WO2023115251A1 - 无线通信中保障ai模型有效性方法、装置、终端及介质 - Google Patents
无线通信中保障ai模型有效性方法、装置、终端及介质 Download PDFInfo
- Publication number
- WO2023115251A1 WO2023115251A1 PCT/CN2021/139547 CN2021139547W WO2023115251A1 WO 2023115251 A1 WO2023115251 A1 WO 2023115251A1 CN 2021139547 W CN2021139547 W CN 2021139547W WO 2023115251 A1 WO2023115251 A1 WO 2023115251A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- model
- terminal
- management process
- execute
- validity
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 265
- 238000004891 communication Methods 0.000 title claims abstract description 74
- 238000010200 validation analysis Methods 0.000 title abstract 4
- 230000008569 process Effects 0.000 claims abstract description 176
- 230000006870 function Effects 0.000 claims description 18
- 230000008859 change Effects 0.000 claims description 16
- 230000000737 periodic effect Effects 0.000 claims description 14
- 238000005265 energy consumption Methods 0.000 claims description 8
- 238000013473 artificial intelligence Methods 0.000 claims description 6
- 230000006978 adaptation Effects 0.000 claims description 5
- 102100022734 Acyl carrier protein, mitochondrial Human genes 0.000 claims 4
- 101000678845 Homo sapiens Acyl carrier protein, mitochondrial Proteins 0.000 claims 4
- 238000013461 design Methods 0.000 description 31
- 239000010410 layer Substances 0.000 description 16
- 230000000875 corresponding effect Effects 0.000 description 14
- 238000004590 computer program Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 238000010295 mobile communication Methods 0.000 description 7
- 238000001228 spectrum Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 230000003993 interaction Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000010365 information processing Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000000275 quality assurance Methods 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Definitions
- the present application relates to the field of communication, and in particular to a method, device, terminal and medium for ensuring the validity of an AI model in wireless communication.
- AI Artificial Intelligence, artificial intelligence
- the related technology adopts the traditional wireless communication method, based on the theoretical modeling of the actual communication environment, and determines the transmission mode between the terminal and the network device according to the established model.
- Embodiments of the present application provide a method, device, terminal, and medium for ensuring the validity of an AI model in wireless communication, and provide a method for ensuring that the AI model is always valid.
- a method for ensuring the validity of an artificial intelligence AI model in wireless communication is provided, the method is executed by a terminal, and the method includes:
- the effectiveness management process of the first AI model is executed.
- a device for ensuring the validity of an artificial intelligence AI model in wireless communication comprising:
- the execution module is configured to execute the effectiveness management process of the first AI model when the failure condition of the model is met.
- a terminal is provided, and the terminal includes:
- transceiver connected to the processor
- memory for storing processor-executable instructions
- the processor is configured to load and execute executable instructions to implement any of the methods for ensuring the validity of the AI model in wireless communication described above.
- a chip is provided, and the chip is used to implement any method for ensuring the validity of an AI model in wireless communication described above.
- a computer-readable storage medium stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, all The at least one program, the code set or the instruction set is loaded and executed by the processor to implement any of the methods for ensuring the validity of an AI model in wireless communication described above.
- the effectiveness management process can be executed according to actual needs, and the AI model can be processed accordingly, so as to ensure the validity of the AI model and improve the performance of wireless communication.
- Fig. 1 is a schematic diagram of a communication system according to an exemplary embodiment
- Fig. 2 is a flowchart showing a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 3 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 4 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 5 is a flowchart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 6 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 7 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 8 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 9 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 10 is a flowchart showing a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 11 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 12 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 13 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 14 is a flow chart of a method for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 15 is a block diagram of a device for ensuring the validity of an AI model in wireless communication according to an exemplary embodiment
- Fig. 16 is a schematic structural diagram of a terminal according to an exemplary embodiment
- Fig. 17 is a schematic structural diagram of a network device according to an exemplary embodiment.
- the network architecture and business scenarios described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute limitations on the technical solutions provided by the embodiments of the present application.
- the evolution of the technology and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.
- the technical solution of the embodiment of the present application can be applied to various communication systems, such as: Global System of Mobile communication (Global System of Mobile communication, GSM) system, code division multiple access (Code Division Multiple Access, CDMA) system, broadband code division multiple access (Wideband Code Division Multiple Access, WCDMA) system, General Packet Radio Service (GPRS), Long Term Evolution (LTE) system, Advanced long term evolution (LTE-A) system , New Radio (NR) system, evolution system of NR system, LTE (LTE-based access to unlicensed spectrum, LTE-U) system on unlicensed spectrum, NR (NR-based access to unlicensed spectrum) on unlicensed spectrum unlicensed spectrum (NR-U) system, Non-Terrestrial Networks (NTN) system, Universal Mobile Telecommunications System (UMTS), Wireless Local Area Networks (WLAN), Wireless Fidelity (Wireless Fidelity, WiFi), fifth-generation communication (5th-Generation, 5G) system or other communication systems, etc.
- GSM Global System of Mobile
- Fig. 1 shows a schematic diagram of a mobile communication system provided by an embodiment of the present application.
- the mobile communication system may include: a terminal 10 and a network device 20 .
- the terminal 10 may include various handheld devices with mobile 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 (User Equipment, UE), mobile station ( Mobile Station, MS) and so on.
- UE User Equipment
- MS Mobile Station
- the network device 20 is a device deployed in an access network for providing a mobile communication function for the terminal 10 .
- the network device 20 may include various forms of macro base stations, micro base stations, relay stations, access points, location management function entities (Location Management Function, LMF) and so on.
- LMF Location Management Function
- the names of devices with access network device functions may be different.
- gNodeB or gNB are called gNodeB or gNB.
- the term "network equipment" may change as communications technology evolves.
- network devices For the convenience of description, in the embodiment of the present application, the above-mentioned devices that provide mobile communication functions for the terminal 10 are collectively referred to as network devices.
- a connection can be established between the network device 20 and the terminal 10 through an air interface, so as to communicate through the connection, including signaling and data interaction.
- the number of network devices 20 may be multiple, and communication between two adjacent network devices 20 may also be performed in a wired or wireless manner.
- the terminal 10 can switch between different network devices 20, that is, establish connections with different network devices 20.
- the network device 20 is regarded as an access network device.
- the "5G NR system" in the embodiments of the present disclosure may also be called a 5G system or an NR system, but those skilled in the art can understand its meaning.
- the technical solution described in the embodiments of the present disclosure can be applied to the 5G NR system, and can also be applied to the subsequent evolution system of the 5G NR system.
- Fig. 2 shows a flow chart of a method for ensuring the validity of an AI model in wireless communication provided by an exemplary embodiment of the present application. This embodiment is illustrated by taking the application in the communication system shown in Figure 1 as an example, and the method includes:
- step 201 when the failure condition of the model is met, the validity management process of the first AI model is executed.
- Model failure conditions include but are not limited to the following 10 situations:
- the terminal receives a system message broadcast by the network device; in the case that the area identifier in the system message is different from the model effective area identifier of the first AI model, execute the validity management process of the first AI model.
- the area identifier in the system message is used to indicate the area where the terminal is located.
- the effective area of the first AI model is at least one of a tracking area (tracking area), a radio access network RAN area, and a self-defined area, and the effective area includes at least one cell.
- the validity period of the first AI model is determined according to a first timer, and the first timer is used to ensure the validity of the first AI model.
- the first timer of the first AI model expires, the first AI model is considered invalid, and further, the effectiveness management process of the first AI model is executed.
- the first timer is started when the terminal receives the configuration information of the first timer, or the first timer is started when the terminal starts to use the first AI model, or the first timer is started when the terminal receives the first AI model is activated.
- the first timer is stopped.
- the first timer may be configured by the network device for the terminal, or the first timer is configured by the terminal itself.
- the first timer is determined based on at least one of the effective duration of the first AI model, service type, operating track/area of the terminal, and network load/energy consumption.
