Model monitoring method and apparatus, device, system, medium, and computer program product

Abstract

The present application relates to the technical field of communications, and discloses a model monitoring method and apparatus, a device, a system, a medium, and a computer program product, capable of monitoring model performance and reducing air interface overhead. The model monitoring method in embodiments of the present application comprises: when determining that the performance state of a first model is a first state, a terminal sends first information to a network side device, the first information being used for the network side device to determine the performance state of the first model, and the first model being used for CSI feedback.

Description

Model monitoring methods, devices, equipment, systems, media, and computer program products Technical Field

[0001] This application belongs to the field of communication technology, specifically relating to a model monitoring method, device, equipment, system, medium, and computer program product. Background Technology

[0002] In wireless communication systems, channel state information (CSI) feedback can be based on artificial intelligence (AI) / machine learning (ML) models to save signaling overhead for CSI feedback.

[0003] Typically, after the terminal encodes the CSI using an AI model, it sends the CSI feedback to the network-side device. Correspondingly, the network-side device can decode the received CSI feedback using another AI model to obtain the complete CSI.

[0004] However, if the performance of the AI / ML model deteriorates, it may lead to a decrease in the accuracy of CSI feedback. Therefore, how to monitor the performance of the AI / ML model is an urgent problem to be solved. Summary of the Invention

[0005] This application provides a model monitoring method, apparatus, device, system, medium, and computer program product that can monitor model performance and reduce air interface overhead.

[0006] In a first aspect, a model monitoring method is provided, which may include: when a terminal determines that the performance state of a first model is a first state, sending first information to a network-side device, wherein the first information is used by the network-side device to determine the performance state of the first model, and the first model is used for feedback of channel state information (CSI).

[0007] In some possible implementations, the first state mentioned above includes: an abnormal performance state.

[0008] In some possible implementations, the aforementioned first information is sent by the terminal after determining that the performance state of the first model is a first state and receiving the first indication information, which is used to instruct the reporting of information used to determine the performance state of the first model.

[0009] In some possible implementations, the first model mentioned above includes a second model and a third model, where the second model is used to generate CSI feedback and the third model is used to recover CSI.

[0010] In some possible implementations, the above model monitoring method further includes: the terminal determining second information, which is used to indicate the performance status of the first model determined by the terminal.

[0011] In some possible implementations, the second information mentioned above includes any of the following:

[0012] The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI.

[0013] The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

[0014] In some possible implementations, the aforementioned second information is obtained through monitoring by a monitoring model;

[0015] The input information of the above monitoring model is any one of the following: the output information of the second model, or the CSI feedback in the output information of the second model;

[0016] The output information of the above monitoring model is any one of the following: the CSI recovered from the CSI feedback in the output information of the second model, the first performance index.

[0017] In some possible implementations, the aforementioned second information is obtained through monitoring by a monitoring model; the first performance metric is used to indicate the similarity or difference between the output information of the monitoring model and the input information of the second model.

[0018] In some possible implementations, the first performance metric includes K performance metrics, where K is a positive integer;

[0019] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the aforementioned first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0020] If K is greater than 1, and the mean of the above K performance indicators satisfies a preset relationship with the first threshold, then the above first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0021] If K is greater than 1, and the number of performance indicators among the above K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0022] In some possible implementations, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the set of candidate parameters of the Radio Resource Control (RRC) signaling configuration.

[0023] In some possible implementations, the aforementioned first information includes at least the input information of the second model.

[0024] In some possible implementations, the input information of the second model mentioned above includes any of the following:

[0025] CSI measured at the current moment;

[0026] Predict the CSI at at least one moment;

[0027] The CSI measured at the current time and the CSI feature information output by the second model at the previous time.

[0028] In some possible implementations, the above-mentioned model monitoring method also includes: the terminal sending second information to the network-side device.

[0029] In some possible implementations, the second information is sent by the terminal when the performance state indicated by the second information is the first state.

[0030] In some possible implementations, the second information includes a first performance metric, which is reported either in a non-quantized manner or in a quantized manner using Q bits, where Q is a positive integer.

[0031] In some possible implementations, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0032] In some possible implementations, the candidate resource for sending the second information is the first resource, and the candidate resource for sending the first information is the second resource;

[0033] The first or second resource is any of the following: predefined by the protocol, configured by the CSI reporting configuration, configured by the RRC signaling, or pre-configured.

[0034] In some possible implementations, the first or second resource is pre-configured when the AI- or ML-based CSI feedback function is activated on the terminal.

[0035] Among some possible implementations, the method of reporting the configuration of the first or second resource via CSI includes any of the following:

[0036] A CSI reporting configuration includes a first resource and a second resource.

[0037] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources.

[0038] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

[0039] Among some possible implementations, the above model monitoring method also includes:

[0040] The terminal receives a second indication information sent by the network-side device. The second indication information is used to indicate the monitoring model used by the updated terminal.

[0041] Based on the second indication information, the terminal updates the monitoring model used by the terminal.

[0042] Among some possible implementations, the above model monitoring method also includes:

[0043] The terminal receives a third indication message sent by the network-side device. The third indication message is used to indicate any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback.

[0044] The terminal executes the operation corresponding to the third instruction information.

[0045] Among some possible implementations, the above model monitoring method also includes:

[0046] The terminal sends its capability information to the network-side device, the capability information indicating at least one of the following:

[0047] Does the terminal possess the primary capability, which is the ability to monitor models or deploy monitoring models?

[0048] Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner?

[0049] In the model monitoring method provided in this application embodiment, when the terminal determines that the performance state of the first model is a first state, it sends first information to the network-side device. The first information is used by the network-side device to determine the performance state of the first model, and the first model is used for feedback of Channel State Information (CSI). In this scheme, since the terminal can send the first information to the network-side device when it determines that the performance state of the first model is a first state, instructing the network-side device to re-determine the performance state of the first model, it can achieve monitoring of model performance by both the terminal and the network side, improving the accuracy of model performance monitoring. Furthermore, since whether the first information is sent is determined by the performance state of the first model determined by the terminal, and not every time model monitoring is performed, air interface overhead can be saved.

[0050] Secondly, a model monitoring method is provided, which may include: a network-side device receiving second information sent by a terminal, the second information being used to indicate the performance status of a first model determined by the terminal, the first model being used for CSI feedback; when the performance status indicated by the second information is an abnormal performance status, the network-side device receiving the first information sent by the terminal, the first information being used by the network-side device to determine the performance status of the first model.

[0051] In some possible implementations, the above model monitoring method may also include:

[0052] When the performance status indicated by the second information is the first status, the network-side device sends a first indication information to the terminal. The first indication information is used to indicate the information reported to determine the performance status of the first model.

[0053] In some possible implementations, the first model mentioned above includes a second model and a third model, where the second model is used to generate CSI feedback and the third model is used to recover CSI.

[0054] In some possible implementations, the second information mentioned above includes any of the following:

[0055] The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI.

[0056] The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

[0057] In some possible implementations, the first performance metric mentioned above includes K performance metrics, where K is a positive integer;

[0058] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0059] If K is greater than 1, and the mean of the above K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0060] If K is greater than 1, and the number of performance indicators among the above K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0061] In some possible implementations, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0062] In some possible implementations, the aforementioned first information includes at least the input information of the second model.

[0063] In some possible implementations, the input information of the second model mentioned above includes any of the following:

[0064] CSI measured at the current moment;

[0065] Predict the CSI at at least one moment;

[0066] The CSI measured at the current time and the CSI feature information output by the second model at the previous time.

[0067] In some possible implementations, the second information mentioned above includes a first performance metric, which is reported either in a non-quantized manner or in a quantized manner using Q bits, where Q is a positive integer.

[0068] In some possible implementations, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0069] Among some possible implementations, the above model monitoring method also includes:

[0070] If the number of times the performance state of the first model determined by the network-side device differs from the performance state of the first model reaches a preset number, the network-side device sends a second indication message, which is used to indicate the monitoring model to be updated.

[0071] Among some possible implementations, the above model monitoring method also includes:

[0072] When the network-side device determines that the performance state of the first model is a first state, the network-side device performs a first operation, which includes at least one of the following:

[0073] Send a third instruction message, which is used to instruct any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback;

[0074] Update the model used to recover CSI;

[0075] Switch the model used to recover CSI;

[0076] Discontinue the use of the model used to recover CSI.

[0077] In some possible implementations, the candidate transmission resource for the second information configured by the network-side device is the first resource, and the candidate transmission resource for the first information configured by the network-side device is the second resource.

[0078] The first or second resource can be configured through any of the following: CSI reporting configuration, RRC signaling configuration, or pre-configuration.

[0079] In some possible implementations, the first or second resource is pre-configured by the network-side device when the CSI feedback function based on artificial intelligence (AI) or machine learning (ML) is activated.

[0080] Among some possible implementations, the method of reporting the configuration of the first or second resource via CSI includes any of the following:

[0081] A CSI reporting configuration includes a first resource and a second resource.

[0082] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources.

[0083] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

[0084] Among some possible implementations, the above model monitoring method also includes:

[0085] The network-side device receives transmitted capability information, which is used to indicate at least one of the following:

[0086] Does the terminal possess the primary capability, which is the ability to monitor models or deploy monitoring models?

[0087] Does it have a second capability, which is the ability to support reporting information in a configured codebook quantization manner?

[0088] In the model monitoring method provided in this application embodiment, the network-side device can receive second information sent by the terminal. The second information indicates the performance status of a first model determined by the terminal, and the first model is used for CSI feedback. When the performance status indicated by the second information is a first state, the network-side device receives the first information sent by the terminal. The first information is used by the network-side device to determine the performance status of the first model. In this solution, since the network-side device can receive the first information sent by the terminal after receiving the second information, and when the second information indicates that the performance status of the first model determined by the terminal is a first state, to re-determine the performance status of the first model, it can achieve monitoring of model performance by both the terminal side and the network side. Furthermore, since whether the first information is received is determined by the performance status of the first model determined by the terminal, and not by listening to and receiving the first information whenever model monitoring is performed, air interface overhead can be saved.

[0089] Thirdly, a model monitoring device is provided, which may include a transmission module.

[0090] The aforementioned sending module is used to send first information to the network-side device when the terminal determines that the performance state of the first model is in the first state. The first information is used by the network-side device to determine the performance state of the first model, and the first model is used for feedback of channel state information (CSI).

[0091] In some possible implementations, the first state mentioned above includes: an abnormal performance state.

[0092] In some possible implementations, the aforementioned first information is sent by the terminal after determining that the performance state of the first model is a first state and receiving the first indication information, which is used to instruct the reporting of information used to determine the performance state of the first model.

[0093] In some possible implementations, the first model mentioned above includes a second model and a third model, where the second model is used to generate CSI feedback and the third model is used to recover CSI.

[0094] In some possible implementations, the aforementioned model monitoring device further includes: a determination module;

[0095] The aforementioned determining module is used to determine the second information, which is used to indicate the performance status of the first model determined by the terminal.

[0096] In some possible implementations, the second information mentioned above includes any of the following:

[0097] The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI.

[0098] The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

[0099] In some possible implementations, the aforementioned second information is obtained through monitoring by a monitoring model;

[0100] The input information of the above monitoring model is any one of the following: the output information of the second model, or the CSI feedback in the output information of the second model;

[0101] The output information of the above monitoring model is any one of the following: the CSI recovered from the CSI feedback in the output information of the second model, the first performance index.

[0102] In some possible implementations, the aforementioned second information is obtained through monitoring by a monitoring model;

[0103] The aforementioned first performance metric is used to indicate the similarity or difference between the output information of the monitoring model and the input information of the second model.

[0104] In some possible implementations, the first performance metric mentioned above includes K performance metrics, where K is a positive integer;

[0105] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0106] If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0107] If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0108] In some possible implementations, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the Radio Resource Control (RRC) signaling configuration.

[0109] In some possible implementations, the aforementioned first information includes at least the input information of the second model.

[0110] In some possible implementations, the input information of the second model mentioned above includes any of the following:

[0111] CSI measured at the current moment;

[0112] Predict the CSI at at least one moment;

[0113] The CSI measured at the current time and the CSI feature information output by the second model at the previous time.

[0114] In some possible implementations, the aforementioned sending module is also used to send second information to the network-side device.

[0115] In some possible implementations, the aforementioned second information is sent by the terminal when the performance state indicated by the second information is the first state.

[0116] In some possible implementations, the second information mentioned above includes a first performance metric, which is reported either in a non-quantized manner or in a quantized manner using Q bits, where Q is a positive integer.

[0117] In some possible implementations, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0118] In some possible implementations, the candidate transmission resource for the second information is the first resource, and the candidate transmission resource for the first information is the second resource;

[0119] The first or second resource is any of the following: predefined by the protocol, configured by the CSI reporting configuration, configured by the RRC signaling, or pre-configured.

[0120] In some possible implementations, the first resource or the second resource mentioned above is pre-configured when the AI- or ML-based CSI feedback function is activated on the terminal.

[0121] Among some possible implementations, the method of reporting the configuration of the first or second resource via CSI includes any of the following:

[0122] A CSI reporting configuration includes a first resource and a second resource.

[0123] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources.

[0124] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

[0125] In some possible implementations, the above-mentioned model monitoring device further includes: a receiving module and a processing module;

[0126] The aforementioned receiving module is used to receive second indication information sent by the network-side device. The second indication information is used to indicate the monitoring model used by the updating terminal.

[0127] The aforementioned processing module is used to update the monitoring model used by the terminal based on the second indication information received by the receiving module.

[0128] In some possible implementations, the above-mentioned model monitoring device further includes: a receiving module and a processing module;

[0129] The aforementioned receiving module is used to receive third indication information sent by the network-side device. The third indication information is used to indicate any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback.

[0130] The aforementioned processing module is used to perform the operation corresponding to the third instruction information received by the receiving module.

[0131] In some possible implementations, the aforementioned sending module is further configured to send terminal capability information to the network-side device, the capability information indicating at least one of the following:

[0132] Does the terminal possess the primary capability, which is the ability to monitor models or deploy monitoring models?

[0133] Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner?

[0134] In the model monitoring device provided in this application embodiment, the device can send first information to the network-side device when the terminal determines that the performance state of the first model is a first state. The first information is used by the network-side device to determine the performance state of the first model, and the first model is used for feedback of Channel State Information (CSI). In this solution, since the model monitoring device can send first information to the network-side device when the terminal determines that the performance state of the first model is a first state, instructing the network-side device to re-determine the performance state of the first model, it can achieve monitoring of model performance by both the terminal and the network side. Furthermore, since whether the first information is sent is determined by the performance state of the first model determined by the terminal, and not by feeding back first information to the network-side device every time model monitoring is performed, air interface overhead can be saved.

[0135] Fourthly, a model monitoring device is provided, which may include: a receiving module;

[0136] The aforementioned receiving module is used to receive second information sent by the terminal. The second information is used to indicate the performance status of the first model determined by the terminal. The first model is used for CSI feedback.

[0137] The aforementioned receiving module is further configured to receive first information sent by the terminal when the performance state indicated by the second information is the first state, wherein the first information is used by the network-side device to determine the performance state of the first model.

[0138] In some possible implementations, the aforementioned model monitoring device further includes: a transmission module;

[0139] The aforementioned sending module is used to send first indication information to the terminal when the performance status indicated by the second information received by the aforementioned receiving module is a first status. The first indication information is used to indicate the information to be reported to determine the performance status of the first model.

[0140] In some possible implementations, the first model includes a second model and a third model, where the second model is used to generate CSI feedback and the third model is used to recover CSI.

[0141] In some possible implementations, the second information includes any of the following:

[0142] The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI.

[0143] The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

[0144] In some possible implementations, the first performance metric includes K performance metrics, where K is a positive integer;

[0145] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0146] If K is greater than 1, and the mean of the above K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0147] If K is greater than 1, and the number of performance indicators among the above K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0148] In some possible implementations, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0149] In some possible implementations, the first information includes at least the input information of the second model.

[0150] In some possible implementations, the input information for the second model includes any of the following:

[0151] CSI measured at the current moment;

[0152] Predict the CSI at at least one moment;

[0153] The CSI measured at the current time and the CSI feature information output by the second model at the previous time.

[0154] In some possible implementations, the second information includes a first performance metric, which is reported either in a non-quantized manner or in a quantized manner using Q bits, where Q is a positive integer.

