匹配，包括：所述第一资源配置组合中的第一类资源配置信息与所述终端设备的待用无线资源相匹配，且，所述第一资源配置组合中的第二类资源配置信息与所述终端设备的待用算力相匹配。
在一个示例中，所述收发器226，还用于向所述终端设备发送资源激活信息，所述资源激活信息用于指示所述第一资源配置信息中的m个资源配置组合，所述m为小于或等于所述n的正整数。
在一个示例中，所述资源激活信息承载于DCI中；或者，所述资源激活信息承载于MAC CE中。
本申请实施例还提供了一种计算机可读存储介质，所述存储介质中存储有计算机程序，所述计算机程序用于被终端设备的处理器执行，以实现上述终端设备侧的资源配置方法。
本申请实施例还提供了一种计算机可读存储介质，所述存储介质中存储有计算机程序，所述计算机程序用于被网络设备的处理器执行，以实现上述网络设备侧的资源配置方法。
本申请实施例还提供了一种芯片，所述芯片包括可编程逻辑电路和/或程序指令，当所述芯片在终端设备上运行时，用于实现如上述终端设备侧的资源配置方法。
本申请实施例还提供了一种芯片，所述芯片包括可编程逻辑电路和/或程序指令，当所述芯片在网络设备上运行时，用于实现如上述网络设备侧的资源配置方法。
本申请还提供了一种计算机程序产品，当计算机程序产品在终端设备上运行时，使得计算机执行上述终端设备侧的资源配置方法。
本申请还提供了一种计算机程序产品，当计算机程序产品在网络设备上运行时，使得计算机执行上述网络设备侧的资源配置方法。
本领域技术人员应该可以意识到，在上述一个或多个示例中，本申请实施例所描述的功能可以用硬件、软件、固件或它们的任意组合来实现。当使用软件实现时，可以将这些功能存储在计算机可读介质中或者作为计算机可读介质上的一个或多个指令或代码进行传输。计算机可读介质包括计算机存储介质和通信介质，其中通信介质包括便于从一个地方向另一个地方传送计算机程序的任何介质。存储介质可以是通用或专用计算机能够存取的任何可用介质。
以上所述仅为本申请的示例性实施例，并不用以限制本申请，凡在本申请的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本申请的保护范围之内。

Claims (72)