- the periodic timer may be configured by the network device to the terminal, or the periodic timer may be configured by the terminal itself.
- the periodic timer is started/restarted when the terminal receives the configuration information of the periodic timer, or the periodic timer is started/restarted when the terminal starts to use the first AI model, or the periodic timer is started/restarted when the terminal Started/restarted when the first AI model is received.
- the periodic timer may be configured by the network device for the terminal, or the periodic timer may be configured by the terminal itself.
- the period timer is determined based on at least one of the effective duration of the first AI model, service type, terminal running track/area, and network load/energy consumption.
- the first threshold and/or the second threshold may be preconfigured by the network device, or may be customized by the terminal.
- the third threshold and/or the fourth threshold may be preconfigured by the network device, or may be customized by the terminal.
- the fifth threshold and/or the sixth threshold may be preconfigured by the network device, or may be customized by the terminal.
- the terminal when the terminal needs to adjust the matching AI model based on service requirements, the terminal executes the validity management process of the first AI model.
- the handover decision based on the first AI model results in N consecutive handover failures, it is considered that the accuracy of the first AI model is insufficient, and it is not suitable for the optimization of the current scene, and it is necessary to execute The effectiveness management process of the first AI model.
- the effectiveness management process of the first AI model is executed.
- the validity management process of the first AI model is executed.
- the scene characteristics of the wireless environment include different indoor environments/outdoor environments, dense cells/open field, LOS (Line Of Sight, line of sight)/NLOS (Non Line Of Sight, non-line of sight), high speed/low speed medium at least one of .
- the validity management process of the first AI model is executed.
- the channel environment index feature of the wireless environment includes at least one of delay power spectrum information, multipath information, angle information, and speed information.
- the validity management process of the first AI model is executed.
- the validity management process of the first AI model is executed.
- the validity management process of the first AI model is executed.
- the terminal can autonomously execute the effectiveness management process of the first AI model.
- the terminal autonomously executes the validity management process of the first AI model.
- the terminal autonomously executes the effectiveness management process of the first AI model, including at least one of autonomously updating the first AI model, autonomously switching the first AI model to the second AI model, and autonomously stopping using the first AI model.
- the terminal autonomously updates the configuration information of the first AI model.
- the terminal updates the first AI model according to the updated configuration information.
- the configuration information of the first AI model includes but is not limited to at least one of structure information and parameter information of the first AI model.
- at least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the terminal stores at least two sets of candidate AI models.
- the terminal autonomously determines the second AI model from at least two sets of AI models, and switches the first AI model to the second AI model.
- At least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the at least two sets of candidate AI models may be pre-configured by the network device to the terminal, or configured by the terminal itself.
- the terminal autonomously stops using the first AI model when the model invalidation condition is met.
- the terminal stops the first timer related to the first AI model, and deletes the configuration information of the first AI model.
- the terminal may also execute the validity management process of the first AI model by sending request information to the network device, where the request information is used to request the network device to execute the validity management process of the first AI model.
- the request information is reported to the network device; according to the instruction information provided by the network device, the validity management process of the first AI model is executed.
- the terminal executes the effectiveness management process of the first AI model according to the instruction information provided by the network device, including at least one of updating the first AI model, switching the first AI model to the second AI model, and stopping using the first AI model. kind.
- the request information includes configuration information of the first AI model and/or auxiliary information for updating the model.
- the terminal reports the request information to the network device.
- the terminal updates the first AI model according to the configuration information of the first AI model provided by the network device and/or the auxiliary information used for model updating.
- the configuration information of the first AI model includes structure information and/or parameter information of the first AI model.
- the terminal stores at least two sets of candidate AI models, and the indication information includes an identifier of the second AI model.
- the terminal reports the request information to the network device.
- the terminal determines the second AI model from at least two sets of candidate AI models according to the identifier of the second AI model provided by the network device; and switches the first AI model to the second AI model.
- the request information includes a suggested AI model identifier, so that the network device determines the AI model to be switched based on the terminal's suggestion.
- the terminal reports the request information to the network device.
- the terminal stops using the first AI model according to the instruction information provided by the network device.
- the terminal stops the first timer related to the first AI model, and deletes the configuration information of the first AI model.
- the terminal reports a completion message to the network device, where the completion message is used to indicate that the terminal has completed the validity management process of the first AI model.
- the completion message includes at least one of the updated structure information and/or parameter information of the first AI model, an identifier of the second AI model, and a message to stop using the first AI model.
- the manner of updating the first AI model includes at least one of adjusting the number of layers of the first AI model, updating network parameters and/or network topology of the first AI model, and clipping or compressing the first AI model.
- the first AI model is used to implement mobility enhancement, CSI (Channel State Information, channel state information) feedback, channel estimation, load balancing, terminal/network energy saving, beam management, terminal trajectory prediction, service prediction, positioning enhancement, wireless resource At least one of management.
- the mobility enhancement is used to enhance the service continuity of the terminal in the mobile state, such as reducing handover interruption by selecting a more suitable target cell.
- the CSI feedback is used to realize the interaction of channel state information between the terminal and the network device.
- Channel estimation is used in the process of estimating model parameters of the first AI model from channel data.
- Load balancing is used to balance the business and distribute it to multiple operating units for operation.
- Terminal/network energy saving is used to reduce invalid energy consumption of terminals/networks.
- Beam management is used to manage the beams that transmit signals.
- the trajectory prediction of the terminal is used for the mobile trajectory of the terminal.
- Service prediction is used to predict the service of the terminal.
- Positioning enhancements are used to determine the location of the terminal.
- Radio resource management is used to provide service quality assurance for terminals in the network under the condition of limited bandwidth.
- the first AI model runs, trains or infers in RRC (Radio Resource Control, radio resource control), SDAP (Service Data Adaptation Protocol, service data adaptation protocol), PDCP (Packet Data Convergence Protocol, packet data convergence protocol), RLC ( Radio Link Control, wireless link control), MAC (Medium Access Control, media access control), PHY (PHYsical, physical layer) in any layer.
- RRC Radio Resource Control, radio resource control
- SDAP Service Data Adaptation Protocol, service data adaptation protocol
- PDCP Packet Data Convergence Protocol, packet data convergence protocol
- RLC Radio Link Control, wireless link control
- MAC Medium Access Control, media access control
- PHY Physical layer
- the function optimized by the first AI model is a function corresponding to any layer of RRC, SDAP, PDCP, RLC, MAC, and PHY.
- the validity management process may be executed according to actual requirements, and the AI model may be processed accordingly, so as to ensure the validity of the AI model and improve the performance of wireless communication.
- FIG. 3 shows an example of this application A flow chart of a method for ensuring the validity of an AI model in wireless communication provided by an exemplary embodiment. This embodiment is illustrated by taking the application in the communication system shown in Figure 1 as an example, and the method includes:
- a network device broadcasts a system message to a terminal.
- system message includes an area identifier.
- the effective area of the first AI model is at least one of a tracking area (tracking area), a radio access network RAN area, and a self-defined area, and the effective area includes at least one cell.
- the network device broadcasts the system message to the terminal.
- step 302 when the model invalidation condition is satisfied, the terminal autonomously executes the validity management process of the first AI model.
- the model invalidation condition means that the area identifier in the system message is different from the model effective area identifier of the first AI model.
- the terminal autonomously updates the first AI model according to model input information during use of the first AI model.
- the first AI model is an AI model for trajectory prediction.
- the terminal may update the first AI model based on historical trajectory information input during use of the first AI model.
- the manner of updating the first AI model includes at least one of adjusting the number of layers of the first AI model, updating network parameters and/or network topology of the first AI model, and clipping or compressing the first AI model.
- the terminal stores at least two sets of candidate AI models.
- the terminal determines the second AI model from at least two sets of candidate AI models.
- the terminal autonomously switches the first AI model to the second AI model.
- at least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the at least two sets of candidate AI models may be pre-configured by the network device to the terminal, or may be configured by the terminal itself.