[0155] In some possible implementations, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0156] In some possible implementations, the aforementioned model monitoring device further includes: a transmission module;

[0157] The aforementioned sending module is used to send a second indication message to the terminal when the number of times the performance status of the first model determined by the network-side device differs from the performance status of the first model determined by the terminal reaches a preset number. The second indication message is used to instruct the terminal to update the monitoring model used.

[0158] In some possible implementations, the aforementioned model monitoring device may also include a processing module.

[0159] The aforementioned processing module is used to perform a first operation when the network-side device determines that the performance state of the first model is a first state. The first operation includes at least one of the following:

[0160] Send a third instruction to the terminal, the third instruction being used to instruct any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback;

[0161] Update the model used to recover CSI;

[0162] Switch the model used to recover CSI;

[0163] Discontinue the use of the model used to recover CSI.

[0164] In some possible implementations, the candidate transmission resource for the second information configured by the network-side device is the first resource, and the candidate transmission resource for the first information configured by the network-side device is the second resource.

[0165] The first or second resource can be configured through any of the following: CSI reporting configuration, RRC signaling configuration, or pre-configuration.

[0166] In some possible implementations, the first or second resource is pre-configured by the network-side device when the terminal activates the CSI feedback function based on artificial intelligence (AI) or machine learning (ML).

[0167] Among some possible implementations, the method of reporting the configuration of the first or second resource via CSI includes any of the following:

[0168] A CSI reporting configuration includes a first resource and a second resource.

[0169] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources.

[0170] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

[0171] In some possible implementations, the receiving module described above is further configured to receive terminal capability information sent by the terminal, the capability information being used to indicate at least one of the following:

[0172] Does the terminal possess the primary capability, which is the ability to monitor models or deploy monitoring models?

[0173] Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner?

[0174] In the model monitoring device provided in this application embodiment, the model monitoring device includes a receiving module. This receiving module can be used to receive second information sent by a terminal, the second information indicating the performance status of a first model determined by the terminal, and the first model being used for CSI feedback. The receiving module is also used to receive first information sent by the terminal when the performance status indicated by the second information is a first status, the first information being used by the network-side device to determine the performance status of the first model. In this solution, since the model monitoring device can receive the first information sent by the terminal after receiving the second information, and when the second information indicates that the performance status of the first model determined by the terminal is a first status, to re-determine the performance status of the first model, it can realize the monitoring of model performance by both the terminal side and the network side. Furthermore, since whether the first information is received is determined by the performance status of the first model determined by the terminal, and not by listening to and receiving the first information whenever model monitoring is performed, air interface overhead can be saved.

[0175] Fifthly, a terminal is provided, the terminal including a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method as described in the first aspect or any implementation thereof.

[0176] In a sixth aspect, embodiments of this application provide a network-side device, which includes a processor and a memory. The memory stores a program or instructions that can run on the processor, and when the program or instructions are executed by the processor, they implement the steps of the method as described in the second aspect or any implementation thereof.

[0177] In a seventh aspect, embodiments of this application provide a communication system, which includes a model monitoring device as described in the third aspect or any implementation thereof, and a model monitoring device as described in the fourth aspect or any implementation thereof; or the communication system includes a terminal as described in the fifth aspect and a network-side device as described in the sixth aspect.

[0178] Eighthly, embodiments of this application provide a readable storage medium storing a program or instructions that, when executed by a processor, implement the model monitoring method as described in the first aspect or any implementation thereof, or implement the steps of the model monitoring method as described in the second aspect or any implementation thereof.

[0179] Ninthly, embodiments of this application provide a computer program product stored in a storage medium, which is executed by at least one processor to perform the model monitoring method as described in the first aspect or any implementation thereof, or to implement the steps of the model monitoring method as described in the second aspect or any implementation thereof. Attached Figure Description

[0180] Figure 1 is a block diagram of a wireless communication system provided in an embodiment of this application;

[0181] Figure 2 is a schematic diagram of the terminal's CSI reporting method;

[0182] Figure 3 is one of the schematic diagrams of the CSI feedback architecture applied to the model monitoring method provided in the embodiments of this application;

[0183] Figure 4 is a second schematic diagram of the CSI feedback architecture applied to the model monitoring method provided in the embodiments of this application;

[0184] Figure 5 is a schematic diagram of the CSI feedback architecture applied to the model monitoring method provided in the embodiments of this application (third of the three).

[0185] Figure 6 is a block diagram of the CSI monitoring system provided in an embodiment of this application;

[0186] Figure 7 is one of the flowcharts of the model monitoring method provided in the embodiments of this application;

[0187] Figure 8 is a second schematic flowchart of the model monitoring method provided in the embodiments of this application;

[0188] Figure 9 is one of the interactive schematic diagrams of the model monitoring method provided in the embodiments of this application;

[0189] Figure 10 is one of the interactive schematic diagrams of the model monitoring method provided in the embodiments of this application;

[0190] Figure 11 is a second interactive schematic diagram of the model monitoring method provided in the embodiments of this application;

[0191] Figure 12 is the third interactive schematic diagram of the model monitoring method provided in the embodiments of this application;

[0192] Figure 13 is one of the structural schematic diagrams of the model monitoring device provided in the embodiments of this application;

[0193] Figure 14 is a second schematic diagram of the structure of the model monitoring device provided in the embodiment of this application;

[0194] Figure 15 is a third structural schematic diagram of the model monitoring device provided in the embodiments of this application;

[0195] Figure 16 is a fourth structural schematic diagram of the model monitoring device provided in the embodiments of this application;

[0196] Figure 17 is a fifth schematic diagram of the structure of the model monitoring device provided in the embodiments of this application;

[0197] Figure 18 is a sixth schematic diagram of the model monitoring device provided in the embodiment of this application;

[0198] Figure 19 is a seventh structural schematic diagram of the model monitoring device provided in the embodiments of this application;

[0199] Figure 20 is a schematic diagram of the structure of the communication device provided in the embodiments of this application. Detailed Implementation

[0200] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0201] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, without limiting the number of objects; for example, the first object can be one or more. Furthermore, "or" in this application indicates at least one of the connected objects. For example, "A or B" covers three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B. The character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0202] The technical solutions of this application can be applied to various communication systems, including but not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, and can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), or other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used with the systems and radio technologies mentioned above, as well as with other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description; however, these technologies can also be applied to systems other than NR systems, such as 6th generation (6G) radio systems. th Generation 6G (6G) communication system, 7th generation communication system.

[0203] The wireless communication system used in this application embodiment is shown in Figure 1. This wireless communication system includes a terminal 11 (or terminal device) and a network-side device 12. The terminal 11 can be any terminal device, including but not limited to terminal devices connected to the network-side device 12 or other terminal devices via wired or wireless connections. Terminal devices can refer to access terminals, user equipment (UE), user units, user stations, mobile stations, mobile stations, remote stations, remote terminals, mobile devices, user terminals, terminals, wireless communication devices, user agents, or user devices. The access terminal can be a cellular phone, cordless phone, Session Initiation Protocol (SIP) phone, Wireless Local Loop (WLL) station, Personal Digital Assistant (PDA), handheld device with wireless communication capabilities, computing device, or other processing device connected to a wireless modem, in-vehicle device, wearable device, terminal device in a 5G network, or terminal device in a future-evolved network, etc. Network-side equipment 12 may include access network equipment or core network equipment. Access network equipment may also be referred to as Radio Access Network (RAN) equipment, Radio Access Network Function, or Radio Access Network Unit. Access network equipment may include base stations, Wireless Local Area Network (WLAN) access points (AS), or Wireless Fidelity (WiFi) nodes, etc. Base stations may be referred to as Node B (NB). As long as the same technical effect is achieved, the base station is not limited to specific technical terms. It should be noted that this application embodiment only uses a base station in an NR system as an example for description and does not limit the specific type of base station.

[0204] Core network equipment may include, but is not limited to, at least one of the following: core network node, core network function, Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (or L-NEF), and Binding Support. Core network functions (BSF), application functions (AF), etc. It should be noted that this application embodiment only uses core network equipment in the NR system as an example for description, and does not limit the specific type of core network equipment.

[0205] The following explains the nouns or terms used in the embodiments of this application.

[0206] 1. CSI feedback in NR systems

[0207] In current NR systems, for CSI feedback schemes, terminals typically employ codebook-based feature vector feedback to enable the base station to acquire downlink CSI. Specifically, the base station sends a downlink channel state information-reference signal (CSI-RS) to the terminal. The terminal uses the CSI-RS to estimate the downlink channel CSI and performs eigenvalue decomposition on the estimated CSI to obtain the feature vector corresponding to that downlink channel CSI. NR systems support various codebook types, including Type I, eType II, and Doppler eType II, to meet CSI feedback requirements with varying precision, overhead, and scenarios. The Doppler eType II codebook allows the terminal to extract time-domain related information of the channel at multiple measurement times, predict the channel state information at the next N4 times, and jointly compress and feed back the predicted CSI at the next N4 times to the base station.

[0208] 2. Terminal CSI reporting method

[0209] Figure 2 shows a schematic diagram of the CSI reporting method of the terminal. As shown in Figure 2, the terminal can report CSI in three ways: periodic CSI, quasi-persistent CSI, and non-periodic CSI. Among them, quasi-persistent CSI can also be called semi-persistent CSI.

[0210] Periodic CSIs can be transmitted on the Physical Uplink Control Channel (PUCCH). The CSI reporting configuration corresponding to the periodic CSI is configured by Radio Resource Control (RRC) signaling. After receiving the corresponding RRC signaling configuration, the terminal periodically reports the CSI.

[0211] Quasi-persistent CSI can be transmitted on the PUCCH or the Physical Uplink Shared Channel (PUSCH). The CSI or CSI feedback transmitted on the PUCCH has its corresponding CSI reporting configuration pre-configured by RRC signaling and activated or deactivated by Media Access Control (MAC) layer signaling. The CSI or CSI configuration transmitted on the PUSCH has its corresponding CSI reporting configuration dynamically indicated (e.g., activated or deactivated) by Downlink Control Information (DCI) signaling. After receiving activation or indication signaling from the network configuration, the terminal periodically transmits CSI on the PUCCH or PUSCH until it receives deactivation signaling and stops reporting.

[0212] The CSI reporting configuration for non-periodic CSIs is also pre-configured via RRC signaling, and some of these configurations can be activated via MAC layer signaling. The CSI trigger signaling in the DCI then indicates the CSI reporting configuration used for CSI reporting or feedback. After receiving the CSI trigger signaling, the terminal reports the corresponding CSI on the PUSCH scheduled by the base station in one go, according to the CSI reporting configuration indicated by the CSI trigger signaling.

[0213] 3. AI-based CSI feedback

[0214] Artificial intelligence (AI) technology relies on the development of different types of neural networks and deep learning algorithms, and has already achieved widespread application in various fields such as image, speech, and video processing. Typical neural network architectures include fully connected networks, convolutional neural networks (CNNs), recurrent neural networks (RNNs), and transformer structures with self-attention mechanisms. Different neural network architectures can accomplish different task objectives. Drawing on the rapid development of AI technology, the combination of artificial intelligence and wireless communication technology has also attracted widespread interest from academia and industry.

[0215] Within the Third Generation Partnership Project (3GPP), extensive research, evaluation, and standardization work has been carried out on AI-based CSI feedback, beam management, and positioning technologies. Given the significant success of AI technologies, especially deep learning, in computer vision and natural language processing, the communications field has begun to explore the use of deep learning to address technical challenges that traditional model-based monitoring methods struggle with. Deep learning's commonly used neural network architectures are non-linear and data-driven, enabling feature extraction from actual channel information, such as channel matrix information, and reconstructing the compressed channel matrix information from the terminal side at the base station as accurately as possible. This provides the possibility of reducing CSI feedback overhead for the UE while ensuring the accurate reconstruction of channel information.

[0216] Deep learning-based CSI feedback treats channel information as an image to be compressed, uses a deep learning autoencoder to compress the input channel information, and reconstructs the compressed channel image on the network side, which can preserve the actual channel information to a greater extent.

[0217] In 3GPP discussions, AI-based CSI feedback, as one of the main use cases for AI projects, underwent multiple rounds of discussions regarding simulation results and potential standard impact. Its main implementation framework is as follows:

[0218] The first type of CSI feedback architecture employs an AI-based CSI autoencoder method. The entire feedback system consists of an encoder and a decoder, deployed on the terminal side and the base station side, respectively. Specifically, as shown in Figure 3, the terminal obtains channel information, such as CSI, through channel estimation. This information serves as the CSI input to the encoder. The encoder's neural network compresses and encodes the channel information, such as the channel information matrix, and feeds the compressed bitstream back to the base station via the air interface feedback link. The base station uses the decoder to recover the channel information based on the feedback bitstream, obtaining the CSI output, such as the decoder outputting complete feedback channel information. The neural networks of the encoder and decoder shown in Figure 3 can be any of the following: Deep Neural Networks (DNNs) composed of multiple fully connected layers; CNNs composed of multiple convolutional layers; RNNs with structures such as Long Short-Term Memory (LSTM) or Gated Recurrent Units (GRUs); or neural networks with residual or self-attention mechanisms.

[0219] It is understandable that the CSI input of the encoder and the CSI output of the decoder can both be full-channel information or feature vectors obtained based on full-channel information. Therefore, current deep learning-based channel information feedback methods are mainly divided into full-channel information-based feedback methods and feature vector-based feedback methods. While channel information-based feedback methods can achieve full-channel information compression and feedback, they have high feedback bitstream overhead and are not currently supported in NR systems. Feature vector-based feedback methods, on the other hand, are the feedback architecture currently supported in NR systems. Furthermore, AI-based feature vector feedback methods can achieve higher CSI feedback accuracy with the same feedback bit overhead, or significantly reduce feedback overhead while achieving the same CSI feedback accuracy.

[0220] It should be noted that the "channel information" and "full channel information" in the first CSI feedback architecture can be: CSI measured or predicted by the terminal side.

[0221] To further enhance the performance of AI-based CSI feedback, the temporal correlation between CSI at different times can be utilized to improve the compressed feedback performance of CSI. Specifically, this can include the second and third CSI feedback architectures described below.

[0222] The second CSI feedback architecture is an AI-based spatial-frequency-temporal joint CSI compression feedback, as shown in Figure 4. The CSI input to the user-side encoder includes not only the current CSI measurement information but also historical CSI information from past moments (such as the previous moment). This historical information is represented by the feature output of the latent vector space of the past CSI output by the encoder and does not have explicit physical meaning. Similarly, the input to the network-side decoder includes not only the current CSI feedback information but also historical CSI information from past moments. This historical information is represented by the feature output of the latent space of past moments from the decoder and also does not have explicit physical meaning. Therefore, for the feedback information at time t on the air interface, it not only implicitly represents the features of the current CSI but also includes information features from historical CSI that help in the recovery of the current CSI, which can be used by the network side to better recover the current CSI.

[0223] The third CSI feedback architecture: AI-based joint CSI prediction and compression, as shown in Figure 5. The user-side CSI prediction model takes the CSI measurement information within the measurement window as input and outputs predicted CSI information for at least one time step. After preprocessing (e.g., if the predicted CSI information is full-channel information, Singular Value Decomposition (SVD) can be used to extract feature vectors from different layers for the CSI compression process), this information is used as input to the encoder. In this case, when the encoder input is CSI information from multiple time steps, the encoder output is joint feedback information, and the decoder can simultaneously recover CSI information from multiple prediction time steps. By extracting the correlation information of CSI from multiple time steps, the performance of CSI feedback can be further enhanced with minimal air interface feedback overhead. Meanwhile, the user-side CSI prediction and CSI coding compression process can adopt either the separate processing process shown in Figure 5, which is called Separated Prediction and Compression (SPC) in the 3GPP discussion; or the joint processing process, as shown in the dashed box, which is to complete Joint Prediction and Compression (JPC) through an AI model.

[0224] The model monitoring method provided in this application embodiment can monitor the performance status of models in the above three CSI feedback architectures, as well as models in future AI-based CSI feedback architectures, in order to ensure the link performance of the CSI feedback link.

[0225] The model monitoring architecture used in this embodiment is shown in Figure 6. This architecture may include a monitoring model 61, a second model 62, and a third model 63. During the model inference phase, the second model 62 is deployed on the terminal side, and the third model 63 is deployed on the network side. To implement model monitoring on the terminal side—that is, to monitor the performance status of the second model 62 and the third model 63 on the terminal side to monitor the link performance of the CSI feedback link—monitoring model 61 can be deployed on the terminal side. It can be understood that the aforementioned second model and third model can be collectively referred to as the first model, used for CSI feedback; the first model can also be called the CSI feedback model.