  1. 一种资源配置方法，其特征在于，应用于终端设备中，所述方法包括：
    接收来自于网络设备的第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
    其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
  2. 根据权利要求1所述的方法，其特征在于，所述第二类资源配置信息包括以下至少一项信息：
    模型使用信息，用于指示所述终端设备使用的AI模型；
    模型运行信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的模型部分；
    模型操作信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的操作部分；
    模型下载信息，用于指示所述终端设备下载的AI模型；
    数据使用信息，用于指示所述终端设备训练AI模型时所使用的训练数据；
    数据上报信息，用于指示所述终端设备上报AI模型的训练结果的频率；
    资源使用信息，用于指示所述终端设备执行AI模型相关操作时所使用的资源量。
  3. 根据权利要求2所述的方法，其特征在于，所述模型使用信息包括：所述终端设备使用的AI模型的标识。
  4. 根据权利要求2或3所述的方法，其特征在于，所述模型运行信息包括：所述终端设备使用的AI模型的模型分割点的信息。
  5. 根据权利要求2至4任一项所述的方法，其特征在于，所述模型操作信息包括：所述终端设备使用的AI模型的操作划分点的信息。
  6. 根据权利要求2至5任一项所述的方法，其特征在于，所述模型下载信息包括：所述终端设备下载的AI模型的标识。
  7. 根据权利要求2至6任一项所述的方法，其特征在于，所述数据使用信息包括：所述终端设备训练AI模型时所使用的训练数据量。
  8. 根据权利要求2至7任一项所述的方法，其特征在于，所述数据上报信息包括：所述终端设备上报AI模型的训练结果的上报周期。
  9. 根据权利要求2至8任一项所述的方法，其特征在于，所述资源使用信息包括：所述终端设备执行AI模型相关操作时所使用的算力。
  10. 根据权利要求1至9任一项所述的方法，其特征在于，所述第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
  11. 根据权利要求1至10任一项所述的方法，其特征在于，
    所述第一资源配置信息承载于无线资源控制RRC配置信息中；
    或者，
    所述第一资源配置信息承载于系统信息中。
  12. 根据权利要求1至11任一项所述的方法，其特征在于，所述方法还包括：
    从所述第一资源配置信息中选择第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
  13. 根据权利要求1至11任一项所述的方法，其特征在于，所述方法还包括：
    接收来自于网络设备的资源指示信息，所述资源指示信息用于指示所述第一资源配置信息中的第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配；
    根据所述资源指示信息，选择所述第一资源配置组合。
  14. 根据权利要求13所述的方法，其特征在于，
    所述资源指示信息承载于下行控制信息DCI中；
    或者，
    所述资源指示信息承载于媒体接入控制的控制单元MAC CE中。
  15. 根据权利要求1至11任一项所述的方法，其特征在于，所述方法还包括：
    接收来自于网络设备的资源激活信息，所述资源激活信息用于指示所述第一资源配置信息中的m个资源配置组合，所述m为小于或等于所述n的正整数；
    根据所述资源激活信息，激活所述m个资源配置组合。
  16. 根据权利要求15所述的方法，其特征在于，所述根据所述资源激活信息，激活所述m个资源配 置组合之后，还包括：
    从所述m个资源配置组合中选择第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
  17. 根据权利要求15或16所述的方法，其特征在于，
    所述资源激活信息承载于DCI中；
    或者，
    所述资源激活信息承载于MAC CE中。
  18. 根据权利要求12、13和16任一项所述的方法，其特征在于，所述终端设备的设备运行信息包括：所述终端设备的待用无线资源和所述终端设备的待用算力；
    所述第一资源配置组合与所述终端设备的设备运行信息相匹配，包括：所述第一资源配置组合中的第一类资源配置信息与所述终端设备的待用无线资源相匹配，且，所述第一资源配置组合中的第二类资源配置信息与所述终端设备的待用算力相匹配。
  19. 一种资源配置方法，其特征在于，应用于网络设备中，所述方法包括：
    向终端设备发送第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
    其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
  20. 根据权利要求19所述的方法，其特征在于，所述第二类资源配置信息包括以下至少一项信息：