- At least two sets of candidate AI models are associated with the region.
- the terminal autonomously stops using the first AI model when the model invalidation condition is met. Since the model effective region identifier of the first AI model does not match the region identifier provided by the system message, the first AI model will no longer be applicable, and the terminal can autonomously stop using the first AI model, and rely on pre-configured rules or algorithms to perform corresponding actions. process, or execute the corresponding process based on the existing protocol specification.
- the terminal stops using the first timer related to the first AI model, and deletes configuration information of the first AI model.
- step 303 the terminal reports a completion message to the network device.
- the completion message is used to indicate that the terminal completes the validity management process of the first AI model.
- the completion message includes at least one of the updated structure information and/or parameter information of the first AI model, an identifier of the second AI model, and a message to stop using the first AI model.
- the completion message can be transmitted through at least one of RRC message or MAC CE (Media Access Control Control Element, media access layer control unit) or UCI (Uplink Control Information, uplink control information).
- RRC message Media Access Control Control Element, media access layer control unit
- UCI Uplink Control Information, uplink control information
- the terminal reports a completion message to the network device.
- this step is optional and may or may not be performed.
- the terminal when the terminal leaves the effective area of the first AI model, it will automatically update, switch or return the first AI model, so as to improve the performance of wireless communication.
- the model invalidation condition is taken as an example where the terminal leaves the valid area of the first AI model.
- the terminal needs to report request information to the network device to execute the validity management process of the first AI model, as shown in FIG. 5
- a flow chart of a method for ensuring the validity of an AI model in wireless communication provided by an exemplary embodiment of the present application is shown. This embodiment is illustrated by taking the application in the communication system shown in Figure 1 as an example, and the method includes:
- a network device broadcasts a system message to a terminal.
- system message includes an area identifier.
- the effective area of the first AI model is at least one of a tracking area (tracking area), a radio access network RAN area, and a self-defined area, and the effective area includes at least one cell.
- the network device broadcasts the system message to the terminal.
- step 502 when the model failure condition is met, the terminal reports request information to the network device.
- the model invalidation condition means that the area identifier in the system message is different from the model effective area identifier of the first AI model.
- the request information may include an identifier of the proposed AI model. So that the network device determines the AI model to be switched based on the terminal's suggestion.
- the terminal reports request information to the network device.
- step 503 the network device provides indication information to the terminal.
- the indication information includes that the indication information includes configuration information of the first AI model and/or auxiliary information for model updating, so that the terminal updates the first AI model according to the indication information.
- the configuration information of the first AI model includes structure information and/or parameter information of the first AI model.
- the indication information may include the identifier of the second AI model, so that the terminal determines the AI model to be switched according to the identifier of the second AI model.
- At least two sets of AI models may be preconfigured on the network device.
- At least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the instruction information is used to instruct the terminal to stop using the first AI model.
- the network device provides indication information to the terminal.
- step 504 the terminal executes the validity management process of the first AI model according to the instruction information provided by the network device.
- the terminal when the indication information includes configuration information of the first AI model and/or auxiliary information for model update, the terminal provides the configuration information of the first AI model and/or the auxiliary information for model update provided by the network device
- the auxiliary information is to update the first AI model.
- the manner of updating the first AI model includes at least one of adjusting the number of layers of the first AI model, updating network parameters and/or network topology of the first AI model, and clipping or compressing the first AI model.
- the indication information includes an identifier of the second AI model; the terminal stores at least two sets of candidate AI models.
- the terminal stores at least two sets of candidate AI models.
- the terminal autonomously switches the first AI model to the second AI model.
- At least two sets of candidate AI models are associated with the region.
- the terminal stops using the first AI model according to the instruction information provided by the network device. Since the terminal has left the effective area of the first AI model, the first AI model will no longer be applicable, and the terminal can stop using the first AI model according to the instruction information of the network device, and rely on pre-configured rules or algorithms to execute the corresponding process, or Execute the corresponding process based on the existing protocol specification.
- the terminal stops using the first timer related to the first AI model, and deletes configuration information of the first AI model.
- step 505 the terminal reports a completion message to the network device.
- the completion message is used to indicate that the terminal completes the validity management process of the first AI model.
- the completion message can be transmitted by at least one of RRC message or MAC CE or UCI.
- the terminal reports a completion message to the network device.
- this step is optional and may or may not be performed.
- the terminal when the terminal leaves the effective area of the first AI model, it will report request information to the network device for updating, switching or returning the first AI model, so as to improve the performance of wireless communication.
- FIG. 7 shows an exemplary embodiment of the present application
- the flow chart of the method for ensuring the validity of AI models in wireless communication is provided. This embodiment is illustrated by taking the application in the communication system shown in Figure 1 as an example, and the method includes:
- step 701 the terminal determines that the first timer expires.
- the first timer is used to ensure the validity of the first AI model.
- the first timer is determined based on at least one of the effective duration of the first AI model, service type, operating track/area of the terminal, and network load/energy consumption.
- the first timer is started when the terminal receives the configuration information of the first timer, or the first timer is started when the terminal starts to use the first AI model, or the first timer is started when the terminal receives the first AI model An AI model is activated.
- the first timer may be configured by the network device for the terminal, or the first timer is configured by the terminal itself.
- step 702 the terminal autonomously executes the effectiveness management process of the first AI model.
- the terminal autonomously updates the first AI model according to model input information during use of the first AI model.
- the first AI model is an AI model for trajectory prediction.
- the terminal may update the first AI model based on historical trajectory information input during use of the first AI model.
- the manner of updating the first AI model includes at least one of adjusting the number of layers of the first AI model, updating network parameters and/or network topology of the first AI model, and clipping or compressing the first AI model.
- the terminal stores at least two sets of candidate AI models.
- the terminal determines the second AI model from at least two sets of candidate AI models.
- the terminal autonomously switches the first AI model to the second AI model.
- at least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the at least two sets of candidate AI models may be preconfigured by the network device to the terminal.
- the terminal autonomously stops using the first AI model. Since the first AI model has expired, the terminal can autonomously stop using the first AI model, and rely on pre-configured rules or algorithms to execute the corresponding process, or execute the corresponding process based on the existing protocol specification. Optionally, the terminal stops using the first timer, and deletes configuration information of the first AI model.
- step 703 the terminal reports a completion message to the network device.
- the completion message is used to indicate that the terminal completes the validity management process of the first AI model.
- the completion message includes at least one of the updated structure information and/or parameter information of the first AI model, an identifier of the second AI model, and a message to stop using the first AI model.
- the completion message can be transmitted by at least one of RRC message or MAC CE or UCI.
- the terminal reports a completion message to the network device.
- this step is optional and may or may not be performed.
- the first AI model when the first timer expires, the first AI model will be automatically updated, switched or rolled back, so as to improve the performance of wireless communication.
- FIG. 9 shows the A flow chart of a method for ensuring the validity of an AI model in wireless communication provided by an exemplary embodiment of the application. This embodiment is illustrated by taking the application in the communication system shown in Figure 1 as an example, and the method includes:
- step 901 the terminal determines that the first timer expires.
- the first timer is used to ensure the validity of the first AI model.
- the first timer is determined based on at least one of the effective duration of the first AI model, service type, operating track/area of the terminal, and network load/energy consumption.
- the first timer is started when the terminal receives the configuration information of the first timer, or the first timer is started when the terminal starts to use the first AI model, or the first timer is started when the terminal receives the first AI model An AI model is activated.
- the first timer may be configured by the network device for the terminal, or the first timer is configured by the terminal itself.
- step 902 the terminal reports request information to the network device.
- the request information may include an identifier of the recommended AI model, so that the network device determines the AI model to be switched based on the terminal's suggestion.
- the terminal reports request information to the network device.
- step 903 the network device provides indication information to the terminal.
- the indication information includes configuration information of the first AI model and/or auxiliary information for model updating, so that the terminal updates the first AI model according to the indication information.