[0226] In Figure 6, the terminal can be terminal 11 in the communication system shown in Figure 1, and the network-side device in Figure 6 can be network-side device 12 in the communication system shown in Figure 1.

[0227] The monitoring process of the model monitoring method provided in this application embodiment will be described below with reference to the model monitoring architecture shown in Figure 6.

[0228] To facilitate the description of the model monitoring process, the CSI feedback process is briefly described here: The terminal can input the measured or predicted CSI into the second model 62 to obtain CSI feedback, and then feed back the CSI feedback to the network-side device through the CSI feedback link. After receiving the CSI feedback, the network-side device can recover the CSI feedback through the third model 63 to obtain the recovered first CSI.

[0229] During the model monitoring process, after receiving CSI feedback, the terminal can recover the CSI feedback through monitoring model 61 to obtain the recovered second CSI. Based on the recovered second CSI and the CSI measured or predicted by the terminal, the performance status of the second model 62 and the third model 63 is determined. If the performance status determined by the terminal is abnormal, the terminal can send the full CSI, i.e., the measured or predicted CSI, to the network-side device. This allows the network-side device to re-determine the performance status of the second model 62 and the third model 63 based on the full CSI and the first SCI recovered by the network-side device through the third model 63. If the performance status determined by the network-side device is normal, it indicates a false alarm by the terminal, meaning that the terminal's monitoring result has an error. Furthermore, if the number of false alarms by the terminal reaches a certain number, the network-side device can instruct the terminal to update monitoring model 61 to improve the monitoring accuracy of the terminal. If the network-side device determines that the performance status is abnormal, it indicates that the performance of the CSI feedback link is poor. Therefore, the network-side device can trigger an update or switch between the second model 62 and the third model 63, or it can trigger a fallback to the traditional solution, i.e., using a non-AI / ML CSI feedback method, thus stopping the use of model-based CSI feedback. In this way, since model monitoring on the network side can be triggered based on the user-side model monitoring results, and label information is only reported when the user side determines that the model's performance status or working state is poor, i.e., a complete CSI report for terminal prediction or measurement is performed, it can achieve model performance monitoring and avoid the overhead of frequent label information reporting.

[0230] It should be noted that in this article, CSI information, channel information, and full channel information have the same meaning and can be used interchangeably. They all refer to the CSI measured or predicted by the terminal.

[0231] The wireless model monitoring method provided in this application will be described in detail below with reference to the accompanying drawings, through some embodiments and application scenarios.

[0232] Figure 7 is a flowchart illustrating a model monitoring method provided in some embodiments of this application. As shown in Figure 7, the model monitoring method may include the following steps 700.

[0233] Step 700: When the terminal determines that the performance status of the first model is the first state, it sends the first information to the network-side device.

[0234] The first information can be used by network-side devices to determine the performance status of the first model, and the first model can be used for CSI feedback.

[0235] It is understandable that the transmission of the first information is determined by the performance status of the first model as determined by the terminal; that is, whether the first information is transmitted is determined by the performance status of the first model as determined by the terminal.

[0236] In some embodiments of this application, the performance state of the first model can characterize the CSI feedback accuracy of the first model.

[0237] In some embodiments of this application, the performance state of the first model may include: a first state and a second state.

[0238] In some embodiments of this application, the first state may include an abnormal performance state, and the second state may include a normal performance state.

[0239] For example, if the terminal determines that the performance status of the first model is abnormal, it means that the accuracy of the CSI fed back by the first model has deteriorated, such as the fed-back CSI not meeting the accuracy or precision requirements of CSI feedback.

[0240] For example, if the terminal determines that the performance state of the first model is the second state, it means that the accuracy of the CSI fed back by the first model is high, such as the CSI feedback meeting the accuracy or precision requirements of CSI feedback.

[0241] Thus, since the first state includes performance state abnormality, and whether the performance state of the first model is abnormal can characterize the CSI feedback accuracy of the first model, the terminal can monitor whether the CSI feedback accuracy of the first model meets the CSI feedback accuracy requirements by determining whether the performance state of the first model is abnormal.

[0242] In some embodiments of this application, when the terminal determines that the performance state of the first model is in a first state, the terminal can send first information to the network-side device. Thus, since the terminal only sends first information to the network-side device when it determines that the performance state of the first model is abnormal, air interface overhead can be saved.

[0243] In some embodiments of this application, if the terminal determines that the performance state of the first model is in the second state, the terminal may not send the first information to the network-side device. Thus, since the terminal does not need to send the first information to the network-side device when it determines that the performance state of the first model is in the second state, air interface overhead can be saved.

[0244] In some embodiments of this application, the first model may include a second model and a third model, wherein the second model may be used to generate CSI feedback and the third model may be used to recover CSI.

[0245] In some embodiments of this application, the CSI feedback generated by the second model can be the bitstream sequence corresponding to the CSI.

[0246] In some embodiments of this application, the performance status of the first model can be used to evaluate the combined performance of the second and third models, such as the CSI feedback performance of the first model.

[0247] It is understandable that the second model can be called an encoder or CSI feedback generation model, and the third model can be called a decoder or CSI recovery model.

[0248] In some embodiments of this application, during the model inference phase, i.e., the CSI feedback phase, a second model is deployed on the terminal side and a third model is deployed on the network side. For example, during the CSI feedback phase, the second model can be deployed within the terminal, and the third model can be deployed within the network-side device.

[0249] In some embodiments of this application, the input information of the second model may include any of the following:

[0250] CSI measured at the current moment;

[0251] Predict the CSI at at least one moment;

[0252] The CSI measured at the current time and the CSI feature information output by the second model at the previous time.

[0253] In some embodiments of this application, the CSI feature information output by the second model at the previous time step can be characterized by the feature output of the latent vector space of the second model at past time steps, and does not have explicit physical meaning.

[0254] For example, the CSI feature information output by the second model at the previous time step can be the feature output of the latent vector space when the second model outputs the CSI measured at the previous time step as CSI feedback; or it can be the feature of the latent vector space of the third model when the third model outputs the CSI measured at the previous time step.

[0255] In some embodiments of this application, the CSI feature information output by the second model at the previous time step has a different format than the CSI measured at the current time step.

[0256] For further descriptions of CSI feature information, please refer to the relevant descriptions of "historical information" in the second type of CSI feedback architecture mentioned above.

[0257] In some embodiments of this application, "current moment" can be any of the following: frame, frame group, subframe, subframe group, time slot, time slot group, symbol, symbol group.

[0258] It can be understood that a frame group can be a group or set of at least two frames, a subframe group can be a group or set of at least two subframes, a time slot group can be a group or set of at least two time slots, and a symbol can be a group or set of at least two symbols.

[0259] In some embodiments of this application, when the input information of the second model includes the predicted CSI at at least one moment, the at least one moment can be one, two, three or more moments, and there is no specific limitation.

[0260] In some embodiments of this application, the output information of the second model may include any of the following:

[0261] CSI feedback for CSI measured at the current moment;

[0262] CSI feedback for at least one predicted CSI moment;

[0263] The CSI feedback of the CSI measured at the current time, and the CSI characteristic information of the CSI measured at the current time.

[0264] It should be noted that the CSI feedback can be obtained by the terminal compressing or encoding the CSI through the second model.

[0265] In some embodiments of this application, "CSI feature information of CSI measured at the current time" is: the feature output representation of the latent vector space output when the second model outputs the CSI feedback of CSI measured at the current time, which does not have explicit physical meaning.

[0266] In some embodiments of this application, the format of "CSI feedback of CSI measured at the current time" is different from the format of "CSI feature information of CSI measured at the current time".

[0267] It is understandable that the input information and output information of the second model may differ for different CSI feedback architectures.

[0268] The following example illustrates the input and output information of the second model, using the input of the second model as the first input information and the output of the second model as the first output information.

[0269] Example 1: In the first type of CSI feedback architecture described above, the first input information includes the CSI measured at the current time, and the first output information includes a set of bit stream sequences for reporting to the network side; wherein, the set of bit stream sequences is: the CSI feedback corresponding to the CSI measured at the current time.

[0270] Example 2: In the second CSI feedback architecture described above, namely the AI-based spatial-frequency-time joint CSI compressed feedback implementation method, the first input information may include the CSI measured at the current time and the CSI feature information of the previous time output by the second model; the first output information may include a set of bit stream sequences for reporting to the network side, and may also include the CSI feature information of the CSI measured at the current time.

[0271] Example 3: In the third CSI feedback architecture described above, namely the AI-based joint CSI prediction and compression implementation method, the first input information may include CSI information at multiple predicted times, and the first output information includes a set of bit stream sequences for reporting to the network side.

[0272] Thus, since the input information of the second model can include the CSI measured at the current time, or the CSI predicted at at least one time, or the CSI measured at the current time and the CSI feature information output by the second model at the previous time, the second model can generate CSI feedback in different ways, thereby improving the flexibility of generating CSI feedback.

[0273] In some embodiments of this application, the input information of the third model may include any of the following:

[0274] CSI feedback from the terminal, such as CSI feedback of CSI measured at the current time, or CSI feedback of CSI predicted at multiple times;

[0275] The CSI feedback from the terminal and the CSI feature information output by the third model at the previous moment.

[0276] In some embodiments of this application, the CSI feature information output by the third model at the previous time step can be characterized by the feature output of the latent vector space of the third model at past time steps, and does not have explicit physical meaning.

[0277] For example, the CSI feature information output by the third model at the previous time step can be the feature output of the latent vector space when the third model outputs the CSI measured at the previous time step as CSI feedback; or it can be the feature of the latent vector space of the third model when the third model outputs the CSI measured at the previous time step.

[0278] In some embodiments of this application, the CSI feature information output by the third model at the previous time step has a different format than the CSI measured at the current time step.

[0279] It is understandable that for other descriptions of the CSI feature information output by the third model at the previous moment, please refer to the relevant description of "historical information" in the second type of CIS feedback architecture mentioned above.

[0280] In some embodiments of this application, the output information of the third model may include any of the following:

[0281] The CSI obtained by recovering the CSI feedback from the terminal;

[0282] The CSI feedback from the terminal is recovered to obtain the CSI and the CSI characteristic information of the measurement time corresponding to the CSI feedback.

[0283] In some embodiments of this application, "CSI feature information at the measurement time corresponding to CSI feedback" is: the feature output representation of the latent vector space output by the third model when the CSI recovery at the measurement time is output, which does not have explicit physical meaning.

[0284] For example, if the CSI feedback from the terminal includes the CSI feedback of the CSI measured at time 1, then the third model can output the recovered CSI measured at time 1 and the CSI feature information of the measured CSI at time 1. This CSI feature information can also be called the historical information of the measured CSI at time 1.

[0285] In some embodiments of this application, it is assumed that the input information of the third model is the second input information and the output information of the third model is the second output information. Then, the second input information and the second output information may include different contents in different CSI feedback architectures.

[0286] Example 4, in conjunction with Example 1 above, in the first type of CSI feedback architecture, the second input information only includes the bit stream sequence reported in the first output information, that is, the CSI feedback of the CSI measured at the current time; the second output information can be the CSI at the current time recovered by the third model.

[0287] Example 5, in conjunction with Example 2 above, in the second type of CSI feedback architecture, the second input information may include the bit stream sequence reported in the first output information, namely the CSI feedback of the CSI measured at the current time and the historical CSI feature information of the previous time output by the third model; the second output information may include the recovered CSI at the current time and the historical CSI feature information at the current time.

[0288] Example 6, in conjunction with Example 3 above, in the third CSI feedback architecture described above, the second input information may include the bit stream sequence reported in the first output information, i.e., the CSI feedback of the predicted CSI at multiple time points; the second output information may include the predicted CSI at multiple time points recovered by the third model.

[0289] In some embodiments of this application, the performance status of the first model can be monitored by monitoring the model.

[0290] The input information of the monitoring model can be any of the following: the output information of the second model, or the CSI feedback in the output information of the second model;

[0291] The output information of the monitoring model can be any of the following: the CSI recovered from the CSI feedback in the output information of the second model, or the first performance index. The first performance index can be used to characterize the similarity or difference between the CSI and the recovered CSI.

[0292] It is understandable that the terminal can evaluate the second and third models based on the first performance indicator, which is the performance status of the first model used for CSI feedback.

[0293] It should be noted that in this application, the terminal can simulate the CSI recovery process of the third model through the monitoring model. In other words, the terminal can use the CSI recovered by the monitoring model as the CSI recovered by the third model, thereby realizing performance monitoring of the second and third models on the terminal side.

[0294] In some embodiments of this application, the first performance index may indicate the similarity or difference between the CSI recovered by the monitoring model from the CSI feedback generated by the second model and the CSI monitored or measured by the terminal, or the first performance index may indicate the similarity or difference between the output information of the monitoring model and the input information of the second model.

[0295] In some embodiments of this application, the input information of the monitoring model and the input information of the third model may be the same, or the input information of the monitoring model may include CSI feedback from the input information of the third model. This can improve the similarity between the CSI recovered by the monitoring model and the CSI recovered by the third model, thereby improving the accuracy of the terminal side in monitoring the performance status of the first model.

[0296] In some embodiments of this application, it is assumed that the input information of the monitoring model is the third input information and the output information of the monitoring model is the third output information. Then, the third input information may include different contents in different CSI feedback architectures.

[0297] For example, in the first, second, or third CSI feedback architecture described above, the third input information can be the same as the second input information; or the third input information can be only the bitstream sequence from the first output information in the second input information, which can be determined according to actual usage requirements, and this application does not limit it. In the first, second, or third CSI feedback architecture described above, the third output information may include: (1) the predicted CSI at the current time or multiple future times, so that the terminal can calculate the first performance index based on the third output information and the first input information. Or, (2) the third output information may include the first performance index.

[0298] For example, in the first type of CSI feedback architecture described above, the third input information is the output information of the second model or the CSI feedback in the output information, and the third output information is the first performance index, or the CSI recovered by the monitoring model from the CSI feedback.

[0299] For example, in the second CSI feedback architecture described above, the third input information is the output information of the second model, and the third output information is the first performance index, or the CSI recovered by the monitoring model from the CSI feedback in the output information of the second model.

[0300] For example, in the third CSI feedback architecture described above, the third input information is the output information of the second model, and the third output information is the first performance index, or the CSI recovered by the monitoring model from the CSI feedback in the output information of the second model.

[0301] In some embodiments of this application, in (1) above, the terminal can compare the CSI recovered by the monitoring model with the CSI measured or predicted by the terminal (i.e., tag information or true value) to obtain a first performance index. In (2) above, the monitoring model can recover the CSI feedback to obtain the recovered CSI, and then compare the recovered CSI with the CSI measured or predicted by the terminal to obtain and output the first performance tag.

[0302] In some embodiments of this application, the method for determining the first performance metric may differ depending on the CSI feedback architecture during a single model inference and monitoring process.

[0303] Specifically, in the first or second CSI feedback architecture described above, the first performance indicator can be based on the CSI at a single moment, such as the calculation result of the CSI measured by the terminal at a single moment. For example, the first performance indicator is obtained by comparing the CSI measured by the terminal at a single moment with the CSI recovered by the CSI feedback of the monitoring model.

[0304] In the third CSI feedback architecture described above, the first performance index can be the average of the calculated results of the predicted CSI at multiple times. For example, for the CSI predicted by the terminal at multiple times, referred to as multiple predicted CSI, a performance index can be determined based on each predicted CSI and the CSI feedback of each predicted CSI, and then the average of the performance indices corresponding to the multiple predicted CSI can be calculated to obtain the first performance index.

[0305] It can be understood that "determining a performance metric based on each predicted CSI and the CSI feedback of each predicted CSI" may include: using the predicted CSI as label information or standard information, comparing the predicted CSI with the CSI recovered from the CSI feedback of the predicted CSI to obtain a performance metric.

[0306] In some embodiments of this application, when the output information of the monitoring model is the CSI recovered from the CSI feedback in the output information of the second model, the first performance metric may include at least one of the following: the squared generalized cosine similarity (SGCS) between the output information of the monitoring model and the input information of the second model, the normalized mean squared error (NMSE) between the output information of the monitoring model and the input information of the second model, and the mean squared error (MSE) between the output information of the monitoring model and the input information of the second model. Of course, the first performance metric may also include any other metric that can measure or evaluate the performance status of the model.