    模型使用信息，用于指示所述终端设备使用的AI模型；
    模型运行信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的模型部分；
    模型操作信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的操作部分；
    模型下载信息，用于指示所述终端设备下载的AI模型；
    数据使用信息，用于指示所述终端设备训练AI模型时所使用的训练数据；
    数据上报信息，用于指示所述终端设备上报AI模型的训练结果的频率；
    资源使用信息，用于指示所述终端设备执行AI模型相关操作时所使用的资源量。
  21. 根据权利要求20所述的方法，其特征在于，所述模型使用信息包括：所述终端设备使用的AI模型的标识。
  22. 根据权利要求20或21所述的方法，其特征在于，所述模型运行信息包括：所述终端设备使用的AI模型的模型分割点的信息。
  23. 根据权利要求20至22任一项所述的方法，其特征在于，所述模型操作信息包括：所述终端设备使用的AI模型的操作划分点的信息。
  24. 根据权利要求20至23任一项所述的方法，其特征在于，所述模型下载信息包括：所述终端设备下载的AI模型的标识。
  25. 根据权利要求20至24任一项所述的方法，其特征在于，所述数据使用信息包括：所述终端设备训练AI模型时所使用的训练数据量。
  26. 根据权利要求20至25任一项所述的方法，其特征在于，所述数据上报信息包括：所述终端设备上报AI模型的训练结果的上报周期。
  27. 根据权利要求20至26任一项所述的方法，其特征在于，所述资源使用信息包括：所述终端设备执行AI模型相关操作时所使用的算力。
  28. 根据权利要求19至27任一项所述的方法，其特征在于，所述第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
  29. 根据权利要求19至28任一项所述的方法，其特征在于，
    所述第一资源配置信息承载于无线资源控制RRC配置信息中；
    或者，
    所述第一资源配置信息承载于系统信息中。
  30. 根据权利要求19至29任一项所述的方法，其特征在于，所述方法还包括：
    向所述终端设备发送资源指示信息，所述资源指示信息用于指示所述第一资源配置信息中的第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
  31. 根据权利要求30所述的方法，其特征在于，
    所述资源指示信息承载于下行控制信息DCI中；
    或者，
    所述资源指示信息承载于媒体接入控制的控制单元MAC CE中。
  32. 根据权利要求30或31所述的方法，其特征在于，所述终端设备的设备运行信息包括：所述终端设备的待用无线资源和所述终端设备的待用算力；
    所述第一资源配置组合与所述终端设备的设备运行信息相匹配，包括：所述第一资源配置组合中的第一类资源配置信息与所述终端设备的待用无线资源相匹配，且，所述第一资源配置组合中的第二类资源配置信息与所述终端设备的待用算力相匹配。
  33. 根据权利要求19至29任一项所述的方法，其特征在于，所述方法还包括：
    向所述终端设备发送资源激活信息，所述资源激活信息用于指示所述第一资源配置信息中的m个资源配置组合，所述m为小于或等于所述n的正整数。
  34. 根据权利要求33所述的方法，其特征在于，
    所述资源激活信息承载于DCI中；
    或者，
    所述资源激活信息承载于MAC CE中。
  35. 一种资源配置装置，其特征在于，应用于终端设备中，所述装置包括：
    配置信息接收模块，用于接收来自于网络设备的第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
    其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
  36. 根据权利要求35所述的装置，其特征在于，所述第二类资源配置信息包括以下至少一项信息：
    模型使用信息，用于指示所述终端设备使用的AI模型；
    模型运行信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的模型部分；
    模型操作信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的操作部分；
    模型下载信息，用于指示所述终端设备下载的AI模型；
    数据使用信息，用于指示所述终端设备训练AI模型时所使用的训练数据；
    数据上报信息，用于指示所述终端设备上报AI模型的训练结果的频率；
    资源使用信息，用于指示所述终端设备执行AI模型相关操作时所使用的资源量。
  37. 根据权利要求36所述的装置，其特征在于，所述模型使用信息包括：所述终端设备使用的AI模型的标识。
  38. 根据权利要求36或37所述的装置，其特征在于，所述模型运行信息包括：所述终端设备使用的AI模型的模型分割点的信息。
  39. 根据权利要求36至38任一项所述的装置，其特征在于，所述模型操作信息包括：所述终端设备使用的AI模型的操作划分点的信息。