- the configuration information of the first AI model includes structural information and/or parameter information of the first AI model.
- the indication information may include the identifier of the second AI model, so that the terminal determines the AI model to be switched according to the identifier of the second AI model.
- At least two sets of AI models may be preconfigured on the network device.
- At least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the network device provides indication information to the terminal.
- step 904 the terminal executes the validity management process of the first AI model according to the instruction information provided by the network device.
- the terminal when the indication information includes configuration information of the first AI model and/or auxiliary information for model update, the terminal provides the configuration information of the first AI model and/or the auxiliary information for model update provided by the network device
- the auxiliary information is to update the first AI model.
- the manner of updating the first AI model includes at least one of adjusting the number of layers of the first AI model, updating network parameters and/or network topology of the first AI model, and clipping or compressing the first AI model.
- the indication information includes an identifier of the second AI model; the terminal stores at least two sets of candidate AI models.
- the terminal stores at least two sets of candidate AI models.
- the terminal autonomously switches the first AI model to the second AI model.
- the terminal stops using the first AI model according to the instruction information provided by the network device. Since the first timer has expired, the first AI model will no longer be applicable, and the terminal can stop using the first AI model according to the instruction information of the network device, and rely on pre-configured rules or algorithms to execute the corresponding process, or based on the existing protocol Execute the corresponding process in a standardized manner.
- the terminal stops using the first timer related to the first AI model, and deletes configuration information of the first AI model.
- step 905 the terminal reports a completion message to the network device.
- the completion message is used to indicate that the terminal completes the validity management process of the first AI model.
- the completion message can be transmitted by at least one of RRC message or MAC CE or UCI.
- the terminal reports a completion message to the network device.
- this step is optional and may or may not be performed.
- the first AI model when the validity period of the first timer expires, the first AI model will report request information to the network device for updating, switching or returning, so as to improve the performance of wireless communication.
- FIG. 11 shows an exemplary The flow chart of the method for ensuring the validity of an AI model in wireless communication provided by the embodiment. This embodiment is illustrated by taking the application in the communication system shown in Figure 1 as an example, and the method includes:
- step 1101 the terminal determines that the accuracy of the first AI model does not meet the model accuracy requirements.
- the model accuracy requirement can be preconfigured by the network device or customized by the terminal.
- the first AI model is an AI model used for trajectory prediction, and the handover decision based on the first AI model results in N consecutive handover failures, it is considered that the accuracy of the first AI model does not support the optimization of the current scene .
- step 1102 the terminal autonomously executes the effectiveness management process of the first AI model.
- the terminal autonomously updates the first AI model according to model input information during use of the first AI model.
- the first AI model is an AI model for trajectory prediction.
- the terminal may update the first AI model based on historical trajectory information input during use of the first AI model.
- the manner of updating the first AI model includes at least one of adjusting the number of layers of the first AI model, updating network parameters and/or network topology of the first AI model, and clipping or compressing the first AI model.
- the terminal stores at least two sets of candidate AI models.
- the terminal determines the second AI model from at least two sets of candidate AI models.
- the terminal autonomously switches the first AI model to the second AI model.
- at least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the at least two sets of candidate AI models may be pre-configured by the network device to the terminal, or may be configured by the terminal itself.
- the terminal autonomously stops using the first AI model. Since the first AI model has been invalidated, the terminal may execute the corresponding process depending on pre-configured rules or algorithms, or execute the corresponding process in a manner based on existing protocol specifications. Optionally, the terminal stops using the first timer related to the first AI model, and deletes configuration information of the first AI model.
- step 1103 the terminal reports a completion message to the network device.
- the completion message is used to indicate that the terminal completes the validity management process of the first AI model.
- the completion message includes at least one of the updated structure information and/or parameter information of the first AI model, an identifier of the second AI model, and a message to stop using the first AI model.
- the completion message can be transmitted by at least one of RRC message or MAC CE or UCI.
- the terminal reports a completion message to the network device.
- this step is optional and may or may not be performed.
- the first AI model when the accuracy of the first AI model does not meet the model accuracy requirements, the first AI model will be updated, switched or returned autonomously, so as to improve the performance of wireless communication.
- the accuracy of the first AI model does not meet the model accuracy requirements as an example.
- the terminal needs to report request information to the network device to execute the effectiveness management process of the first AI model, as shown in FIG. 13 A flow chart of a method for ensuring the validity of an AI model in wireless communication provided by an exemplary embodiment of the present application is shown. This embodiment is illustrated by taking the application in the communication system shown in Figure 1 as an example, and the method includes:
- step 1301 the terminal determines that the accuracy of the first AI model does not meet the model accuracy requirements.
- the model accuracy requirement can be preconfigured by the network device or customized by the terminal.
- the first AI model is an AI model used for trajectory prediction, and the handover decision based on the first AI model results in N consecutive handover failures, it is considered that the accuracy of the first AI model does not support the optimization of the current scene .
- step 1302 the terminal reports request information to the network device.
- the request information may include an identifier of the proposed AI model. So that the network device determines the AI model to be switched based on the terminal's suggestion.
- the terminal reports request information to the network device.
- step 1303 the network device provides indication information to the terminal.
- the indication information includes that the indication information includes configuration information of the first AI model and/or auxiliary information for model updating, so that the terminal updates the first AI model according to the indication information.
- the configuration information of the first AI model includes structure information and/or parameter information of the first AI model.
- the indication information may include the identifier of the second AI model, so that the terminal determines the AI model to be switched according to the identifier of the second AI model.
- At least two sets of AI models may be preconfigured on the network device.
- At least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the network device provides indication information to the terminal.
- step 1304 the terminal executes the validity management process of the first AI model according to the indication information provided by the network device.
- the terminal when the indication information includes configuration information of the first AI model and/or auxiliary information for model update, the terminal provides the configuration information of the first AI model and/or the auxiliary information for model update provided by the network device
- the auxiliary information is to update the first AI model.
- the manner of updating the first AI model includes at least one of adjusting the number of layers of the first AI model, updating network parameters and/or network topology of the first AI model, and clipping or compressing the first AI model.
- the indication information includes an identifier of the second AI model; the terminal stores at least two sets of candidate AI models.
- the terminal stores at least two sets of candidate AI models. According to the identification of the second AI model provided by the network device, determine the second AI model from at least two sets of candidate AI models; and switch the first AI model to the second AI model.
- the terminal stops using the first AI model according to the instruction information provided by the network device. Since the accuracy of the first AI model does not meet the model accuracy requirements, the first AI model will no longer be applicable, and the terminal can stop using the first AI model according to the instruction information of the network device, and rely on pre-configured rules or algorithms to execute the corresponding process , or execute the corresponding process based on the existing protocol specification.
- the terminal stops using the first timer related to the first AI model, and deletes configuration information of the first AI model.
- step 1305 the terminal reports a completion message to the network device.
- the completion message is used to indicate that the terminal completes the validity management process of the first AI model.
- the completion message can be transmitted by at least one of RRC message or MAC CE or UCI.
- the terminal reports a completion message to the network device.
- this step is optional and may or may not be performed.
- the first AI model when the accuracy of the first AI model does not meet the model accuracy requirements, the first AI model will be requested to the network device to update, switch or return, so as to improve the performance of wireless communication.
- Fig. 15 shows a block diagram of an apparatus for ensuring the validity of an AI model in wireless communication provided by an exemplary embodiment of the present application.
- the device 150 includes:
- the execution module 151 is configured to execute the validity management process of the first AI model when the failure condition of the model is satisfied.