[0307] It is understandable that the larger the SGCS value, the better the model's performance. A larger SGCS value indicates that two vectors are more similar, such as the higher the similarity between the feature vector input to the second model's CSI and the feature vector output by the third model. On the other hand, the smaller the values ​​of NMSE and MSE, the better the model's performance. Smaller NMSE and MSE values ​​indicate a smaller gap between the CSI recovered by the third model and the CSI measured or predicted by the terminal, thus resulting in better CSI feedback performance of the first model.

[0308] Thus, since the input information of the monitoring model is the output information of the second model or the CSI feedback in the output information of the second model, and the output information of the monitoring model is the CSI or the first performance index recovered from the CSI feedback in the output information of the second model; while in the CSI feedback process, the CSI feedback in the output information of the second model is the input information of the third model, and the output information of the third model includes the CSI recovered from the CSI feedback, the terminal can realize the simulation of the processing of the CSI feedback of the second model output by the third model on the terminal side, or determine the similarity or difference between the CSI output by the third model and the CSI input to the second model through the monitoring model, thereby improving the monitoring performance of model monitoring on the terminal side.

[0309] In some embodiments of this application, the format of the first information can be any of the following: floating-point number (float) 32, float 16, eType II, doppler eType II, etc., any format with high data precision.

[0310] In some embodiments of this application, the method for triggering the transmission of the first information may include the following method 1 and method 2.

[0311] Method 1: Send the first information based on the model monitoring results on the terminal side.

[0312] In some embodiments of this application, the first information is sent by the terminal when it determines that the performance state of the first model is in a first state.

[0313] Thus, since the terminal only reports the tag information, such as the first information, used by the network side to monitor the performance status of the first model when the terminal detects that the performance status of the first model is in the first state, the terminal can save the signaling overhead of reporting tag information.

[0314] Method 2: Send the first information based on instructions from the network-side device.

[0315] In some embodiments of this application, the first information is sent by the terminal when it determines that the performance state of the first model is a first state and receives the first indication information. The first indication information can be used to indicate the information to be reported to determine the performance state of the first model.

[0316] In some embodiments of this application, the first information may also be sent when the terminal receives the first instruction information.

[0317] For example, if the performance state of the first model determined by the terminal is the first state, the terminal can send the first information to the network-side device after receiving the first indication information.

[0318] Thus, since the first information is reported after the terminal side detects that the performance status of the first model is in the first state and receives the reporting indication information (i.e., the first indication information) from the network side device, it is convenient for the network side device to accurately listen to or receive the first information, thereby improving the reporting accuracy of the first information.

[0319] In some embodiments of this application, after receiving the first indication information, if the performance status of the first model determined by the terminal is the second status, the terminal can send the first information to the network-side device.

[0320] In some embodiments of this application, the first information may include at least the input information of the second model. For a description of the input information of the second model, please refer to the relevant descriptions of the input information of the second model in the above embodiments, which will not be repeated here.

[0321] Thus, since the first information can include at least the input information of the second model, the network-side device can use the first information as the label information for model monitoring, thereby accurately determining whether the performance status of the first model is abnormal.

[0322] In the model monitoring method provided in this application embodiment, since the terminal can send first information to the network-side device when it determines that the performance state of the first model is in a first state, instructing the network-side device to re-determine the performance state of the first model, it can realize the monitoring of model performance by both the terminal side and the network side, thereby improving the accuracy of model performance monitoring. Furthermore, since whether the first information is sent is determined by the performance state of the first model determined by the terminal, and not by feeding back the first information to the network-side device every time model monitoring is performed, air interface overhead can be saved.

[0323] It is understood that by adopting the model monitoring method provided in the application embodiment, model monitoring on the network side can be triggered based on the model monitoring results on the terminal side; and only when the terminal side considers the working state of the model to be poor, that is, the performance state of the model is abnormal, will the label information, such as the first information, be reported. Therefore, the air interface overhead for reporting label information can be saved, and the monitoring performance of the model can be guaranteed.

[0324] In some embodiments of this application, the model monitoring method provided in this application may further include step 701.

[0325] Step 701: The terminal determines the second information.

[0326] The second information can be used to indicate the performance status of the first model determined by the terminal.

[0327] In some embodiments of this application, the terminal can determine the second information through a monitoring model. In other words, the second information is obtained by the terminal through monitoring by the monitoring model.

[0328] In some embodiments of this application, the second information may be used to indicate whether the performance state of the first model determined by the terminal is a first state or a second state.

[0329] Step 701 can be performed before step 700.

[0330] Thus, since the terminal can determine the second information, and the second information can indicate the performance status of the first model determined by the terminal, the terminal can send the first information to the network-side device only when it determines that the performance status of the first model is the first state. Therefore, the accuracy of the first information can be improved, thereby avoiding unnecessary information transmission and saving signaling overhead.

[0331] In some embodiments of this application, the second information described above may include any of the following:

[0332] The first performance metric can be used to characterize the similarity or difference between the CSI and the recovered CSI.

[0333] The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

[0334] In some embodiments of this application, the first monitoring result can indicate whether the performance state of the first model is abnormal.

[0335] For example, the first monitoring result can be indicated by a single bit, where a "1" bit indicates that the performance status of the first model is normal, and a "0" bit indicates that the performance status of the first model is abnormal; or, a "1" bit indicates that the performance status of the first model is abnormal, and a "0" bit indicates that the performance status of the first model is normal. Of course, the first monitoring result can also be represented in any other possible way, such as by quantization.

[0336] In some embodiments of this application, when the first monitoring result indicates an abnormal performance state of the first model, that is, when the first monitoring result indicates a first state, it means that the model monitoring on the terminal side believes that the model currently used for CSI feedback, that is, the first model, is not performing well.

[0337] In some embodiments of this application, the second information is obtained by the terminal through a monitoring model; the aforementioned first performance index can be used to indicate the similarity or difference between the output information of the monitoring model and the input information of the second model.

[0338] For example, when the output information of the monitoring model is the CSI recovered from the CSI feedback in the output information of the second model, the terminal can compare the CSI output by the monitoring model with the CSI output by the second model to obtain the first performance index.

[0339] In some embodiments of this application, the first performance metric may indicate the similarity or difference between the CSI recovered by the monitoring model and the CSI measured or predicted by the terminal (i.e., the CSI input to the second model).

[0340] In some embodiments of this application, the first performance index may include K performance indices, where K can be a positive integer;

[0341] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0342] If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0343] If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0344] In some embodiments of this application, the first performance index may include K performance indices, where K can be a positive integer;

[0345] If K equals 1, and one performance indicator does not satisfy a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the second state; or,

[0346] If K is greater than 1, and the mean of the K performance indicators does not satisfy a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the second state; or,

[0347] If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is less than M, then the first monitoring result is used to indicate that the performance state of the first model is the second state, where M is a positive integer less than or equal to K.

[0348] In some embodiments of this application, the performance indicators and preset relationships may differ.

[0349] For example, if the performance metric is GCS, the preset relationship is: the performance metric is greater than or equal to the first threshold.

[0350] For example, if the performance metric is NMSE or MSE, the default relationship is: the performance metric is less than or equal to the first threshold.

[0351] Thus, when the first performance indicator includes K performance indicators, the first monitoring result can be determined in different ways based on these K performance indicators, thereby increasing the flexibility of the method for determining the first monitoring result.

[0352] In some embodiments of this application, K, M, or the first threshold can be predefined by the protocol as a preset value; or, K, M, or the first threshold can be configured by the network-side device through downlink signaling.

[0353] In some embodiments of this application, K is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration. This improves the flexibility of how the terminal obtains or determines K.

[0354] In some embodiments of this application, M is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0355] In some embodiments of this application, the first threshold is any one of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0356] For example, K, M, and the first threshold can all be configured in the CSI reporting configuration.

[0357] For example, K, M, and the first threshold are all indicated by indicator fields added in the CSI reporting configuration.

[0358] For example, K and M are configured for CSI reporting, and the first threshold is predefined by the protocol.

[0359] For example, K and M are configured for CSI reporting, and the first threshold is activated from the set of candidate parameters for RRC signaling configuration.

[0360] For example, K and M are indicated by the indication fields added in the CSI reporting configuration, and the first threshold is activated in the candidate parameter set of the RRC signaling configuration.

[0361] For example, K, M, and the first threshold are all active parameters in the candidate parameter set of the RRC signaling configuration.

[0362] It is understood that the above example is only used to illustrate K, M and the first threshold. In actual implementation, K, M and the first threshold can also be determined or obtained by any other possible means.

[0363] In some embodiments of this application, when K, M and the first threshold are indicated by indication fields added in the CSI reporting configuration, K, M and the first threshold are indicated by different indication fields added in the CSI reporting configuration, such as indication fields used to indicate K, M and the first threshold in the CSI reporting configuration.

[0364] In some embodiments of this application, when K, M, and the first threshold are all active in the candidate parameter set configured by RRC signaling: the network-side device can configure a candidate parameter set through RRC signaling, and the Media Access Control-Control Element (MAC-CE) selects one parameter configuration from the candidate parameter set for activation; or, the network-side device can configure a candidate parameter set through RRC signaling, and the MAC-CE selects a portion of the parameter configurations from the candidate parameter set, and the DCI signaling dynamically triggers (e.g., activates) one of the parameter configurations selected by the MAC-CE.

[0365] In some embodiments of this application, when K, M, and the first threshold are all active in the candidate parameter set of RRC signaling configuration:

[0366] This set of candidate parameters can include multiple parameter configurations, each configuration containing: a K, an M, and a first threshold; or...

[0367] The candidate parameter set can include three subsets, namely subset 1 to subset 3. Subset 1 is configured with K, subset 2 is configured with M, and subset 3 is configured with a first threshold. This increases the flexibility in selecting or activating M, K, and the first threshold.

[0368] In some embodiments of this application, K, M and the first threshold may be configured through different RRC signaling or through one RRC signaling, and this application does not limit them.

[0369] It is understandable that if the first monitoring result is determined based on whether the average of K performance indicators and the first threshold satisfy a preset relationship, the network-side device may not need to configure M.

[0370] Thus, since the second information may include the first performance index or the first monitoring result, and the first monitoring result can directly indicate the performance status of the first model, and the first performance index can indirectly indicate the performance status of the first model, the flexibility of the second information in indicating the performance status of the model can be improved.

[0371] Furthermore, it can improve the flexibility of the way the terminal side determines the performance status of the first model, thereby improving the monitoring flexibility of the terminal side in model monitoring.

[0372] In some embodiments of this application, the model monitoring method provided in this application may further include step 702.

[0373] Step 702: The terminal sends the second information to the network-side device.

[0374] It is understood that the execution order of steps 700 and 702 is not limited. For example, step 700 can be executed first, followed by step 702; or step 702 can be executed first, followed by step 700; or steps 700 and 702 can be executed simultaneously.

[0375] In some embodiments of this application, the second information is sent by the terminal when the performance state indicated by the second information is the first state.

[0376] In some embodiments of this application, the first indication information may be sent by the network-side device to the terminal when the network-side device receives the second information reported by the terminal and the performance status indicated by the second information is the first status.

[0377] In some embodiments of this application, when the second information indicates that the performance state of the first model is in the first state, it indicates that the model monitoring results on the terminal side believe that the current model used for CSI feedback is not performing well. Thus, the terminal can send the second information to the network side device so that the network side device can know the model monitoring results on the terminal side.

[0378] For example, when the performance status indicated by the second information is abnormal, the terminal can send a first monitoring result or a first performance indicator to the network-side device.

[0379] In some embodiments of this application, when the second information includes the first monitoring result, the second information can be reported through a 1-bit indication field in the Uplink Control Information (UCI).

[0380] For example, when a bit in the UCI is "1", it means that the terminal determines the performance state of the first model as the first state.

[0381] In some embodiments of this application, regardless of whether the performance status indicated by the second information is abnormal, the terminal can send the second information to the network-side device so that the network-side device can know the performance monitoring results of the first model on the terminal side.

[0382] For example, when a bit in the UCI is "1", it means that the terminal determines the performance state of the first model to be in the first state; when a bit in the UCI is "0", it means that the terminal determines the performance state of the first model to be in the second state.

[0383] For example, when a bit in the UCI is "0", it means that the terminal determines the performance state of the first model to be in the first state; when a bit in the UCI is "1", it means that the terminal determines the performance state of the first model to be in the second state.

[0384] Thus, since the terminal can report the second information to the network-side device when the performance state of the first model determined by the terminal is the first state, the air interface overhead on the terminal side can be further saved.

[0385] In some embodiments of this application, when the second information includes a first performance indicator, the first performance indicator can be reported in a non-quantized manner or in a quantized manner using Q bits, where Q can be a positive integer. Thus, since the first performance indicator can be reported in either a non-quantized or quantized manner, the flexibility of the reporting method for the first performance indicator can be improved.

[0386] In some embodiments of this application, Q can be any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0387] In some embodiments of this application, when Q is active in the candidate parameter set configured by RRC signaling: the network-side device can configure a candidate parameter set through RRC signaling, and the MAC-CE selects one parameter configuration from the candidate parameter set for activation; or, the MAC-CE selects some parameter configurations from the candidate parameter set, and the DCI signaling dynamically triggers (e.g., activates) one of the parameter configurations selected by the MAC-CE.

[0388] In some embodiments of this application, Q can be configured simultaneously with or separately from M, K and the first threshold.

[0389] In some embodiments of this application, the candidate transmission resource for the second information is the first resource, and the candidate transmission resource for the first information is the second resource; the first resource or the second resource is any one of the following: predefined by the protocol, configured by the CSI reporting configuration, configured by the RRC signaling, or pre-configured.

[0390] In some embodiments of this application, the second resource can be used to report the first information during the model monitoring process. Correspondingly, the first resource can be used to report CSI feedback and the second information during the inference phase.

[0391] In some embodiments of this application, the first resource may include at least one of frequency domain resources and time domain resources.

[0392] In some embodiments of this application, the second resource may include at least one of frequency domain resources and time domain resources.

[0393] In some embodiments of this application, frequency domain resources can be: subcarriers, carriers, resource blocks (RBs), resource elements or resource particles (REs), and subbands.

[0394] In some embodiments of this application, the time-domain resource can be any of the following: frame, frame group, subframe, subframe group, time slot, time slot group, symbol, symbol group.

[0395] In some embodiments of this application, the first resource and the second resource are different. For example, the first resource and the second resource are different symbols or different RBs.

[0396] In some embodiments of this application, when the reporting of the first information is not required or there is no need to report the first information, the second resource can be left idle or can be used for the reporting of other information.

[0397] In some embodiments of this application, when the first resource or the second resource is configured by the CSI reporting configuration, the network-side device may by default associate the first resource and the second resource configured in the same CSI reporting configuration with each other and use them to report the second information and the first information, respectively.

[0398] In some embodiments of this application, the method of reporting the configuration of the first or second resource via CSI may include any one of the following methods (i), (ii), and (iii):

[0399] Method (1): Configure the first and second resources in a CSI reporting configuration;

[0400] Method (II): Configure the first resource in one type I CSI reporting configuration, and configure the first resource and the second resource in one type II CSI reporting configuration;

[0401] Method (3): Configure the first resource in one type 1 CSI reporting configuration and configure the second resource in one type 2 CSI reporting configuration.

[0402] In some embodiments of this application, the first type of CSI reporting configuration differs from the second type of CSI reporting configuration.

[0403] For example, the first type of CSI reporting configuration is a periodic CSI reporting configuration, and the second type of CSI reporting configuration is a semi-continuous CSI reporting configuration.

[0404] For example, the first type of CSI reporting configuration is a periodic CSI reporting configuration, and the second type of CSI reporting configuration is a non-periodic CSI reporting configuration.

[0405] For example, the first type of CSI reporting configuration is a semi-persistent CSI reporting configuration, and the second type of CSI reporting configuration is a periodic CSI reporting configuration.

[0406] For example, the first type of CSI reporting configuration is a semi-persistent CSI reporting configuration, and the second type of CSI reporting configuration is a non-periodic CSI reporting configuration.

[0407] For example, the first type of CSI reporting configuration is a non-periodic CSI reporting configuration, while the second type of CSI reporting configuration is a periodic CSI reporting configuration.

[0408] For example, the first type of CSI reporting configuration is a non-periodic CSI reporting configuration, and the second type of CSI reporting configuration is a semi-persistent CSI reporting configuration.

[0409] In some embodiments of this application, in method (a) described above, the network-side device can configure the first resource and the second resource separately in the same CSI configuration report. The number of CSI configuration reports with both the first and second resources configured is not limited.