  40. 根据权利要求36至39任一项所述的装置，其特征在于，所述模型下载信息包括：所述终端设备下载的AI模型的标识。
  41. 根据权利要求36至40任一项所述的装置，其特征在于，所述数据使用信息包括：所述终端设备训练AI模型时所使用的训练数据量。
  42. 根据权利要求36至41任一项所述的装置，其特征在于，所述数据上报信息包括：所述终端设备上报AI模型的训练结果的上报周期。
  43. 根据权利要求36至42任一项所述的装置，其特征在于，所述资源使用信息包括：所述终端设备执行AI模型相关操作时所使用的算力。
  44. 根据权利要求35至43任一项所述的装置，其特征在于，所述第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
  45. 根据权利要求35至44任一项所述的装置，其特征在于，
    所述第一资源配置信息承载于无线资源控制RRC配置信息中；
    或者，
    所述第一资源配置信息承载于系统信息中。
  46. 根据权利要求35至45任一项所述的装置，其特征在于，所述装置还包括：
    配置组合选择模块，用于从所述第一资源配置信息中选择第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
  47. 根据权利要求35至45任一项所述的装置，其特征在于，所述装置还包括：
    指示信息接收模块，用于接收来自于网络设备的资源指示信息，所述资源指示信息用于指示所述第一资源配置信息中的第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配；
    配置组合选择模块，用于根据所述资源指示信息，选择所述第一资源配置组合。
  48. 根据权利要求47所述的装置，其特征在于，
    所述资源指示信息承载于下行控制信息DCI中；
    或者，
    所述资源指示信息承载于媒体接入控制的控制单元MAC CE中。
  49. 根据权利要求35至45任一项所述的装置，其特征在于，所述装置还包括：
    激活信息接收模块，用于接收来自于网络设备的资源激活信息，所述资源激活信息用于指示所述第一资源配置信息中的m个资源配置组合，所述m为小于或等于所述n的正整数；
    配置组合激活模块，用于根据所述资源激活信息，激活所述m个资源配置组合。
  50. 根据权利要求49所述的装置，其特征在于，所述装置还包括：
    配置组合选择模块，用于从所述m个资源配置组合中选择第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
  51. 根据权利要求49或50所述的装置，其特征在于，
    所述资源激活信息承载于DCI中；
    或者，
    所述资源激活信息承载于MAC CE中。
  52. 根据权利要求46、47和50任一项所述的装置，其特征在于，所述终端设备的设备运行信息包括：所述终端设备的待用无线资源和所述终端设备的待用算力；
    所述第一资源配置组合与所述终端设备的设备运行信息相匹配，包括：所述第一资源配置组合中的第一类资源配置信息与所述终端设备的待用无线资源相匹配，且，所述第一资源配置组合中的第二类资源配置信息与所述终端设备的待用算力相匹配。
  53. 一种资源配置装置，其特征在于，应用于网络设备中，所述装置包括：
    配置信息发送模块，用于向终端设备发送第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
    其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
  54. 根据权利要求53所述的装置，其特征在于，所述第二类资源配置信息包括以下至少一项信息：
    模型使用信息，用于指示所述终端设备使用的AI模型；
    模型运行信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的模型部分；
    模型操作信息，用于指示在所述终端设备使用的AI模型中，所述终端设备负责运行的操作部分；
    模型下载信息，用于指示所述终端设备下载的AI模型；
    数据使用信息，用于指示所述终端设备训练AI模型时所使用的训练数据；
    数据上报信息，用于指示所述终端设备上报AI模型的训练结果的频率；
    资源使用信息，用于指示所述终端设备执行AI模型相关操作时所使用的资源量。
  55. 根据权利要求54所述的装置，其特征在于，所述模型使用信息包括：所述终端设备使用的AI模型的标识。
  56. 根据权利要求54或55所述的装置，其特征在于，所述模型运行信息包括：所述终端设备使用的AI模型的模型分割点的信息。
  57. 根据权利要求54至56任一项所述的装置，其特征在于，所述模型操作信息包括：所述终端设备使用的AI模型的操作划分点的信息。
  58. 根据权利要求54至57任一项所述的装置，其特征在于，所述模型下载信息包括：所述终端设备下载的AI模型的标识。
  59. 根据权利要求54至58任一项所述的装置，其特征在于，所述数据使用信息包括：所述终端设备训练AI模型时所使用的训练数据量。