- the execution module 151 is further configured to execute the validity management process of the first AI model when the terminal leaves the valid area of the first AI model; Or, when the validity period of the first AI model expires, execute the validity management process of the first AI model; or, execute the first AI model based on the periodic update of the first AI model
- the validity management process of the model or, when the channel quality of the terminal meets the first threshold condition, execute the validity management process of the first AI model; or, when the available computing power of the terminal meets the first threshold condition
- execute the validity management process of the first AI model or, in the case that the storage capacity of the terminal meets the third threshold condition, execute the validity management process of the first AI model ; or, when the service type of the terminal changes or the quality of service QoS changes, execute the effectiveness management process of the first AI model; or, when the accuracy of the first AI model does not meet the model In the case of accuracy requirements, execute the validity management process of the first AI model; or, in the case of
- the receiving module 152 is configured to receive system messages broadcast by network devices
- the executing module 151 is further configured to execute the validity management process of the first AI model when the region identifier in the system message is different from the model valid region identifier of the first AI model.
- the effective area is at least one of a tracking area, a radio access network RAN area, and a self-defined area, and the effective area includes at least one cell.
- the execution module 151 is further configured to execute the validity management of the first AI model when the validity period of the first timer of the first AI model expires In the process, the first timer is used to ensure the validity of the first AI model.
- the executing module 151 is further configured to start the first timer when the terminal receives the configuration information of the first timer; or, start the first timer when the terminal When using the first AI model, start the first timer.
- the first timer is based on at least one of the effective duration of the first AI model, service type, operating trajectory/area of the terminal, network load/energy consumption definite.
- the execution module 151 is further configured to execute the validity management process of the first AI model when the channel quality of the terminal is higher than a first threshold; or, In the case that the channel quality of the terminal is lower than the second threshold, the validity management process of the first AI model is executed.
- the execution module 151 is further configured to execute the validity management process of the first AI model when the available computing power of the terminal is higher than a third threshold; or , when the available computing power of the terminal is lower than a fourth threshold, execute the validity management process of the first AI model.
- the execution module 151 is further configured to execute the validity management process of the first AI model when the storage capacity of the terminal is higher than a fifth threshold; or, In a case where the storage capability of the terminal is lower than the sixth threshold, the validity management process of the first AI model is executed.
- the execution module 151 is further configured to execute the validity management process of the first AI model when the classification of the wireless environment changes; or, in the When the scene characteristics of the wireless environment change, execute the effectiveness management process of the first AI model; or, when the channel environment index characteristics of the wireless environment change, execute the first AI model The validity management process; or, when the time domain characteristic information of the wireless environment changes, execute the validity management process of the first AI model; or, in the frequency domain characteristic information of the wireless environment of the terminal In case of a change, execute the validity management process of the first AI model; or, in the case of a change in the spatial feature information of the wireless environment, execute the validity management process of the first AI model.
- the execution module 151 is further configured to autonomously execute the validity management process of the first AI model when the model invalidation condition is satisfied.
- the execution module 151 is further configured to autonomously update the first AI model when the model failure condition is met; or, when the model failure condition is met , autonomously switch the first AI model to a second AI model; or, autonomously stop using the first AI model when the failure condition of the model is met.
- the execution module 151 is further configured to autonomously update the first AI model according to the model input information during use of the first AI model when the model failure condition is met AI model.
- the terminal stores at least two sets of candidate AI models; the execution module 151 is further configured to select the at least two sets of candidate AI
- the second AI model is determined in the model; the first AI model is automatically switched to the second AI model.
- the at least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the reporting module 153 is configured to report request information to the network device when the model failure condition is satisfied, and the request information is used to request the network device to execute the first A validity management process of an AI model; the execution module 151 is further configured to execute the validity management process of the first AI model according to the instruction information provided by the network device.
- the execution module 151 is further configured to update the first AI model according to the indication information provided by the network device; or, according to the indicating information, switching the first AI model to a second AI model; or, according to the indicating information provided by the network device, stopping using the first AI model.
- the indication information includes configuration information of the first AI model and/or auxiliary information for model updating; the execution module 151 is further configured to The configuration information of the first AI model and/or the auxiliary information used for model updating are used to update the first AI model.
- the configuration information of the first AI model includes structure information and/or parameter information of the first AI model.
- the indication information includes the identification of the second AI model; the terminal stores at least two sets of candidate AI models; the execution module 151 is further configured to Identifying the second AI model, determining the second AI model from the at least two sets of candidate AI models; switching the first AI model to the second AI model.
- the at least two sets of candidate AI models are related to at least one of computing power, service, and channel quality of the terminal.
- the request information includes an identification of the proposed AI model.
- the request information is transmitted through at least one of a radio resource control RRC message, a media access layer control unit MAC CE, uplink control information UCI, and a random access process.
- the indication information is transmitted by at least one of RRC message, MAC CE, downlink control information DCI, and system message broadcast.
- the way of updating the first AI model includes adjusting the number of layers of the first AI model, updating the network parameters and/or network topology of the first AI model, cutting or compressing At least one of the first AI models.
- the reporting module 153 is configured to report a completion message to the network device, where the completion message is used to indicate that the terminal completes the validity management process of the first AI model.
- the first AI model is used to implement mobility enhancement, CSI feedback, channel estimation, load balancing, energy saving of the terminal/network, beam management, trajectory prediction of the terminal, service At least one of prediction, positioning enhancement, and radio resource management.
- the first AI model runs, trains or infers in RRC, Service Data Adaptation Protocol SDAP, Packet Data Convergence Protocol PDCP, Radio Link Control RLC, Media Access Control MAC, Physical Any layer in the layer PHY.
- RRC Service Data Adaptation Protocol SDAP
- Packet Data Convergence Protocol PDCP Packet Data Convergence Protocol
- Radio Link Control RLC Radio Link Control RLC
- Media Access Control MAC Physical Any layer in the layer PHY.
- the function optimized by the first AI model is a function corresponding to any layer of RRC, SDAP, PDCP, RLC, MAC, and PHY.
- FIG. 16 shows a schematic structural diagram of a terminal 1600 provided by an embodiment of the present application.
- the terminal 1600 may include: a processor 1601 , a transceiver 1602 and a memory 1603 .
- the processor 1601 includes one or more processing cores, and the processor 1601 executes various functional applications and information processing by running software programs and modules.
- the transceiver 1602 may include a receiver and a transmitter.
- the receiver and the transmitter may be implemented as the same wireless communication component, and the wireless communication component may include a wireless communication chip and a radio frequency antenna.
- the memory 1603 may be connected to the processor 1601 and the transceiver 1602 .
- the memory 1603 may be used to store a computer program executed by the processor, and the processor 1601 is used to execute the computer program, so as to implement various steps performed by the terminal in the wireless communication system in the above method embodiments.
- volatile or non-volatile storage device includes but not limited to: magnetic disk or optical disk, electrically erasable and programmable Read Only Memory, Erasable Programmable Read Only Memory, Static Anytime Access Memory, Read Only Memory, Magnetic Memory, Flash Memory, Programmable Read Only Memory.
- the process performed by the transceiver 1602 and the processor 1601 in the terminal 1600 can refer to the above-mentioned methods shown in FIG. 2, FIG. 3, FIG. 5, FIG. 8, FIG. 9, FIG. 11 and FIG.
- SMF Service Management Function, business management function
- FIG. 17 shows a schematic structural diagram of a network device 1700 provided by an embodiment of the present application.
- the network device 1700 may include: a processor 1701 , a transceiver 1702 and a memory 1703 .
- the processor 1701 includes one or more processing cores, and the processor 1701 executes various functional applications and information processing by running software programs and modules.
- Transceiver 1702 may include a receiver and a transmitter.
- the transceiver 1702 may include a wired communication component, and the wired communication component may include a wired communication chip and a wired interface (such as an optical fiber interface).
- the transceiver 1702 may also include a wireless communication component, and the wireless communication component may include a wireless communication chip and a radio frequency antenna.
- the memory 1703 may be connected to the processor 1701 and the transceiver 1702 .
- the memory 1703 may be used to store a computer program executed by the processor, and the processor 1701 is used to execute the computer program, so as to implement various steps performed by the network device in the wireless communication system in the foregoing method embodiments.
- the memory 1703 can be implemented by any type of volatile or non-volatile storage device or their combination.