[0410] For example, network-side devices can configure a first resource and a second resource in each CSI reporting configuration.

[0411] For example, network-side devices can configure a first resource and a second resource in at least one CSI reporting configuration sent within a preset time period.

[0412] In some embodiments of this application, in method (ii) above, the network-side device can configure a first resource in each first type of CSI reporting configuration, and configure a first resource and a second resource in each second type of CSI reporting configuration. The first resource configured in the first type of CSI reporting configuration and the first resource configured in the second type of CSI reporting configuration can be the same or different.

[0413] In some embodiments of this application, in method (ii) above, the network-side device can configure a first resource in each first type of CSI reporting configuration within a preset time period, and configure a first resource and a second resource in each second type of CSI reporting configuration within the preset time period. The first resource configured in the first type of CSI reporting configuration and the first resource configured in the second type of CSI reporting configuration can be the same or different.

[0414] In some embodiments of this application, in method (iii) described above, the network-side device can configure a first resource in each first type of CSI reporting configuration, and configure a first resource and a second resource in each second type of CSI reporting configuration. The first resource configured in the first type of CSI reporting configuration and the first resource configured in the second type of CSI reporting configuration can be the same or different.

[0415] In some embodiments of this application, in method (iii) described above, the network-side device can configure a first resource in each first type of CSI reporting configuration within a preset time period, and configure resources in each second type of CSI reporting configuration within the preset time period. The first resource configured in the first type of CSI reporting configuration and the first resource configured in the second type of CSI reporting configuration can be the same or different.

[0416] It should be noted that the above-mentioned preset time period can be the supervision or monitoring time period of the first model.

[0417] It is understood that in the above methods (ii) and (iii), the network-side device can, based on the reporting requirements of the first information, indicate the selection range of the second information or the sending resources of the first information through additional signaling, such as indicating to select from the resources configured in the first type of CSI reporting configuration, and / or indicating to select from the resources configured in the second type of CSI reporting configuration.

[0418] In some embodiments of this application, the methods of configuring the first and second resources through CSI reporting configuration are different, and the transmission resources used by the terminal to send the second information and the first information may be different.

[0419] For example, in the above method (a), the terminal can use the first resource configured by any CSI reporting configuration to send the second information, and use the second resource configured by any SCI reporting configuration to send the first information.

[0420] For example, in the above method (II), the first type of CSI reporting configuration includes A CSI reporting configurations, each of which is configured with a first resource. The configuration identifiers of the A CSI reporting configurations are as follows: The second type of CSI reporting configuration includes B CSI reporting configurations, each of which is configured with a first resource and a second resource. For example, each of the second type of CSI reporting configurations is configured with one first resource and one second resource. The configuration identifiers of the B CSI reporting configurations are as follows: There are a total of B, where A and B are both positive integers.

[0421] When the reporting of the first information is not required, the network device can pre-configure via RRC signaling, or configure via RRC signaling, or activate via MAC-CE / DCI, to instruct the terminal to select the first resource configured in one of the A CSI reporting configurations, so that the terminal can use the first resource configured in that CSI reporting configuration to send the second information. When the reporting of the first information is required, the network device can reconfigure via RRC signaling, or activate via MAC-CE / DCI, to instruct the terminal to activate the resource configured in one of the B CSI reporting configurations, so that the terminal can use the first resource configured in a second type of CSI reporting configuration to send the second information, and use the second resource configured in that second type of CSI reporting configuration to send the first information.

[0422] For example, in the above method (iii), the first type of CSI reporting configuration includes C CSI reporting configurations, each of which is configured with a first resource. The configuration identifiers of the C CSI reporting configurations are as follows: The second type of CSI reporting configuration includes D CSI reporting configurations, each of which is configured with a second resource. For example, each of the second type of CSI reporting configurations is configured with one first resource and one second resource. The configuration identifiers of the D CSI reporting configurations are as follows: There are a total of D, where C and D are both positive integers.

[0423] When the reporting of the first information is not required, the network-side device can instruct the terminal to select the first resource configured in one of the C CSI reporting configurations as the transmission resource for the second information through RRC signaling pre-configuration, RRC signaling configuration, or MAC-CE / DCI activation. When the reporting of the first information is required, the network-side device can instruct the terminal to additionally select the second resource configured in one of the D CSI reporting configurations as the transmission resource for the first information through RRC signaling pre-configuration or MAC-CE / DCI activation.

[0424] It is understandable that in method (iii), the terminal needs to use a first resource configured in a first type of CSI reporting configuration to send the second information, and use a second resource configured in a second type of CSI reporting configuration to send the first information.

[0425] Thus, since configuring the first or second resource through CSI reporting configuration includes: configuring the first and second resources in one CSI reporting configuration; or configuring the first resource in one first-type CSI reporting configuration and configuring the first and second resources in one second-type CSI reporting configuration; or configuring the first resource in one first-type CSI reporting configuration and configuring the second resource in one second-type CSI reporting configuration, the flexibility of configuring the first or second resource through CSI reporting configuration can be improved.

[0426] In some embodiments of this application, the first or second resource is pre-configured when the terminal activates the AI- or ML-based CSI feedback function. Thus, since the first or second resource is pre-configured when the terminal activates the AI- or ML-based CSI feedback function, resource waste can be avoided.

[0427] In some embodiments of this application, in order for the network-side device to know that the terminal has activated the AI / ML-based CSI feedback function, the terminal may send an indication message to the network-side device after activating the CSI feedback function to instruct the terminal to activate the AI / ML-based CSI feedback function; or, assuming that the terminal activates the AI / ML-based CSI feedback function based on the first activation indication message sent by the network-side device, then the network-side device can determine that the terminal activates the AI / ML-based CSI feedback function after a preset time after the sending time of the first activation indication message.

[0428] In some embodiments of this application, when the network-side device activates the AI / ML-based CSI feedback function or activates the first model, the network-side device can pre-configure the first and second resources through RRC signaling.

[0429] Thus, since the terminal can send the second information to the network-side device, the network-side device can know the terminal's monitoring results of the first model based on the second information, and determine whether the monitoring results conflict with the monitoring results determined by the network-side device, thereby further improving the monitoring accuracy of the CSI feedback model.

[0430] In some embodiments of this application, the model monitoring method provided in this application may further include steps 703 and 704.

[0431] Step 703: The terminal receives the second instruction information sent by the network-side device.

[0432] Step 704: The terminal updates the monitoring model used by the terminal based on the second indication information.

[0433] The second indication information can be used to indicate the monitoring model used by the updated terminal.

[0434] In some embodiments of this application, "updating the monitoring model" may include: updating the model parameters of the monitoring model to obtain better model monitoring results on the terminal side.

[0435] In some embodiments of this application, the monitoring model used by the terminal can be used to monitor the performance status of the first model.

[0436] It is understandable that when the terminal receives the second indication information, it means that the monitoring capability or accuracy of the monitoring results used by the terminal is poor. For example, the performance status of the first model determined by the terminal is different from the performance status of the first model determined by the network-side device. For instance, the performance status of the first model determined by the terminal is abnormal, but the performance status of the first model determined by the network-side device is the second state.

[0437] In some embodiments of this application, step 703 may be performed after step 700.

[0438] Thus, since the terminal can update the monitoring model used by the terminal based on the received second indication information, the accuracy of monitoring the performance of the first model on the terminal side can be improved.

[0439] In some embodiments of this application, the model monitoring method provided in this application may further include the following steps 705 and 706.

[0440] Step 705: The terminal receives the third instruction information sent by the network-side device.

[0441] Step 706: The terminal executes the operation corresponding to the third instruction information.

[0442] The third instruction information can be used to instruct any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback.

[0443] For example, if the third instruction information indicates that the model used to generate CSI feedback should be updated, the terminal can update the second model, such as updating the model parameters of the second model.

[0444] For example, if the third instruction indicates a switch to the model used to generate CSI feedback, the terminal can switch from the second model to the fourth model, where the fourth model is also the model used to generate CSI feedback. It is understood that the third instruction may carry the model identifier of the fourth model.

[0445] For example, if the third instruction message indicates to stop using the model used to generate CSI feedback, then the terminal can stop using the second model to generate CSI feedback and generate CSI feedback using the conventional method.

[0446] It is understandable that when the terminal receives the third indication information, it can indicate that the performance status of the first model determined by the terminal is the same as the performance status of the first model determined by the network-side device. For example, if both the terminal and the network-side device determine that the performance status of the first model is abnormal, that is, the CSI feedback performance of the first model is poor, the network-side device can instruct the terminal to update or switch the second model through the third indication information, or directly fall back to the traditional CSI feedback scheme to improve the accuracy of the CSI feedback generated by the terminal.

[0447] In some embodiments of this application, step 705 may be performed after step 700.

[0448] Thus, after the terminal sends the first information to the network-side device, it can receive the third instruction information and perform the corresponding operation, such as updating the model used to generate CSI feedback, switching the model used to generate CSI feedback, or stopping the use of the model used to generate CSI feedback. Therefore, the terminal can use an appropriate CSI feedback generation method to generate CSI feedback, thereby ensuring the accuracy of CSI feedback.

[0449] In some embodiments of this application, the model monitoring method provided in this application may further include the following step 707.

[0450] Step 707: The terminal sends its capability information to the network-side device. This capability information can be used to indicate at least one of the following:

[0451] Does the terminal have a primary capability? The primary capability can be the ability to monitor models or the ability to deploy monitoring models.

[0452] Does the terminal have a secondary capability? The secondary capability could be the ability to report information in a configured codebook quantization manner.

[0453] In some embodiments of this application, the terminal can send its capability information to the network-side device based on the capability reporting instruction information of the network-side device.

[0454] In some embodiments of this application, the first capability is the ability to monitor the performance status of the first model on the terminal side, or in other words, the terminal has the ability to deploy a monitoring model.

[0455] In some embodiments of this application, the "configured codebook quantization method" may include any of the following: a configured eTypeII codebook quantization method, or a configured doppler eTypeII codebook quantization method. It is understood that either of these two codebook quantization methods can achieve the reporting of high-precision tag information, such as the first piece of information.

[0456] In some embodiments of this application, whether the terminal has a second capability can also be described as: whether the terminal supports the reporting of first information.

[0457] It is understandable that when the terminal possesses both the first and second capabilities, it means that the terminal can perform model monitoring on the terminal side and can report the first information. In other words, the terminal performs the above step 700 while possessing both the first and second capabilities.

[0458] In some embodiments of this application, step 707 may be performed before step 700 or after step 702.

[0459] Thus, since the terminal can report capability information to the network-side device indicating whether the terminal has the first capability / second capability, the network-side device can know whether the terminal can realize the terminal-side monitoring of the first model and whether it can report the first information.

[0460] Figure 8 is a flowchart illustrating a model monitoring method provided in some embodiments of this application. As shown in Figure 8, the model monitoring method may include the following steps 801 and 802.

[0461] Step 801: The network-side device receives the second information sent by the terminal.

[0462] The second information can be used to indicate the performance status of the first model determined by the terminal, and the first model can be used for CSI feedback.

[0463] In some embodiments of this application, the performance state of the first model may include: a first state and a second state.

[0464] In some embodiments of this application, the first state may include an abnormal performance state, and the second state may include a normal performance state.

[0465] In some embodiments of this application, the second information may be used to indicate whether the performance state of the first model determined by the terminal is a first state or a second state.

[0466] In some embodiments of this application, the first model may include a second model and a third model, wherein the second model may be used to generate CSI feedback and the third model may be used to recover CSI.

[0467] In some embodiments of this application, the CSI feedback generated by the second model can be the bitstream sequence corresponding to the CSI.

[0468] In some embodiments of this application, the performance status of the first model may include: the accuracy of the CSI feedback generated by the second model, such as CSI encoding performance, and the CSI recovery performance of the third model, such as CSI decoding performance.

[0469] It is understandable that the second model can be called an encoder or CSI feedback generation model, and the third model can be called a decoder or CSI recovery model.

[0470] In some embodiments of this application, during the model inference phase, i.e., the CSI feedback phase, a second model is deployed on the terminal side and a third model is deployed on the network side. For example, during the CSI feedback phase, the second model can be deployed within the terminal, and the third model can be deployed within the network-side device.

[0471] In some embodiments of this application, the input information of the second model may include any of the following:

[0472] CSI measured at the current moment;

[0473] Predict the CSI at at least one moment;

[0474] The CSI measured at the current time and the CSI feature information output by the second model at the previous time.

[0475] In some embodiments of this application, the CSI feature information output by the second model at the previous time step can be characterized by the feature output of the latent vector space of the second model at past time steps, and does not have explicit physical meaning.

[0476] For example, the CSI feature information output by the second model at the previous time step can be the feature output of the latent vector space when the second model outputs the CSI measured at the previous time step as CSI feedback; or it can be the feature of the latent vector space of the third model when the third model outputs the CSI measured at the previous time step.

[0477] In some embodiments of this application, the CSI feature information output by the second model at the previous time step has a different format than the CSI measured at the current time step.

[0478] For further descriptions of CSI feature information, please refer to the relevant descriptions of "historical information" in the second type of CSI feedback architecture mentioned above.

[0479] In some embodiments of this application, "current moment" can be any of the following: frame, frame group, subframe, subframe group, time slot, time slot group, symbol, symbol group.

[0480] It can be understood that a frame group can be a group or set of at least two frames, a subframe group can be a group or set of at least two subframes, a time slot group can be a group or set of at least two time slots, and a symbol can be a group or set of at least two symbols.

[0481] In some embodiments of this application, when the input information of the second model includes the predicted CSI at at least one moment, the at least one moment can be one, two, three or more moments, and there is no specific limitation.

[0482] In some embodiments of this application, the output information of the second model may include any of the following:

[0483] CSI feedback for CSI measured at the current moment;

[0484] CSI feedback for at least one predicted CSI moment;

[0485] The CSI feedback of the CSI measured at the current time, and the CSI characteristic information of the CSI measured at the current time.

[0486] It should be noted that the CSI feedback can be obtained by the terminal compressing or encoding the CSI through the second model.

[0487] In some embodiments of this application, "CSI feature information of CSI measured at the current time" is: the feature output representation of the latent vector space output when the second model outputs the CSI feedback of CSI measured at the current time, which does not have explicit physical meaning.

[0488] In some embodiments of this application, the format of "CSI feedback of CSI measured at the current time" is different from the format of "CSI feature information of CSI measured at the current time".

[0489] In some embodiments of this application, the input information of the third model may include any of the following:

[0490] CSI feedback from the terminal, such as CSI feedback of CSI measured at the current time, or CSI feedback of CSI predicted at multiple times;

[0491] The CSI feedback from the terminal and the CSI feature information output by the third model at the previous moment.

[0492] In some embodiments of this application, the CSI feature information output by the third model at the previous time step can be characterized by the feature output of the latent vector space of the third model at past time steps, and does not have explicit physical meaning.

[0493] For example, the CSI feature information output by the third model at the previous time step can be the feature output of the latent vector space when the third model outputs the CSI measured at the previous time step as CSI feedback; or it can be the feature of the latent vector space of the third model when the third model outputs the CSI measured at the previous time step.

[0494] In some embodiments of this application, the CSI feature information output by the third model at the previous time step has a different format than the CSI measured at the current time step.

[0495] It is understood that for further descriptions of the CSI feature information output by the third model at the previous time step, please refer to the relevant description of "historical information" in the second CIS feedback architecture described above. In some embodiments of this application, the output information of the third model may include any of the following:

[0496] The CSI obtained by recovering the CSI feedback from the terminal;

[0497] The CSI feedback from the terminal is recovered to obtain the historical CSI feature information of the measurement time corresponding to the CSI feedback.

[0498] For further descriptions of the second and third models, please refer to the relevant descriptions in the above terminal-side method embodiments.

[0499] In some embodiments of this application, the performance state indicated by the second information can be either a normal performance state or an abnormal performance state, that is, the performance state indicated by the second information can be either a first state or a second state.

[0500] In some embodiments of this application, the second information may include any of the following:

[0501] The first performance metric can be used to characterize the similarity or difference between the CSI and the recovered CSI.

[0502] The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

[0503] In some embodiments of this application, the first monitoring result can indicate whether the performance state of the first model is abnormal.

[0504] For example, the first monitoring result can be indicated by a single bit, where a "1" bit indicates that the performance status of the first model is normal, and a "0" bit indicates that the performance status of the first model is abnormal; or, a "1" bit indicates that the performance status of the first model is abnormal, and a "0" bit indicates that the performance status of the first model is normal. Of course, the first monitoring result can also be represented in any other possible way, such as by quantization.