  60. 根据权利要求54至59任一项所述的装置，其特征在于，所述数据上报信息包括：所述终端设备上报AI模型的训练结果的上报周期。
  61. 根据权利要求54至60任一项所述的装置，其特征在于，所述资源使用信息包括：所述终端设备执行AI模型相关操作时所使用的算力。
  62. 根据权利要求53至61任一项所述的装置，其特征在于，所述第一类资源配置信息包括以下至少一项：时域资源信息、频域资源信息、空间域资源信息、码域资源信息。
  63. 根据权利要求53至62任一项所述的装置，其特征在于，
    所述第一资源配置信息承载于无线资源控制RRC配置信息中；
    或者，
    所述第一资源配置信息承载于系统信息中。
  64. 根据权利要求53至63任一项所述的装置，其特征在于，所述装置还包括：
    指示信息发送模块，用于向所述终端设备发送资源指示信息，所述资源指示信息用于指示所述第一资源配置信息中的第一资源配置组合，所述第一资源配置组合与所述终端设备的设备运行信息相匹配。
  65. 根据权利要求64所述的装置，其特征在于，
    所述资源指示信息承载于下行控制信息DCI中；
    或者，
    所述资源指示信息承载于媒体接入控制的控制单元MAC CE中。
  66. 根据权利要求64或65所述的装置，其特征在于，所述终端设备的设备运行信息包括：所述终端设备的待用无线资源和所述终端设备的待用算力；
    所述第一资源配置组合与所述终端设备的设备运行信息相匹配，包括：所述第一资源配置组合中的第一类资源配置信息与所述终端设备的待用无线资源相匹配，且，所述第一资源配置组合中的第二类资源配置信息与所述终端设备的待用算力相匹配。
  67. 根据权利要求53至63任一项所述的装置，其特征在于，所述装置还包括：
    激活信息发送模块，用于向所述终端设备发送资源激活信息，所述资源激活信息用于指示所述第一资源配置信息中的m个资源配置组合，所述m为小于或等于所述n的正整数。
  68. 根据权利要求67所述的装置，其特征在于，
    所述资源激活信息承载于DCI中；
    或者，
    所述资源激活信息承载于MAC CE中。
  69. 一种终端设备，其特征在于，所述终端设备包括处理器和与所述处理器相连的收发器；其中：
    所述收发器，用于接收来自于网络设备的第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
    其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
  70. 一种网络设备，其特征在于，所述网络设备包括处理器和与所述处理器相连的收发器；其中：
    所述收发器，用于向终端设备发送第一资源配置信息，所述第一资源配置信息包括n个资源配置组合，所述资源配置组合包括第一类资源配置信息和第二类资源配置信息，所述n为正整数；
    其中，所述第一类资源配置信息用于指示无线资源配置，所述第二类资源配置信息用于指示人工智能AI资源配置。
  71. 一种计算机可读存储介质，其特征在于，所述存储介质中存储有计算机程序，所述计算机程序用于被终端设备的处理器执行，以实现如权利要求1至18任一项所述的资源配置方法。
  72. 一种计算机可读存储介质，其特征在于，所述存储介质中存储有计算机程序，所述计算机程序用于被网络设备的处理器执行，以实现如权利要求19至34任一项所述的资源配置方法。
PCT/CN2020/084306 2020-04-10 2020-04-10 资源配置方法、装置、设备及存储介质 WO2021203437A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
EP20930151.4A EP4135438A4 (en) 2020-04-10 2020-04-10 RESOURCE ALLOCATION METHOD, DEVICE, APPARATUS AND RECORDING MEDIUM
CN202080099659.0A CN115380594A (zh) 2020-04-10 2020-04-10 资源配置方法、装置、设备及存储介质
PCT/CN2020/084306 WO2021203437A1 (zh) 2020-04-10 2020-04-10 资源配置方法、装置、设备及存储介质
US17/961,964 US20230038071A1 (en) 2020-04-10 2022-10-07 Resource configuration method and apparatus, device, and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/084306 WO2021203437A1 (zh) 2020-04-10 2020-04-10 资源配置方法、装置、设备及存储介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/961,964 Continuation US20230038071A1 (en) 2020-04-10 2022-10-07 Resource configuration method and apparatus, device, and storage medium