- the volatile or non-volatile storage device includes but not limited to: magnetic disk or optical disk, electrically erasable and programmable Read Only Memory, Erasable Programmable Read Only Memory, Static Anytime Access Memory, Read Only Memory, Magnetic Memory, Flash Memory, Programmable Read Only Memory.
- the transceiver 1702 is configured to receive a second service access request sent by a non-relay terminal in a relay side link scenario; the second service access request is used to request Accessing the non-relay terminal to the first broadcast/multicast service.
- the process performed by the transceiver 1702 and the processor 1701 in the above-mentioned network device 1700 can refer to the above-mentioned methods shown in FIG. 2, FIG. 3, FIG. 5, FIG. 8, FIG. 9, FIG. 11 and FIG. The various steps performed by the UPF unit in .
- the process performed by the transceiver 1702 and the processor 1701 in the above-mentioned network device 1700 can refer to the above-mentioned methods shown in FIG. 2, FIG. 3, FIG. 5, FIG. 8, FIG. 9, FIG. 11 and FIG. The various steps performed by the SMF unit in .
- the embodiment of the present application also provides a computer-readable storage medium, where a computer program is stored in the storage medium, and the computer program is loaded and executed by a processor to realize the above-mentioned FIG. 2 , FIG. 3 , FIG. 5 , and FIG. 8 .
- each step is performed by a terminal or a network device.
- the present application also provides a computer program product including computer instructions stored in a computer-readable storage medium.
- the processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the above-mentioned Fig. 2, Fig. 3, Fig. 5, Fig. 8, Fig. 9, Fig. 11 and Fig. 13
- each step is performed by a terminal or a network device.
- the present application also provides a chip, which is used to run in a computer device, so that the computer device executes the above-mentioned steps shown in Fig. 2, Fig. 3, Fig. 5, Fig. 8, Fig. 9, Fig. 11 and Fig. 13.
- each step is performed by a terminal or a network device.
- the present application also provides a computer program, which is executed by a processor of a computer device, so as to implement the methods shown in the above-mentioned Fig. 2 , Fig. 3 , Fig. 5 , Fig. 8 , Fig. 9 , Fig. 11 and Fig. 13 , various steps performed by the terminal or the network device.
- the functions described in the embodiments of the present application may be implemented by hardware, software, firmware or any combination thereof.
- the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.
- Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. Storage media may be any available media that can be accessed by a general purpose or special purpose computer.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
Description
Claims (62)
- 一种无线通信中保障人工智能AI模型有效性方法,其特征在于,所述方法由终端执行,所述方法包括:在满足模型失效条件的情况下,执行第一AI模型的有效性管理流程。
- 根据权利要求1所述的方法,其特征在于,所述在满足模型失效条件的情况下,执行第一AI模型的有效性管理流程,包括:在所述终端离开所述第一AI模型的有效区域的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述第一AI模型的有效期到期的情况下,执行所述第一AI模型的有效性管理流程;或者,基于所述第一AI模型的周期性更新,执行所述第一AI模型的有效性管理流程;或者,在所述终端的信道质量满足第一阈值条件的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述终端的可用算力满足第二阈值条件的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述终端的存储能力满足第三阈值条件的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述终端的业务类型发生变化或服务质量QoS发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述第一AI模型的精确度不符合模型精度要求的情况下,执行所述第一AI模型的有效性管理流程;或者,在无线环境发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在接收到网络设备指示的模型切换指示消息的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求2所述的方法,其特征在于,所述在所述终端离开所述第一AI模型的有效区域的情况下,执行所述第一AI模型的有效性管理流程,包括:接收网络设备广播的系统消息;在所述系统消息中的区域标识与所述第一AI模型的模型有效区域标识不同的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求2所述的方法,其特征在于,所述有效区域是跟踪区、无线接入网RAN区域、自定义区域中的至少一种,所述有效区域包括至少一个小区。
- 根据权利要求2所述的方法,其特征在于,所述在所述第一AI模型的有效期到期的情况下,执行所述第一AI模型的有效性管理流程,包括:在所述第一AI模型的第一定时器超时的情况下,执行所述第一AI模型的有效性管理流程,所述第一定时器用于保障所述第一AI模型的有效性。
- 根据权利要求5所述的方法,其特征在于,所述方法还包括:在所述终端收所述第一定时器的配置信息时,启动所述第一定时器;或,在所述终端开始使用所述第一AI模型时,启动所述第一定时器;或,在所述终端收到所述第一AI模型时,启动所述第一定时器。
- 根据权利要求6所述的方法,其特征在于,所述第一定时器是基于所述第一AI模型的有效时长、业务类型、所述终端的运行轨迹/区域、网络负载/能耗中的至少一种确定的。
- 根据权利要求2所述的方法,其特征在于,所述在所述终端的信道质量满足第一阈值条件的情况下,执行第一AI模型的有效性管理流程,包括:在所述终端的信道质量高于第一阈值的情况下,执行所述第一AI模型的有效性管理流程;或,在所述终端的信道质量低于第二阈值的情况下,执行所述第一AI模型的有效 性管理流程。
- 根据权利要求2所述的方法,其特征在于,所述在所述终端可用算力满足阈值条件的情况下,执行第一AI模型的有效性管理流程,包括:在所述终端的可用算力高于第三阈值的情况下,执行所述第一AI模型的有效性管理流程;或,在所述终端的可用算力低于第四阈值的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求2所述的方法,其特征在于,所述在所述终端的存储能力满足第三阈值条件的情况下,执行所述第一AI模型的有效性管理流程,包括:在所述终端的存储能力高于第五阈值的情况下,执行所述第一AI模型的有效性管理流程;或,在所述终端的存储能力低于第六阈值的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求2所述的方法,其特征在于,所述在无线环境发生变化的情况下,执行所述第一AI模型的有效性管理流程,包括:在所述终端的无线环境的分类发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的场景特征发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的信道环境指标特征发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的时域特征信息发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的频域特征信息发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的空间特征信息发生变化的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求1至11任一项所述的方法,其特征在于,所述在满足模型失效条件的情况下,执行第一AI模型的有效性管理流程,包括:在满足所述模型失效条件的情况下,自主执行所述第一AI模型的有效性管理流程。