[0505] In some embodiments of this application, when the first monitoring result indicates an abnormal performance state of the first model, that is, when the first monitoring result indicates a first state, it means that the model monitoring on the terminal side believes that the model currently used for CSI feedback, that is, the first model, is not performing well.

[0506] In some embodiments of this application, the second information is obtained by the terminal through a monitoring model; the aforementioned first performance index can be used to indicate the similarity or difference between the output information of the monitoring model and the input information of the second model.

[0507] For example, when the output information of the monitoring model is the CSI recovered from the CSI feedback in the output information of the second model, the terminal can compare the CSI output by the monitoring model with the CSI output by the second model to obtain the first performance index.

[0508] In some embodiments of this application, the first performance metric may indicate the similarity or difference between the CSI recovered by the monitoring model and the CSI measured or predicted by the terminal (i.e., the CSI input to the second model).

[0509] In some embodiments of this application, when the output information of the monitoring model is the CSI recovered by the monitoring model from the CSI feedback of the output information of the second model, the first performance indicator may include at least one of the following: the SGCS between the output information of the monitoring model and the input information of the second model, the NMSE between the output information of the monitoring model and the input information of the second model, and the MSE between the output information of the monitoring model and the input information of the second model. Of course, the first performance indicator may also include any other indicator that can measure or evaluate the performance status of the model.

[0510] It is understood that the first performance metric can be used to represent or measure at least one of SGCS, NMSE, and MSE between the CSI recovered by the third model and the CSI input to the second model.

[0511] For further descriptions of the first performance metric, please refer to the relevant descriptions in the above embodiments.

[0512] In some embodiments of this application, when the first monitoring result characterizes the performance state of the first model as a first state, it indicates that the model monitoring on the terminal side believes that the model currently used for CSI feedback, i.e., the first model, is not performing well.

[0513] In some embodiments of this application, the second information includes a first performance indicator, which can be reported in a non-quantized manner or in a quantized manner using Q bits, where Q can be a positive integer. Thus, since the first performance indicator can be reported in either a non-quantized or quantized manner, the flexibility of the reporting method for the first performance indicator can be improved.

[0514] In some embodiments of this application, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0515] In some embodiments of this application, Q can be any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0516] In some embodiments of this application, when Q is active in the candidate parameter set configured by RRC signaling: the network-side device can configure a candidate parameter set through RRC signaling, and the MAC-CE selects one parameter configuration from the candidate parameter set for activation; or, the MAC-CE selects some parameter configurations from the candidate parameter set, and the DCI signaling dynamically triggers (e.g., activates) one of the parameter configurations selected by the MAC-CE.

[0517] In some embodiments of this application, the first performance index may include K performance indices, where K can be a positive integer;

[0518] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0519] If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0520] If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0521] In some embodiments of this application, the first performance index may include K performance indices, where K can be a positive integer;

[0522] If K equals 1, and one performance indicator does not satisfy a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the second state; or,

[0523] If K is greater than 1, and the mean of the K performance indicators does not satisfy a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the second state; or,

[0524] If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is less than M, then the first monitoring result is used to indicate that the performance state of the first model is the second state, where M is a positive integer less than or equal to K.

[0525] In some embodiments of this application, the performance indicators and preset relationships may differ.

[0526] For example, if the performance metric is GCS, the preset relationship is: the performance metric is greater than or equal to the first threshold.

[0527] For example, if the performance metric is NMSE or MSE, the default relationship is: the performance metric is less than or equal to the first threshold.

[0528] Thus, when the first performance indicator includes K performance indicators, the first monitoring result can be determined in different ways based on these K performance indicators, thereby increasing the flexibility of the method for determining the first monitoring result.

[0529] In some embodiments of this application, K, M, or the first threshold can be predefined by the protocol as a preset value; or, K, M, or the first threshold can be configured by the network-side device through downlink signaling.

[0530] In some embodiments of this application, K is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration. This improves the flexibility of how the terminal obtains or determines K.

[0531] In some embodiments of this application, M is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration. This improves the flexibility of how the terminal obtains or determines M.

[0532] In some embodiments of this application, the first threshold is any one of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration. This improves the flexibility of how the terminal obtains or determines the first threshold.

[0533] For example, K, M, and the first threshold can all be configured in the CSI reporting configuration.

[0534] For example, K, M, and the first threshold are all indicated by indicator fields added in the CSI reporting configuration.

[0535] For example, K and M are configured for CSI reporting, and the first threshold is predefined by the protocol.

[0536] For example, K and M are configured for CSI reporting, and the first threshold is activated from the set of candidate parameters for RRC signaling configuration.

[0537] For example, K and M are indicated by the indication fields added in the CSI reporting configuration, and the first threshold is activated in the candidate parameter set of the RRC signaling configuration.

[0538] For example, K, M, and the first threshold are all active parameters in the candidate parameter set of the RRC signaling configuration.

[0539] It is understood that the above example is only used to illustrate K, M and the first threshold. In actual implementation, K, M and the first threshold can also be determined or obtained by any other possible means.

[0540] In some embodiments of this application, when K, M and the first threshold are indicated by indication fields added in the CSI reporting configuration, K, M and the first threshold are indicated by different indication fields added in the CSI reporting configuration, such as indication fields used to indicate K, M and the first threshold in the CSI reporting configuration.

[0541] In some embodiments of this application, when K, M, and the first threshold are all active in the candidate parameter set configured by RRC signaling: the network-side device can configure a candidate parameter set through RRC signaling, and the Media Access Control-Control Element (MAC-CE) selects one parameter configuration from the candidate parameter set for activation; or, the network-side device can configure a candidate parameter set through RRC signaling, and the MAC-CE selects a portion of the parameter configurations from the candidate parameter set, and the DCI signaling dynamically triggers (e.g., activates) one of the parameter configurations selected by the MAC-CE.

[0542] In some embodiments of this application, when K, M, and the first threshold are all active in the candidate parameter set of RRC signaling configuration:

[0543] This set of candidate parameters can include multiple parameter configurations, each configuration containing: a K, an M, and a first threshold; or...

[0544] The candidate parameter set can include three subsets, namely subset 1 to subset 3. Subset 1 is configured with K, subset 2 is configured with M, and subset 3 is configured with a first threshold. This increases the flexibility in selecting or activating M, K, and the first threshold.

[0545] In some embodiments of this application, K, M and the first threshold may be configured through different RRC signaling or through one RRC signaling, and this application does not limit them.

[0546] It is understandable that if the first monitoring result is determined based on whether the average of K performance indicators and the first threshold satisfy a preset relationship, the network-side device may not need to configure M.

[0547] In some embodiments of this application, M, K and the first threshold can be configured simultaneously with Q, or they can be configured separately.

[0548] Thus, since the second information may include the first performance index or the first monitoring result, and the first monitoring result can directly indicate the performance status of the first model, and the first performance index can indirectly indicate the performance status of the first model, the flexibility of the second information in indicating the performance status of the model can be improved.

[0549] Furthermore, after receiving the second information, which includes the first performance indicator, the network-side device can accurately determine the terminal's performance monitoring results for the first model based on the second information.

[0550] Step 802: When the performance status indicated by the second information is in the first state, the network-side device receives the first information sent by the terminal.

[0551] The first piece of information can be used by network-side devices to determine the performance status of the first model.

[0552] In some embodiments of this application, after receiving the second information, if the performance state indicated by the second information is the second state, the network-side device does not need to perform any additional operations. If the performance state indicated by the second information is the first state, the network-side device can receive the first information sent by the terminal. It can be understood that the sending of the first information is determined by the performance state of the first model determined by the terminal, that is, whether the first information is sent is determined by the performance state of the first model determined by the terminal.

[0553] In some embodiments of this application, the first information may include at least the input information of the second model. For a description of the input information of the second model, please refer to the relevant descriptions in the above embodiments.

[0554] In some embodiments of this application, the format of the first information can be any of the following: floating-point number (float) 32, float 16, eTypeII, doppler eTypeII, etc., any format with high data precision.

[0555] In some embodiments of this application, after receiving the first information, the network-side device can determine a second performance index based on the first information and the output information of the third model. The second performance index is used to characterize the similarity or difference between the CSI and the CSI recovered by the third model.

[0556] In some embodiments of this application, the second performance metric may indicate the similarity or difference between the output information of the third model and the input information of the second model. For example, the second performance metric may indicate the similarity or difference between the CSI recovered by the third model and the CSI measured or predicted by the terminal.

[0557] In some embodiments of this application, the second performance metric may include at least one of the following: the SGCS between the output information of the third model and the input information of the second model, the NMSE between the output information of the third model and the input information of the second model, the MSE between the output information of the third model and the input information of the second model, etc. Of course, the second performance metric may also include any other metric that can measure or evaluate the performance state of the model.

[0558] In some embodiments of this application, the method for determining the second performance metric may differ depending on the CSI feedback architecture.

[0559] Specifically, in the first or second CSI feedback architecture described above, the second performance metric can be based on the CSI at a single moment, such as the calculation result of the CSI measured by the terminal at a single moment. For example, the network-side device can compare the measured CSI at a single moment with the CSI recovered by the CSI feedback of the third model to obtain the second performance metric.

[0560] In the third CSI feedback architecture described above, the second performance metric can be the average of the calculated results of predicted CSIs at multiple times. For example, for the CSIs predicted by the terminal at multiple times, hereinafter referred to as multiple predicted CSIs, the network-side device can determine a performance metric based on each predicted CSI and the CSI feedback of each predicted CSI, and then calculate the average of the multiple performance metrics to obtain the second performance metric.

[0561] It can be understood that "determining a performance metric based on each predicted CSI and the CSI feedback of each predicted CSI" may include: using the predicted CSI as label information or standard information, comparing the predicted CSI with the CSI recovered by the CSI feedback of the predicted CSI by the third model to obtain a performance metric.

[0562] In some embodiments of this application, the type and calculation method of the second performance index are consistent with the type and calculation method of the first index, that is, the second performance index is calculated using the same calculation method and sample size as the first performance index.

[0563] For example, if the first performance metric is SGCS, then the second performance metric is also SGCS.

[0564] For example, if the first performance metric is MSE, then the second performance metric is also MSE.

[0565] For example, if the first performance metric is NMSE, then the second performance metric is also NMSE.

[0566] For example, if the first performance indicator is directly output by the monitoring model, then the second performance indicator is directly output by the third model; if the first performance indicator is calculated by the terminal based on the output information of the monitoring model and the input information of the first model, then the second performance indicator is calculated by the network-side device based on the output information of the third model and the first information. The first information includes at least the input information of the first model. Of course, in actual implementation, when the first performance indicator is directly output by the monitoring model, the second performance indicator can also be calculated by the network-side device based on the output information of the third model and the first information, depending on the model function or model structure of the third model.

[0567] In some embodiments of this application, the time corresponding to the sample used to calculate the second performance index should be aligned one-to-one with the time corresponding to the sample used to calculate the first performance index.

[0568] For example, if the first performance indicator is calculated by the terminal based on the measured CSI at time i and the CSI feedback at time i recovered from the monitoring model, then: the first information includes the measured CSI at time i, and the second performance indicator is calculated based on the first information and the CSI at time i recovered from the monitoring model. It should be noted that the "CSI feedback at time i" can be generated by the second model, such as by encoding the measured CSI at time i to obtain the CSI feedback at time i.

[0569] For example, if the first performance index is calculated by the terminal based on the predicted CSI at multiple times and the predicted CSI at those multiple times recovered by the monitoring model, then: the first information includes the predicted CSI at those multiple times, and the second performance index is calculated based on the first information and the CSI recovered by the monitoring model from the predicted CSI at those multiple times.

[0570] In some embodiments of this application, the second performance index includes the same number of performance indices as the first performance index.

[0571] In some embodiments of this application, the network-side device may determine the second monitoring result based on the relationship between the second performance indicator and the first threshold.

[0572] In some embodiments of this application, the second performance index may include K performance indices, where K can be a positive integer;

[0573] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the second monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0574] If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the second monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0575] If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the second monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0576] In some embodiments of this application, the first performance index may include K performance indices, where K can be a positive integer;

[0577] If K equals 1, and one performance indicator does not satisfy a preset relationship with the first threshold, then the second monitoring result is used to indicate that the performance state of the first model is the second state; or,

[0578] If K is greater than 1, and the mean of the K performance indicators does not satisfy a preset relationship with the first threshold, then the second monitoring result is used to indicate that the performance state of the first model is the second state; or,

[0579] If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy the preset relationship with the first threshold is less than M, then the second monitoring result is used to indicate that the performance state of the first model is the second state, where M is a positive integer less than or equal to K.

[0580] In some embodiments of this application, the performance indicators and preset relationships may differ.

[0581] For example, if the performance metric is GCS, the preset relationship is: the performance metric is greater than or equal to the first threshold.

[0582] For example, if the performance metric is NMSE or MSE, the default relationship is: the performance metric is less than or equal to the first threshold.

[0583] In some embodiments of this application, the transmission resource for the second information is selected by the terminal from candidate resources for the second information; similarly, the transmission resource for the first information is selected by the terminal from candidate resources for the first information. It should be noted that the terminal can independently select the transmission resources for the second and first information, or it can select the transmission resources based on indications from the network-side device.

[0584] In some embodiments of this application, the candidate transmission resource for the second information is the first resource, and the candidate transmission resource for the first information is the second resource; the first resource or the second resource is any one of the following: predefined by the protocol, or configured by the network-side device.

[0585] In some embodiments of this application, the second resource can be used to report the first information during the model monitoring process. Correspondingly, the first resource can be used to report CSI feedback and the second information during the inference phase.

[0586] In some embodiments of this application, the first resource may include at least one of frequency domain resources and time domain resources.

[0587] In some embodiments of this application, the second resource may include at least one of frequency domain resources and time domain resources.

[0588] In some embodiments of this application, frequency domain resources can be: subcarriers, carriers, RBs, REs, and subbands.

[0589] In some embodiments of this application, the time-domain resource can be any of the following: frame, subframe, symbol, time slot, symbol.

[0590] In some embodiments of this application, the first resource and the second resource are different. For example, the first resource and the second resource are different symbols or different RBs.

[0591] In some embodiments of this application, when the reporting of the first information is not required, the second resource can be left idle or can be used for the reporting of other information.

[0592] In some embodiments of this application, the candidate transmission resource for the second information configured by the network-side device is the first resource, and the candidate transmission resource for the first information configured by the network-side device is the second resource; the first resource or the second resource is configured through any of the following: CSI reporting configuration, RRC signaling configuration, or pre-configuration.

[0593] In some embodiments of this application, the method of reporting the configuration of the first or second resource via CSI may include any one of the following methods (iv), (v), and (vi):

[0594] Method (4): Configure the first and second resources in a single CSI reporting configuration;

[0595] Method (5): Configure the first resource in one type 1 CSI reporting configuration, and configure the first and second resources in one type 2 CSI reporting configuration;

[0596] Method (6): Configure the first resource in a first-class CSI reporting configuration and configure the second resource in a second-class CSI reporting configuration.

[0597] In some embodiments of this application, the first type of CSI reporting configuration differs from the second type of CSI reporting configuration.

[0598] For example, the first type of CSI reporting configuration is a periodic CSI reporting configuration, and the second type of CSI reporting configuration is a semi-continuous CSI reporting configuration.

[0599] For example, the first type of CSI reporting configuration is a periodic CSI reporting configuration, and the second type of CSI reporting configuration is a non-periodic CSI reporting configuration.

[0600] For example, the first type of CSI reporting configuration is a semi-persistent CSI reporting configuration, and the second type of CSI reporting configuration is a periodic CSI reporting configuration.

[0601] For example, the first type of CSI reporting configuration is a semi-persistent CSI reporting configuration, and the second type of CSI reporting configuration is a non-periodic CSI reporting configuration.

[0602] For example, the first type of CSI reporting configuration is a non-periodic CSI reporting configuration, while the second type of CSI reporting configuration is a periodic CSI reporting configuration.

[0603] For example, the first type of CSI reporting configuration is a non-periodic CSI reporting configuration, and the second type of CSI reporting configuration is a semi-persistent CSI reporting configuration.

[0604] In some embodiments of this application, in the above-described method (iv), the network-side device can configure the first resource and the second resource separately in the same CSI configuration report. The number of CSI configuration reports with both the first and second resources configured is not limited.