Publications (1)

Publication Number Publication Date
WO2021203437A1 true WO2021203437A1 (zh) 2021-10-14

Family

ID=78022355

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/084306 WO2021203437A1 (zh) 2020-04-10 2020-04-10 资源配置方法、装置、设备及存储介质

Country Status (4)

Country Link
US (1) US20230038071A1 (zh)
EP (1) EP4135438A4 (zh)
CN (1) CN115380594A (zh)
WO (1) WO2021203437A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022233294A1 (zh) * 2021-05-07 2022-11-10 展讯通信（上海）有限公司 通信方法、装置和电子设备
WO2023097870A1 (zh) * 2021-12-03 2023-06-08 中国信息通信研究院 一种无线通信人工智能处理方法和设备
WO2023230969A1 (zh) * 2022-06-01 2023-12-07 北京小米移动软件有限公司 人工智能模型的确定方法及装置、通信设备及存储介质
WO2024011581A1 (zh) * 2022-07-15 2024-01-18 华为技术有限公司 一种通信方法及装置

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20240053405A (ko) * 2022-10-17 2024-04-24 고려대학교 산학협력단 서버리스 엣지 컴퓨팅 환경에서의 동적 분할 컴퓨팅 방법

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108347776A (zh) * 2017-01-25 2018-07-31 华为技术有限公司 一种通信系统中资源分配的方法及设备
US20190042884A1 (en) * 2017-12-28 2019-02-07 Francesc Guim Bernat Malleable fabric attached virtual artificial intelligence (ai) training appliances
WO2019047156A1 (zh) * 2017-09-08 2019-03-14 富士通株式会社 资源配置方法、确定方法及其装置、通信系统
CN109615022A (zh) * 2018-12-20 2019-04-12 上海智臻智能网络科技股份有限公司 模型上线配置方法和装置
CN109937596A (zh) * 2016-11-11 2019-06-25 Oppo广东移动通信有限公司 调度终端设备的方法、网络设备和终端设备
CN110767220A (zh) * 2019-10-16 2020-02-07 腾讯科技（深圳）有限公司 一种智能语音助手的交互方法、装置、设备及存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109937596A (zh) * 2016-11-11 2019-06-25 Oppo广东移动通信有限公司 调度终端设备的方法、网络设备和终端设备
CN108347776A (zh) * 2017-01-25 2018-07-31 华为技术有限公司 一种通信系统中资源分配的方法及设备
WO2019047156A1 (zh) * 2017-09-08 2019-03-14 富士通株式会社 资源配置方法、确定方法及其装置、通信系统
US20190042884A1 (en) * 2017-12-28 2019-02-07 Francesc Guim Bernat Malleable fabric attached virtual artificial intelligence (ai) training appliances
CN109615022A (zh) * 2018-12-20 2019-04-12 上海智臻智能网络科技股份有限公司 模型上线配置方法和装置
CN110767220A (zh) * 2019-10-16 2020-02-07 腾讯科技（深圳）有限公司 一种智能语音助手的交互方法、装置、设备及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4135438A4 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022233294A1 (zh) * 2021-05-07 2022-11-10 展讯通信（上海）有限公司 通信方法、装置和电子设备
WO2023097870A1 (zh) * 2021-12-03 2023-06-08 中国信息通信研究院 一种无线通信人工智能处理方法和设备
WO2023230969A1 (zh) * 2022-06-01 2023-12-07 北京小米移动软件有限公司 人工智能模型的确定方法及装置、通信设备及存储介质
WO2024011581A1 (zh) * 2022-07-15 2024-01-18 华为技术有限公司 一种通信方法及装置

Also Published As

Publication number Publication date
EP4135438A1 (en) 2023-02-15
CN115380594A (zh) 2022-11-22
EP4135438A4 (en) 2023-06-07
US20230038071A1 (en) 2023-02-09

Similar Documents

Publication Publication Date Title
WO2021203437A1 (zh) 资源配置方法、装置、设备及存储介质
JP6888082B2 (ja) アップリンク通信波形選択の容易化
CN112470525B (zh) 无线网络中的网络切片约束的传输
WO2021142637A1 (zh) 人工智能操作处理方法、装置、系统、终端及网络设备
CN114172765B (zh) 一种无线通信人工智能信道估计方法和设备
CN106470492B (zh) 资源占用方法、装置及用户设备
CN105490790B (zh) 载波聚合中辅载波的配置方法和装置
US20220264675A1 (en) System and method for processing data
JP2024026381A (ja) 探索空間モニタリング方法及び装置
CN116034381A (zh) 通信方法和通信装置
CN109803422A (zh) 一种资源激活的方法及相关设备
CN115699842A (zh) 模型管理方法、系统、装置、通信设备及存储介质
US20240211770A1 (en) Communication method and apparatus
US20240223652A1 (en) Artificial intelligence model download method, apparatus, and system
CN115442876A (zh) 终端的节电方法、装置、无线接入设备、电子设备和介质
US11943793B2 (en) AI engine-supporting downlink radio resource scheduling method and apparatus
WO2023221111A1 (zh) Ue能力的上报方法、装置、设备和介质
CN116233896B (zh) 无线通信系统中用户设备传送消息的方法和系统
WO2023108497A1 (zh) 能力指示方法、装置、设备及存储介质
US20240056904A1 (en) External exposure of control plane to user plane switch parameter
CN112738245B (zh) 设备应答信息的确定方法及装置、存储介质及电子装置
WO2022247739A1 (zh) 一种数据传输方法和相关装置
US20240056869A1 (en) Method and apparatus for service using tethering in wireless communication system
KR20240045903A (ko) 종단간 어플리케이션-레벨 지연의 성능보장을 위한 무선자원 할당 방법 및 장치
CN117998601A (zh) 一种数据传输方法、装置及存储介质

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: 20930151

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2020930151

Country of ref document: EP

Effective date: 20221109