- 根据权利要求12所述的方法,其特征在于,所述在满足模型失效条件的情况下,自主执行第一AI模型的有效性管理流程,包括:在满足所述模型失效条件的情况下,自主更新所述第一AI模型;或,在满足所述模型失效条件的情况下,自主将所述第一AI模型切换为第二AI模型;或,在满足所述模型失效条件的情况下,自主停止使用所述第一AI模型。
- 根据权利要求13所述的方法,其特征在于,所述在满足所述模型失效条件的情况下,自主更新所述第一AI模型,包括:在满足所述模型失效条件的情况下,根据所述第一AI模型使用期间的模型输入信息,自主更新所述第一AI模型。
- 根据权利要求13所述的方法,其特征在于,所述终端存储有至少两套候选AI模型;所述在满足所述模型失效条件的情况下,自主将所述第一AI模型切换为第二AI模型,包括:在满足所述模型失效条件的情况下,从所述至少两套候选AI模型中确定所述第二AI模型;自主将所述第一AI模型切换为所述第二AI模型。
- 根据权利要求15所述的方法,其特征在于,所述至少两套候选AI模型与所述终端的算力、业务、信道质量中的至少一种相关。
- 根据权利要求1至11任一项所述的方法,其特征在于,所述在满足模型失效条件的情况下,执行第一AI模型的有效性管理流程,包括:在满足所述模型失效条件的情况下,向网络设备上报请求信息,所述请求信息用于向所述网络设备请求执行所述第一AI模型的有效性管理流程;根据所述网络设备提供的指示信息执行所述第一AI模型的有效性管理流程。
- 根据权利要求17所述的方法,其特征在于,所述根据所述网络设备提供的指示信息,执行第一AI模型的有效性管理流程,包括:根据所述网络设备提供的所述指示信息,更新所述第一AI模型;或,根据所述网络设备提供的所述指示信息,将所述第一AI模型切换为第二AI模型;或,根据所述网络设备提供的所述指示信息,停止使用所述第一AI模型。
- 根据权利要求18所述的方法,其特征在于,所述指示信息包括所述第一AI模型的配置信息和/或用于模型更新的辅助信息;所述根据所述网络设备提供的所述指示信息更新所述第一AI模型,包括:根据所述网络设备提供的所述第一AI模型的配置信息和/或用于模型更新的辅助信息,对所述第一AI模型进行更新。
- 根据权利要求19所述的方法,其特征在于,所述第一AI模型的配置信息包括所述第一AI模型的结构信息和/或参数信息。
- 根据权利要求18所述的方法,其特征在于,所述指示信息包括第二AI模型的标识;所述终端存储有至少两套候选AI模型;所述根据所述网络设备提供的所述指示信息将所述第一AI模型切换为第二AI模型,包括:根据所述网络设备提供的所述第二AI模型的标识,从所述至少两套候选AI模型中确定所述第二AI模型;将所述第一AI模型切换为所述第二AI模型。
- 根据权利要求21所述的方法,其特征在于,所述至少两套候选AI模型与所述终端的算力、业务、信道质量中的至少一种相关。
- 根据权利要求21所述的方法,其特征在于,所述请求信息包括建议的AI模型的标识。
- 根据权利要求17所述的方法,其特征在于,所述请求信息通过无线资源控制RRC消息、媒体接入层控制单元MAC CE、上行控制信息UCI、随机接入过程中的至少一种方式进行传输。
- 根据权利要求17所述的方法,其特征在于,所述指示信息通过RRC消息、MAC CE、下行控制信息DCI、系统消息广播中的至少一种方式进行传输。
- 根据权利要求13或18所述的方法,其特征在于,更新所述第一AI模型的方式包括调整所述第一AI模型的层数、更新所述第一AI模型的网络参数和/或网络拓扑、裁剪或压缩所述第一AI模型中的至少一种。
- 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:向网络设备上报完成消息,所述完成消息用于表示所述终端完成所述第一AI模型的有效性管理流程。
- 根据权利要求1至11任一项所述的方法,其特征在于,所述第一AI模型用于实现移动性增强、CSI反馈、信道估计、负载均衡、所述终端/网络节能、波束管理、所述终端的轨迹预测、业务预测、定位增强、无线资源管理中的至少一种。
- 根据权利要求1至11任一项所述的方法,其特征在于,所述第一AI模型运行、训练或推理在RRC、服务数据适配协议SDAP、分组数据汇聚协议PDCP、无线链路控制RLC、媒体接入控制MAC、物理层PHY中的任意一层。
- 根据权利要求1至11任一项所述的方法,其特征在于,所述第一AI模型优化的功能是RRC、SDAP、PDCP、RLC、MAC、PHY中的任意一层对应的功能。
- 一种无线通信中保障人工智能AI模型有效性装置,其特征在于,所述装置包括:执行模块,用于在满足模型失效条件的情况下,执行第一AI模型的有效性管理流程。
- 根据权利要求31所述的装置,其特征在于,所述执行模块,还用于在所述终端离开所述第一AI模型的有效区域的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述第一AI模型的有效期到期的情况下,执行所述第一AI模型的有效性管理流程;或者,基于所述第一AI模型的周期性更新,执行所述第一AI模型的有效性管理流程;或者,在所述终端的信道质量满足第一阈值条件的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述终端的可用算力满足第二阈值条件的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述终端的存储能力满足第三阈值条件的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述终端的业务类型发生变化或服务质量QoS发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述第一AI模型的精确度不符合模型精度要求的情况下,执行所述第一AI模型的有效性管理流程;或者,在无线环境发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在接收到网络设备指示的模型切换指示消息的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求32所述的装置,其特征在于,接收模块,用于接收网络设备广播的系统消息;所述执行模块,还用于在所述系统消息中的区域标识与所述第一AI模型的模型有效区域标识不同的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求32所述的装置,其特征在于,所述有效区域是跟踪区、无线接入网RAN区域、自定义区域中的至少一种,所述有效区域包括至少一个小区。
- 根据权利要求32所述的装置,其特征在于,所述执行模块,还用于在所述第一AI模型的第一定时器的有效期到期的情况下,执行所述第一AI模型的有效性管理流程,所述第一定时器用于保障所述第一AI模型的有效性。
- 根据权利要求35所述的装置,其特征在于,所述执行模块,还用于在所述终端收所述第一定时器的配置信息时,启动所述第一定时器;或,在所述终端开始使用所述第一AI模型时,启动所述第一定时器。
- 根据权利要求36所述的装置,其特征在于,所述第一定时器是基于所述第一AI模型的有效时长、业务类型、所述终端的运行轨迹/区域、网络负载/能耗中的至少一种确定的。
- 根据权利要求32所述的装置,其特征在于,所述执行模块,还用于在所述终端的信道质量高于第一阈值的情况下,执行所述第一AI模型的有效性管理流程;或,在所述终端的信道质量低于第二阈值的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求32所述的装置,其特征在于,所述执行模块,还用于在所述终端的可用算力高于第三阈值的情况下,执行所述第一AI模型的有效性管理流程;或,在所述终端的可用算力低于第四阈值的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求32所述的装置,其特征在于,所述执行模块,还用于在所述终端的存储能力高于第五阈值的情况下,执行所述第一AI模型的有效性管理流程;或,在所述终端的存储能力低于第六阈值的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求32所述的装置,其特征在于,所述执行模块,还用于在所述无线环境的分类发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的场景特征发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的信道环境指标特征发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的时域特征信 息发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的频域特征信息发生变化的情况下,执行所述第一AI模型的有效性管理流程;或者,在所述无线环境的空间特征信息发生变化的情况下,执行所述第一AI模型的有效性管理流程。
- 根据权利要求31至41任一项所述的装置,其特征在于,所述执行模块,还用于在满足所述模型失效条件的情况下,自主执行所述第一AI模型的有效性管理流程。
- 根据权利要求32所述的装置,其特征在于,所述执行模块,还用于在满足所述模型失效条件的情况下,自主更新所述第一AI模型;或,在满足所述模型失效条件的情况下,自主将所述第一AI模型切换为第二AI模型;或,在满足所述模型失效条件的情况下,自主停止使用所述第一AI模型。
- 根据权利要求43所述的装置,其特征在于,所述执行模块,还用于在满足所述模型失效条件的情况下,根据所述第一AI模型使用期间的模型输入信息,自主更新所述第一AI模型。
- 根据权利要求44所述的装置,其特征在于,所述终端存储有至少两套候选AI模型;所述执行模块,还用于在满足所述模型失效条件的情况下,从所述至少两套候选AI模型中确定所述第二AI模型;自主将所述第一AI模型切换为所述第二AI模型。
- 根据权利要求45所述的装置,其特征在于,所述至少两套候选AI模型与所述终端的算力、业务、信道质量中的至少一种相关。
- 根据权利要求31至41任一项所述的装置,其特征在于,上报模块,用于在满足所述模型失效条件的情况下,向网络设备上报请求信息,所述请求信息用于向所述网络设备请求执行所述第一AI模型的有效性管理流程;所述执行模块,还用于根据所述网络设备提供的指示信息执行所述第一AI模型的有效性管理流程。
- 根据权利要求47所述的装置,其特征在于,所述执行模块,还用于根据所述网络设备提供的所述指示信息,更新所述第一AI模型;或,根据所述网络设备提供的所述指示信息,将所述第一AI模型切换为第二AI模型;或,根据所述网络设备提供的所述指示信息,停止使用所述第一AI模型。
- 根据权利要求48所述的装置,其特征在于,所述指示信息包括所述第一AI模型的配置信息和/或用于模型更新的辅助信息;所述执行模块,还用于根据所述网络设备提供的所述第一AI模型的配置信息和/或用于模型更新的辅助信息,对所述第一AI模型进行更新。
- 根据权利要求49所述的装置,其特征在于,所述第一AI模型的配置信息包括所述第一AI模型的结构信息和/或参数信息。