[0605] For example, network-side devices can configure a first resource and a second resource in each CSI reporting configuration.

[0606] For example, network-side devices can configure a first resource and a second resource in at least one CSI reporting configuration sent within a preset time period.

[0607] In some embodiments of this application, in the above-described method (v), the network-side device can configure a first resource in each first type of CSI reporting configuration, and configure a first resource and a second resource in each second type of CSI reporting configuration. The first resource configured in the first type of CSI reporting configuration and the first resource configured in the second type of CSI reporting configuration can be the same or different.

[0608] In some embodiments of this application, in the above-described method (v), the network-side device can configure a first resource in each first type of CSI reporting configuration within a preset time period, and configure a first resource and a second resource in each second type of CSI reporting configuration within the preset time period. The first resource configured in the first type of CSI reporting configuration and the first resource configured in the second type of CSI reporting configuration can be the same or different.

[0609] In some embodiments of this application, in the above-described method (vi), the network-side device can configure a first resource in each first type of CSI reporting configuration, and configure a first resource and a second resource in each second type of CSI reporting configuration. The first resource configured in the first type of CSI reporting configuration and the first resource configured in the second type of CSI reporting configuration can be the same or different.

[0610] In some embodiments of this application, in method (vi) described above, the network-side device can configure a first resource in each first type of CSI reporting configuration within a preset time period, and configure resources in each second type of CSI reporting configuration within the preset time period. The first resource configured in the first type of CSI reporting configuration and the first resource configured in the second type of CSI reporting configuration can be the same or different.

[0611] It should be noted that the above-mentioned preset time period can be the supervision or monitoring time period of the first model.

[0612] It is understood that in the above methods (v) and (vi), the network-side device can, based on the reporting requirements of the first information, indicate the selection range of the second information or the transmission resources of the first information through additional signaling, such as indicating to select from the candidate transmission resources configured in the first type of CSI reporting configuration, and / or indicating to select from the candidate transmission resources configured in the second type of CSI reporting configuration.

[0613] In some embodiments of this application, the methods of configuring the first and second resources through CSI reporting configuration are different, and the transmission resources used by the terminal to send the second information and the first information may be different.

[0614] For example, in the above method (iv), the terminal can use the first resource configured by any CSI reporting configuration to send the second information, and use the second resource configured by any SCI reporting configuration to send the first information.

[0615] For example, in the above method (v), the first type of CSI reporting configuration includes A CSI reporting configurations, each of which is configured with a first resource. The configuration identifiers of the A CSI reporting configurations are as follows: The second type of CSI reporting configuration includes B CSI reporting configurations, each of which is configured with a first resource and a second resource. For example, each of the second type of CSI reporting configurations is configured with one first resource and one second resource. The configuration identifiers of the B CSI reporting configurations are as follows: There are a total of B, where A and B are both positive integers.

[0616] When the reporting of the first information is not required, the network device can pre-configure via RRC signaling, or configure via RRC signaling, or activate via MAC-CE / DCI, to instruct the terminal to select the first resource configured in one of the A CSI reporting configurations, so that the terminal can use the first resource configured in that CSI reporting configuration to send the second information. When the reporting of the first information is required, the network device can reconfigure via RRC signaling, or activate via MAC-CE / DCI, to instruct the terminal to activate the resource configured in one of the B CSI reporting configurations, so that the terminal can use the first resource configured in a second type of CSI reporting configuration to send the second information, and use the second resource configured in that second type of CSI reporting configuration to send the first information.

[0617] For example, in the above method (vi), the first type of CSI reporting configuration includes C CSI reporting configurations, each of which is configured with a first resource. The configuration identifiers of the C CSI reporting configurations are as follows: The second type of CSI reporting configuration includes D CSI reporting configurations, each of which is configured with a second resource. For example, each of the second type of CSI reporting configurations is configured with one first resource and one second resource. The configuration identifiers of the D CSI reporting configurations are as follows: There are a total of D, where C and D are both positive integers.

[0618] When the reporting of the first information is not required, the network-side device can instruct the terminal to select the first resource configured in one of the C CSI reporting configurations as the transmission resource for the second information through RRC signaling pre-configuration, RRC signaling configuration, or MAC-CE / DCI activation. When the reporting of the first information is required, the network-side device can instruct the terminal to additionally select the second resource configured in one of the D CSI reporting configurations as the transmission resource for the first information through RRC signaling pre-configuration or MAC-CE / DCI activation.

[0619] It is understandable that in method (vi), the terminal needs to use a first resource configured in a first type of CSI reporting configuration to send the second information, and use a second resource configured in a second type of CSI reporting configuration to send the first information.

[0620] Thus, since configuring the first or second resource through CSI reporting configuration includes: configuring the first and second resources in one CSI reporting configuration; or configuring the first resource in one first-type CSI reporting configuration and configuring the first and second resources in one second-type CSI reporting configuration; or configuring the first resource in one first-type CSI reporting configuration and configuring the second resource in one second-type CSI reporting configuration, the flexibility of configuring the first or second resource through CSI reporting configuration can be improved.

[0621] In some embodiments of this application, when the first resource or the second resource is configured by the CSI reporting configuration, the network-side device may by default associate the first resource and the second resource configured in the same CSI reporting configuration with each other and use them to report the second information and the first information, respectively.

[0622] In some embodiments of this application, the first or second resource is pre-configured by the network-side device when the terminal activates the CSI feedback function based on artificial intelligence (AI) or machine learning (ML).

[0623] In some embodiments of this application, the network-side device can determine that the terminal has activated the AI- or ML-based CSI feedback function based on indication information from the terminal side. This indication information can indicate that the terminal has activated the AI- or ML-based CSI feedback function. Alternatively, the network-side device can determine that the terminal activates the AI / ML-based CSI feedback function after a preset time following the transmission time of the first activation indication information. The first activation indication information is used to indicate that the terminal activates the AI- or ML-based CSI feedback function.

[0624] In some embodiments of this application, when the network-side device activates the AI / ML-based CSI feedback function or activates the first model, the network-side device can pre-configure the first and second resources through RRC signaling.

[0625] In some embodiments of this application, when the information used to determine the performance status of the first model, such as the transmission of the first information, is based on the model monitoring results sent on the terminal side, i.e., method 1 in the above terminal side method embodiments, the network side device can determine the content reported by the second resource by monitoring whether the second information exists in the first resource, or by judging whether the received second information meets the conditions of the first information.

[0626] For example, if the terminal sends the second information after determining that the performance state of the first model is the first state, and the network-side device detects that the second information exists in the first resource, that is, the network-side device has received the second information, then the network-side device can determine that the content received through the second resource is the first information; otherwise, the network-side device can determine that the content received through the second resource is other information (for scenarios where the second resource can be used to send other information when the first information does not need to be reported).

[0627] For example, if the terminal sends a second message after determining the performance status of the first model each time, and the performance status indicated by the second message received by the network-side device is the first status, then the received second message is determined to meet the reporting conditions of the first message, thereby determining that the content received through the second resource is the first message; otherwise, if the performance status indicated by the second message received by the network-side device is the second status (such as normal performance status), then the received second message is determined to not meet the reporting conditions of the first message, thereby determining that the content received through the second resource is the first message.

[0628] In some embodiments of this application, when the information used to determine the performance status of the first model, such as the transmission of the first information being an indication message sent by the network-side device (i.e., method 2 in the terminal-side method embodiments described above), is sent to the terminal, the network-side device can monitor the first information carried on the second resource after sending the first indication message. The first indication message can be used to indicate the reporting of information used to determine the performance status of the first model. For a description of the first indication message, please refer to the relevant description in the following embodiments.

[0629] In the model monitoring method provided in this application embodiment, since the network-side device can receive the first information sent by the terminal after receiving the second information, and when the second information indicates that the performance state of the first model determined by the terminal is in the first state, in order to re-determine the performance state of the first model, the monitoring of model performance by both the terminal side and the network side can be realized. Furthermore, since whether the first information is received is determined by the performance state of the first model determined by the terminal, and not by listening to and receiving the first information whenever model monitoring is performed, air interface overhead can be saved.

[0630] In some embodiments of this application, the model monitoring method provided in this application may further include the following step 803.

[0631] Step 803: When the performance status indicated by the second information is in the first state, the network-side device sends the first indication information to the terminal.

[0632] The first indication information can be used to indicate the information reported to determine the performance status of the first model.

[0633] In some embodiments of this application, when the terminal determines that the performance status of the first model is abnormal, the network-side device can send a first indication message to the terminal. Thus, since the network-side device only sends the first indication message to the terminal when the terminal determines that the performance status of the first model is abnormal, air interface overhead can be saved.

[0634] In some embodiments of this application, the above embodiments illustrate the example of the network-side device sending a first indication message to the terminal when the performance status indicated by the second information is abnormal. In actual implementation, the network-side device may also periodically send the first indication message to the terminal (i.e., the sending of the first indication message is unrelated to the performance status indicated by the second information or whether the second information has been received); or the network-side device may send the first indication message to the terminal based on the network-side model monitoring time, which is not limited in this application.

[0635] In some embodiments of this application, when the terminal determines that the performance status of the first model is abnormal, the network-side device can send a first indication message to the terminal. Thus, since the terminal determines that the performance status of the first model is normal, the network-side device does not need to send the first indication message to the terminal, thereby saving air interface overhead.

[0636] In some embodiments of this application, step 803 may be performed after step 801.

[0637] Thus, since the network-side device can send the first indication information to the terminal when the performance state indicated by the second information is the first state, it is convenient for the network-side device to accurately listen to or receive the first information, thereby improving the accuracy of the first information reporting.

[0638] In some embodiments of this application, the model monitoring method provided in this application may further include the following step 804.

[0639] Step 804: When the number of times the performance state of the first model determined by the network-side device differs from the performance state of the first model determined by the terminal reaches a preset number, the network-side device sends a second indication message to the terminal.

[0640] The second indication information can be used to indicate the monitoring model used by the updated terminal.

[0641] In some embodiments of this application, the preset number of times can be agreed upon by the protocol or set by the network-side device.

[0642] In some embodiments of this application, "the performance state of the first model determined by the network-side device is different from the performance state of the first model determined by the terminal" may include: the performance state of the first model determined by the network-side device is the second state, but the performance state indicated by the second information is the first state.

[0643] In some embodiments of this application, if the number of times the performance state of the first model determined by the network-side device differs from the performance state of the first model determined by the terminal reaches a preset number, the network-side device can consider that the monitoring accuracy of the monitoring model used by the terminal side is poor.

[0644] It is understandable that the aforementioned preset number of times can also be referred to as the number of false alarms.

[0645] In some embodiments of this application, the network-side device can determine a second monitoring result based on the relationship between a second performance indicator and a first threshold. The second monitoring result can indicate whether the performance status of the first model determined by the network-side device is abnormal.

[0646] In some embodiments of this application, the second instruction information may instruct the updating of model parameters of the monitoring model used by the terminal, so as to improve the model monitoring performance on the terminal side.

[0647] In some embodiments of this application, step 804 may be performed after step 802.

[0648] Thus, since the network-side device can send a second indication message to the terminal to indicate the updating of the monitoring model used by the terminal when the performance state of the first model determined by the network-side device differs from that of the first model determined by the terminal a preset number of times, the terminal can update the monitoring model it uses in a timely manner, thereby improving the accuracy of monitoring the performance of the first model on the terminal side.

[0649] In some embodiments of this application, the model monitoring method provided in this application may further include the following step 805.

[0650] Step 805: If the network-side device determines that the performance state of the first model is the first state, the network-side device performs the first operation.

[0651] The first operation may include at least one of the following:

[0652] Send a third instruction to the terminal. The third instruction can be used to instruct any of the following: update the model that can be used to generate CSI feedback, switch the model that can be used to generate CSI feedback, or stop using the model that can be used to generate CSI feedback.

[0653] Update the model used to recover CSI;

[0654] Switch the model used to recover CSI;

[0655] Discontinue the use of the model used to recover CSI.

[0656] For example, the first operation may include: the network-side device updating the model used to restore CSI feedback and sending third indication information to the terminal, wherein the third indication information is used to indicate the update of the model used to generate CSI feedback.

[0657] For example, the first operation may include: the network-side device switching the model used to restore CSI feedback and sending a third indication message to the terminal, wherein the third indication message is used to indicate the switching of the model used to generate CSI feedback.

[0658] For example, the first operation may include: the network-side device stopping the use of the model for restoring CSI feedback and sending a third indication message to the terminal, wherein the third indication message is used to indicate the cessation of the use of the model for generating CSI feedback.

[0659] It is understandable that after the network-side device performs the first operation, it can optimize the CSI recovery performance on the network side and optimize the CSI generation performance on the terminal side, thereby enabling timely optimization of the link performance of the CSI feedback link and improving the accuracy of CSI feedback.

[0660] In some embodiments of this application, step 805 may be performed after step 802.

[0661] In this way, when the network-side device determines that the performance status of the first model is abnormal, it can instruct the terminal to optimize the CSI generation method and optimize the network-side device's CSI recovery method through the third indication information. Therefore, the terminal can use an appropriate CSI feedback generation method to provide CSI feedback, and the network-side device can use an appropriate CSI recovery method to recover CSI, thereby ensuring the accuracy of CSI feedback and thus ensuring the link performance of the CSI feedback link.

[0662] In some embodiments of this application, the model monitoring method provided in this application may further include the following step 806.

[0663] Step 806: The network-side device receives the terminal's capability information sent by the terminal. This capability information can be used to indicate at least one of the following:

[0664] Does the terminal possess the primary capability, which is the ability to monitor models or deploy monitoring models?

[0665] Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner?

[0666] In some embodiments of this application, the network-side device can send capability reporting indication information to the terminal, and the terminal can send its capability information to the network-side device based on the received capability reporting indication information.

[0667] In some embodiments of this application, the first capability is the ability to monitor the performance status of the first model on the terminal side, or in other words, the terminal has the ability to deploy a monitoring model.

[0668] In some embodiments of this application, the "configured codebook quantization method" may include any of the following: a configured eTypeII codebook quantization method, or a configured doppler eTypeII codebook quantization method. It is understood that either of these two codebook quantization methods can achieve the reporting of high-precision tag information, such as the first piece of information.

[0669] In some embodiments of this application, whether the terminal has a second capability can also be described as: whether the terminal supports the reporting of first information.

[0670] It is understandable that when a terminal has both first and second capabilities, it means that the terminal can perform model monitoring on the terminal side and can report first information.

[0671] In some embodiments of this application, step 806 may be performed before step 801.

[0672] Thus, since the network-side device can receive capability information reported by the terminal indicating whether the terminal has the first capability / second capability, the network-side device can know whether the terminal can realize the terminal-side monitoring of the first model and whether it can report the first information.

[0673] It should be noted that other explanations and descriptions of this embodiment can be found in the relevant descriptions of the terminal-side embodiments described above, and will not be repeated here.

[0674] Figure 9 is an interactive schematic diagram of the model monitoring method provided in some embodiments of this application. As shown in Figure 9, the model monitoring method may include steps 900 to 904.

[0675] Step 900: The terminal sends the second information to the network-side device.

[0676] Step 901: The network-side device receives the second information.

[0677] The second information can be used to indicate the performance status of the first model determined by the terminal.

[0678] In some embodiments of this application, the terminal can send the second information to the network-side device when the performance status indicated by the second information is abnormal. That is, when the terminal determines that the performance status of the first model is abnormal, it reports the second information to the network-side device, thereby achieving early warning while further saving overhead.

[0679] Of course, the terminal can also report the second information when it determines that the performance status of the first model is either the first state or the second state.

[0680] Step 902: If the performance state of the first model determined by the terminal is the second state, the terminal does not send the first information to the network-side device.

[0681] Step 903: When the performance state of the first model determined by the terminal is the first state, the terminal sends the first information to the network-side device.

[0682] Step 904: When the performance status indicated by the second information is the first status, the network-side device receives the first information.

[0683] The first information is used by the network-side device to determine the performance status of the first model, and the first model is used for CSI feedback.

[0684] In some embodiments of this application, after receiving the first information, the network-side device can obtain or determine the performance status of the first model based on the first information and the output information of the third model.

[0685] It should be noted that steps 902 and 903 above can be performed selectively.

[0686] In some embodiments of this application, referring to FIG9 and FIG10, before step 903 above, the model monitoring method provided by the embodiments of this application may further include the following steps 905 and 906, and step 903 may be implemented by the following step 903A.