- 根据权利要求48所述的装置,其特征在于,所述指示信息包括第二AI模型的标识;所述终端存储有至少两套候选AI模型;所述执行模块,还用于根据所述网络设备提供的所述第二AI模型的标识,从所述至少两套候选AI模型中确定所述第二AI模型;将所述第一AI模型切换为所述第二AI模型。
- 根据权利要求51所述的装置,其特征在于,所述至少两套候选AI模型与所述终端的算力、业务、信道质量中的至少一种相关。
- 根据权利要求51所述的装置,其特征在于,所述请求信息包括建议的AI模型的标识。
- 根据权利要求47所述的装置,其特征在于,所述请求信息通过无线资源控制RRC消息、媒体接入层控制单元MAC CE、上行控制信息UCI、随机接入过程中的至少一种方式进行传输。
- 根据权利要求47所述的装置,其特征在于,所述指示信息通过RRC消息、MAC CE、 下行控制信息DCI、系统消息广播中的至少一种方式进行传输。
- 根据权利要求43或48所述的装置,其特征在于,更新所述第一AI模型的方式包括调整所述第一AI模型的层数、更新所述第一AI模型的网络参数和/或网络拓扑、裁剪或压缩所述第一AI模型中的至少一种。
- 根据权利要求31至41任一项所述的装置,其特征在于,所述装置还包括:上报模块,用于向网络设备上报完成消息,所述完成消息用于表示所述终端完成所述第一AI模型的有效性管理流程。
- 根据权利要求31至41任一项所述的装置,其特征在于,所述第一AI模型用于实现移动性增强、CSI反馈、信道估计、负载均衡、所述终端/网络节能、波束管理、所述终端的轨迹预测、业务预测、定位增强、无线资源管理中的至少一种。
- 根据权利要求31至41任一项所述的装置,其特征在于,所述第一AI模型运行、训练或推理在RRC、服务数据适配协议SDAP、分组数据汇聚协议PDCP、无线链路控制RLC、媒体接入控制MAC、物理层PHY中的任意一层。
- 根据权利要求31至41任一项所述的装置,其特征在于,所述第一AI模型优化的功能是RRC、SDAP、PDCP、RLC、MAC、PHY中的任意一层对应的功能。
- 一种终端,其特征在于,所述终端包括:处理器;与所述处理器相连的收发器;用于存储处理器可执行指令的存储器;其中,所述处理器被配置为加载并执行所述可执行指令以实现如权利要求1至30任一所述的无线通信中保障AI模型有效性方法。
- 一种计算机可读存储介质,其特征在于,所述可读存储介质中存储有至少一条指令、至少一段程序、代码集或指令集,所述至少一条指令、所述至少一段程序、所述代码集或指令集由处理器加载并执行以实现如权利要求1至30任一所述的无线通信中保障AI模型有效性方法。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202180104774.7A CN118339567A (zh) | 2021-12-20 | 2021-12-20 | 无线通信中保障ai模型有效性方法、装置、终端及介质 |
PCT/CN2021/139547 WO2023115251A1 (zh) | 2021-12-20 | 2021-12-20 | 无线通信中保障ai模型有效性方法、装置、终端及介质 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2021/139547 WO2023115251A1 (zh) | 2021-12-20 | 2021-12-20 | 无线通信中保障ai模型有效性方法、装置、终端及介质 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023115251A1 true WO2023115251A1 (zh) | 2023-06-29 |
Family
ID=86900978
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2021/139547 WO2023115251A1 (zh) | 2021-12-20 | 2021-12-20 | 无线通信中保障ai模型有效性方法、装置、终端及介质 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN118339567A (zh) |
WO (1) | WO2023115251A1 (zh) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112561332A (zh) * | 2020-12-16 | 2021-03-26 | 北京百度网讯科技有限公司 | 模型管理方法、装置、电子设备、存储介质和程序产品 |
US20210303973A1 (en) * | 2020-03-26 | 2021-09-30 | 10353744 Canada Ltd. | Artificial intelligence-based personalized financial recommendation assistant system and method |
CN113537284A (zh) * | 2021-06-04 | 2021-10-22 | 中国人民解放军战略支援部队信息工程大学 | 基于拟态机制的深度学习实现方法及系统 |
-
2021
- 2021-12-20 WO PCT/CN2021/139547 patent/WO2023115251A1/zh active Application Filing
- 2021-12-20 CN CN202180104774.7A patent/CN118339567A/zh active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20210303973A1 (en) * | 2020-03-26 | 2021-09-30 | 10353744 Canada Ltd. | Artificial intelligence-based personalized financial recommendation assistant system and method |
CN112561332A (zh) * | 2020-12-16 | 2021-03-26 | 北京百度网讯科技有限公司 | 模型管理方法、装置、电子设备、存储介质和程序产品 |
CN113537284A (zh) * | 2021-06-04 | 2021-10-22 | 中国人民解放军战略支援部队信息工程大学 | 基于拟态机制的深度学习实现方法及系统 |
Also Published As
Publication number | Publication date |
---|---|
CN118339567A (zh) | 2024-07-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20230209390A1 (en) | Intelligent Radio Access Network | |
JP6019233B2 (ja) | 端末アクセス方法、システム及び端末 | |
US8379597B2 (en) | Method for preventing ping-pong handover effect in mobile WiMAX networks | |
CN113411850B (zh) | 切换方法和设备 | |
US9215643B2 (en) | Route selecting device and mobile radio communication system | |
US11838833B2 (en) | Grouping method, apparatus, and system | |
US10341891B2 (en) | User equipment adaptation of reporting triggers based on active set size | |
CN110971349B (zh) | 一种重复传输方法、终端和网络侧设备 | |
JP2018524892A (ja) | ワイヤレスネットワークからユーザ装置へのデータのダウンロードを制御するための方法、装置、コンピュータ読み取り可能な媒体及びコンピュータプログラム製品 | |
JP2022502929A (ja) | データ伝送制御方法、ネットワーク機器および記憶媒体 | |
US20230189057A1 (en) | Service traffic steering method and apparatus | |
TW202044901A (zh) | 直通鏈路傳輸方法和終端 | |
CN110463288B (zh) | 直连通信的发送功率控制方法、装置、设备及存储介质 | |
CN111263430B (zh) | Srs功率控制方法和设备 | |
CN114982268A (zh) | 一种无线参数调整方法和装置 | |
JP2024515402A (ja) | ポジショニング設定方法及び電子装置 | |
WO2022000379A1 (zh) | 设备切换方法、装置、设备及可读存储介质 | |
WO2023115251A1 (zh) | 无线通信中保障ai模型有效性方法、装置、终端及介质 | |
WO2023039905A1 (zh) | Ai数据的传输方法、装置、设备及存储介质 | |
US12009980B2 (en) | Method and device for configuring terminal policy, terminal, base station and storage medium | |
CN115567989A (zh) | 小区指示、小区切换方法、装置、服务节点、终端及介质 | |
CN108476545A (zh) | 无线资源控制rrc状态的控制方法、装置及基站 | |
CN114765754A (zh) | 无线服务开放方法及第一执行主体 | |
WO2023207026A1 (zh) | 信息指示和处理方法以及装置 | |
WO2023184419A1 (zh) | 功率控制方法、装置、设备及存储介质 |
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: 21968408 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 202180104774.7 Country of ref document: CN |
|
REG | Reference to national code |
Ref country code: BR Ref legal event code: B01A Ref document number: 112024012359 Country of ref document: BR |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2021968408 Country of ref document: EP Effective date: 20240722 |
|
ENP | Entry into the national phase |
Ref document number: 112024012359 Country of ref document: BR Kind code of ref document: A2 Effective date: 20240618 |