[0687] Step 905: When the performance status indicated by the second information is in the first state, the network-side device sends the first indication information to the terminal.

[0688] Step 906: The terminal receives the first instruction information.

[0689] The first indication information is used to indicate the information to be reported to determine the performance status of the first model.

[0690] Step 903A: When the terminal determines that the performance state of the first model is in the first state and receives the first indication information, the terminal sends the first information to the network-side device.

[0691] In some embodiments of this application, referring to FIG9 and FIG11, after step 904 above, the model monitoring method provided in the embodiments of this application may further include the following steps 907 to 909.

[0692] Step 907: When the number of times the performance state of the first model determined by the network-side device differs from the performance state of the first model determined by the terminal reaches a preset number, the network-side device sends a second indication message to the terminal.

[0693] The second indication information is used to indicate the monitoring model used by the updated terminal.

[0694] Step 908: The terminal receives the second instruction information.

[0695] Step 909: The terminal updates the monitoring model used by the terminal based on the second indication information.

[0696] The second indication information is used to indicate the monitoring model used by the updated terminal.

[0697] In some embodiments of this application, referring to FIG9 and FIG12, after step 904 above, the model monitoring method provided by the embodiments of this application may further include the following steps 910 to 912.

[0698] Step 910: When the network-side device determines that the performance state of the first model is the first state, the network-side device performs a first operation, which includes at least one of the following: sending a third indication message to the terminal, updating the model used to generate CSI feedback, switching the model used to generate CSI feedback, and stopping the use of the model used to generate CSI feedback.

[0699] Step 911: The terminal receives the third instruction information.

[0700] Step 912: The terminal executes the operation corresponding to the third instruction information.

[0701] The third instruction information can be used to instruct any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback.

[0702] In the model monitoring method provided in this application embodiment, model monitoring on the network side can be triggered based on the model monitoring results on the user side, such as the second information. Specifically, the label information, i.e. the first information, is reported only when the user side believes that the model is not working well. Therefore, the air interface overhead for reporting label information can be saved, and the monitoring of model performance can be guaranteed.

[0703] It should be noted that the model monitoring method provided in this application can be used not only in the above embodiments, but also in model performance monitoring for future 6G systems or other communication systems, such as CSI feedback based on Joint Source-Channel Coding (JSCC) and feedback of CSI joint prediction and compression based on JSCC.

[0704] It should be noted that the explanation of this embodiment can be found in the relevant descriptions of the corresponding embodiments of the terminal device and network side device described above, and will not be repeated here.

[0705] Accordingly, this application provides a model monitoring device, which can be any model monitoring device in the communication process, such as a terminal or network-side device. Based on the above method example, the model monitoring device can be divided into functional modules. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware or as software functional modules. It should be noted that the module division in this embodiment is illustrative, and the actual division can be based on the specific circumstances.

[0706] Figure 13 illustrates a possible structural diagram of the model monitoring device involved in the above embodiments, where each functional module is divided according to its corresponding function. As shown in Figure 13, the model monitoring device 1300 may include a sending module 1301. The sending module 1301 is used to send first information to the network-side device when the terminal determines that the performance state of the first model is a first state. The first information is used by the network-side device to determine the performance state of the first model, and the first model is used for feedback of Channel State Information (CSI).

[0707] In some possible implementations, the first state mentioned above includes: an abnormal performance state.

[0708] In some possible implementations, the aforementioned first information is sent by the terminal after determining that the performance state of the first model is a first state and receiving the first indication information, which is used to instruct the reporting of information used to determine the performance state of the first model.

[0709] In some possible implementations, the first model mentioned above includes a second model and a third model, where the second model is used to generate CSI feedback and the third model is used to recover CSI.

[0710] In some possible implementations, the model monitoring device 1300 described above further includes: a determination module;

[0711] The aforementioned determining module is used to determine the second information, which is used to indicate the performance status of the first model determined by the terminal.

[0712] In some possible implementations, the second information mentioned above includes any of the following:

[0713] The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI.

[0714] The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

[0715] In some possible implementations, the aforementioned second information is obtained through monitoring by a monitoring model;

[0716] The input information of the above monitoring model is any one of the following: the output information of the second model, or the CSI feedback in the output information of the second model;

[0717] The output information of the above monitoring model is any one of the following: the CSI recovered from the CSI feedback in the output information of the second model, the first performance index.

[0718] In some possible implementations, the aforementioned second information is obtained through monitoring by a monitoring model;

[0719] The aforementioned first performance metric is used to indicate the similarity or difference between the output information of the monitoring model and the input information of the second model.

[0720] In some possible implementations, the first performance metric mentioned above includes K performance metrics, where K is a positive integer;

[0721] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0722] If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0723] If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0724] In some possible implementations, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the Radio Resource Control (RRC) signaling configuration.

[0725] In some possible implementations, the aforementioned first information includes at least the input information of the second model.

[0726] In some possible implementations, the input information of the second model mentioned above includes any of the following:

[0727] CSI measured at the current moment;

[0728] Predict the CSI at at least one moment;

[0729] The CSI measured at the current time and the CSI feature information output by the second model at the previous time.

[0730] In some possible implementations, the aforementioned sending module 1301 is also used to send second information to the network-side device.

[0731] In some possible implementations, the aforementioned second information is sent by the terminal when the performance state indicated by the second information is the first state.

[0732] In some possible implementations, the second information mentioned above includes a first performance metric, which is reported either in a non-quantized manner or in a quantized manner using Q bits, where Q is a positive integer.

[0733] In some possible implementations, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0734] In some possible implementations, the candidate transmission resource for the second information is the first resource, and the candidate transmission resource for the first information is the second resource;

[0735] The first or second resource is any of the following: predefined by the protocol, configured by the CSI reporting configuration, configured by the RRC signaling, or pre-configured.

[0736] In some possible implementations, the first resource or the second resource mentioned above is pre-configured when the AI- or ML-based CSI feedback function is activated on the terminal.

[0737] Among some possible implementations, the method of reporting the configuration of the first or second resource via CSI includes any of the following:

[0738] A CSI reporting configuration includes a first resource and a second resource.

[0739] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources.

[0740] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

[0741] In some possible implementations, as shown in Figure 14 in conjunction with Figure 13, the above-mentioned model monitoring device 1300 further includes: a receiving module 1302 and a processing module 1303;

[0742] The aforementioned receiving module 1302 is used to receive second indication information sent by the network-side device, the second indication information being used to indicate the monitoring model used by the updating terminal;

[0743] The aforementioned processing module 1303 is used to update the monitoring model used by the terminal based on the second indication information received by the receiving module 1302.

[0744] In some possible implementations, as shown in Figure 14 in conjunction with Figure 13, the above-mentioned model monitoring device 1300 further includes: a receiving module 1302 and a processing module 1303;

[0745] The receiving module 1302 described above is used to receive third indication information sent by the network-side device. The third indication information is used to indicate any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback.

[0746] The aforementioned processing module 1303 is used to perform the operation corresponding to the third instruction information received by the receiving module 1302.

[0747] In some possible implementations, the aforementioned sending module 1301 is further configured to send terminal capability information to the network-side device, the capability information indicating at least one of the following:

[0748] Does the terminal possess the primary capability, which is the ability to monitor models or deploy monitoring models?

[0749] Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner?

[0750] In the model monitoring device provided in this application embodiment, the device can send first information to the network-side device when the terminal determines that the performance state of the first model is a first state. The first information is used by the network-side device to determine the performance state of the first model, and the first model is used for feedback of Channel State Information (CSI). In this solution, since the model monitoring device can send first information to the network-side device when the terminal determines that the performance state of the first model is a first state, instructing the network-side device to re-determine the performance state of the first model, it can achieve monitoring of model performance by both the terminal and the network side. Furthermore, since whether the first information is sent is determined by the performance state of the first model determined by the terminal, and not by feeding back first information to the network-side device every time model monitoring is performed, air interface overhead can be saved.

[0751] The modules of the above-mentioned model monitoring device can also be used to perform other actions executed by the terminal in the above-mentioned method embodiments. All relevant content of each step involved in the above-mentioned method embodiments can be referred to the functional description of the corresponding functional module, and will not be repeated here.

[0752] Figure 15 illustrates another possible structural diagram of the model monitoring device involved in the above embodiments when using integrated units. As shown in Figure 15, the model monitoring device 1500 provided in this application embodiment may include a processing module 1501 and a communication module 1502. The processing module 1501 can be used to control and manage the operation of the wireless model monitoring device. For example, the processing module 1501 can be used to support the model monitoring device in executing steps 704, 706, etc., in the above terminal-side method embodiments, and / or to execute other processes of the terminal-related technologies described herein. The communication module 1502 can be used to support the model monitoring device in communicating with other network entities, such as communicating with other wireless model monitoring devices. In one example, as shown in Figure 15, the model monitoring device may also include a storage module 1503. The storage module 1503 is used to store the program code and data of the model monitoring device, such as CSI reporting configuration.

[0753] The processing module 1501 can be a processor, the communication module 1502 can be a transceiver, a transceiver circuit or a communication interface, etc., and the storage module 1503 can be a memory.

[0754] More details on how the modules included in the above-mentioned model monitoring device achieve the above functions can be found in the descriptions of the previous method embodiments, and will not be repeated here.

[0755] It should be noted that the model monitoring device shown in Figure 15 can be a terminal, which can be used to execute the wireless model monitoring method executed by the terminal in the above method embodiment.

[0756] With each functional module divided according to its corresponding function, Figure 16 shows a possible structural schematic diagram of the model monitoring device involved in the above embodiments. As shown in Figure 16, the model monitoring device 1700 may include a receiving module 1601.

[0757] The aforementioned receiving module 1601 is used to receive second information sent by the terminal. The second information is used to indicate the performance status of the first model determined by the terminal. The first model is used for CSI feedback.

[0758] The aforementioned receiving module 1601 is further configured to receive first information sent by the terminal when the performance state indicated by the second information is the first state, wherein the first information is used by the network-side device to determine the performance state of the first model.

[0759] In some possible implementations, as shown in Figure 17 in conjunction with Figure 16, the above-mentioned model monitoring device 1600 further includes: a transmission module 1602;

[0760] The sending module 1602 is used to send first indication information to the terminal when the performance status indicated by the second information received by the receiving module 1601 is a first status. The first indication information is used to indicate the information to be reported to determine the performance status of the first model.

[0761] In some possible implementations, the first model includes a second model and a third model, where the second model is used to generate CSI feedback and the third model is used to recover CSI.

[0762] In some possible implementations, the second information includes any of the following:

[0763] The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI.

[0764] The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

[0765] In some possible implementations, the first performance metric includes K performance metrics, where K is a positive integer;

[0766] If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0767] If K is greater than 1, and the mean of the above K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or,

[0768] If K is greater than 1, and the number of performance indicators among the above K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

[0769] In some possible implementations, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0770] In some possible implementations, the first information includes at least the input information of the second model.

[0771] In some possible implementations, the input information for the second model includes any of the following:

[0772] CSI measured at the current moment;

[0773] Predict the CSI at at least one moment;

[0774] The CSI measured at the current time and the CSI feature information output by the second model at the previous time.

[0775] In some possible implementations, the second information includes a first performance metric, which is reported either in a non-quantized manner or in a quantized manner using Q bits, where Q is a positive integer.

[0776] In some possible implementations, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added to the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

[0777] In some possible implementations, as shown in Figure 17 in conjunction with Figure 16, the above-mentioned model monitoring device 1600 further includes: a transmission module 1602;

[0778] The aforementioned sending module 1602 is used to send a second indication message to the terminal when the number of times the performance status of the first model determined by the network-side device differs from the performance status of the first model determined by the terminal reaches a preset number. The second indication message is used to indicate that the monitoring model used by the terminal should be updated.

[0779] In some possible implementations, as shown in Figure 18 in conjunction with Figure 16, the above-mentioned model monitoring device 1600 further includes a processing module 1603.

[0780] The aforementioned processing module 1603 is configured to perform a first operation when the network-side device determines that the performance state of the first model is a first state. The first operation includes at least one of the following:

[0781] Send a third instruction to the terminal, the third instruction being used to instruct any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback;

[0782] Update the model used to recover CSI;

[0783] Switch the model used to recover CSI;

[0784] Discontinue the use of the model used to recover CSI.

[0785] In some possible implementations, the candidate transmission resource for the second information configured by the network-side device is the first resource, and the candidate transmission resource for the first information configured by the network-side device is the second resource.

[0786] The first or second resource can be configured through any of the following: CSI reporting configuration, RRC signaling configuration, or pre-configuration.

[0787] In some possible implementations, the first or second resource is pre-configured by the network-side device when the terminal activates the CSI feedback function based on artificial intelligence (AI) or machine learning (ML).

[0788] Among some possible implementations, the method of reporting the configuration of the first or second resource via CSI includes any of the following:

[0789] A CSI reporting configuration includes a first resource and a second resource.

[0790] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources.

[0791] A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

[0792] In some possible implementations, the receiving module described above is further configured to receive terminal capability information sent by the terminal, the capability information being used to indicate at least one of the following:

[0793] Does the terminal possess the primary capability, which is the ability to monitor models or deploy monitoring models?

[0794] Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner?

[0795] In the model monitoring device provided in this application embodiment, the model monitoring device includes a receiving module. This receiving module can be used to receive second information sent by a terminal, the second information indicating the performance status of a first model determined by the terminal, and the first model being used for CSI feedback. The receiving module is also used to receive first information sent by the terminal when the performance status indicated by the second information is a first status, the first information being used by the network-side device to determine the performance status of the first model. In this solution, since the model monitoring device can receive the first information sent by the terminal after receiving the second information, and when the second information indicates that the performance status of the first model determined by the terminal is a first status, to re-determine the performance status of the first model, it can realize the monitoring of model performance by both the terminal side and the network side. Furthermore, since whether the first information is received is determined by the performance status of the first model determined by the terminal, and not by listening to and receiving the first information whenever model monitoring is performed, air interface overhead can be saved.

[0796] Each module of the above-mentioned model monitoring device can also be used to perform other actions performed by the network-side device in the above-mentioned method embodiments. All relevant content of each step involved in the above-mentioned method embodiments can be referred to in the functional description of the corresponding functional module, and will not be repeated here.

[0797] Figure 19 illustrates another possible structural diagram of the model monitoring device involved in the above embodiments when using integrated units. As shown in Figure 19, the model monitoring device 1900 provided in this application embodiment may include a processing module 1901 and a communication module 1902. The processing module 1901 can be used to control and manage the operation of the model monitoring device. For example, the processing module 1901 can be used to support the model monitoring device in executing steps 806 in the above method embodiments, and / or to execute other processes related to network-side device technologies described herein. The communication module 1902 can be used to support communication between the model monitoring device and other network entities, such as communication with other model monitoring devices. In one example, as shown in Figure 19, the model monitoring device may also include a storage module 1903 for storing the program code and data of the model monitoring device, such as storing second information and first information.

[0798] The processing module 1901 can be a processor, the communication module 1902 can be a transceiver, a transceiver circuit or a communication interface, etc., and the storage module 1903 can be a memory.

[0799] More details on how the modules included in the above-mentioned model monitoring device achieve the above functions can be found in the descriptions of the previous method embodiments, and will not be repeated here.

[0800] It should be noted that the model monitoring device 1900 shown in Figure 19 can be a network-side device, which can be used to execute the model monitoring method executed by the network-side device in the above method embodiment.

[0801] It should be noted that the descriptions of the above model monitoring device embodiments are similar to those of the above method embodiments, and have similar beneficial effects. In other words, for technical details not disclosed in the device embodiments of this application, please refer to the descriptions in the preceding method embodiments for understanding; they will not be repeated here.

[0802] As shown in Figure 20, this application embodiment also provides a communication device 2000, including a processor 2001 and a memory 2002. The memory 2002 stores a program or instructions that can run on the processor 2001. For example, when the communication device 2000 is a terminal device, the program or instructions executed by the processor 2001 implement the various steps of the above-described terminal-side method embodiment and achieve the same technical effect. When the communication device 2000 is a network-side device, the program or instructions executed by the processor 2001 implement the various steps of the above-described network-side device method embodiment and achieve the same technical effect. To avoid repetition, this will not be described again here.

[0803] It is understood that the implementation process of each implementation method mentioned in this embodiment can refer to the relevant description of the method embodiment and achieve the same or corresponding technical effect. To avoid repetition, it will not be described again here.

[0804] This application also provides a communication system, including the network-side device and terminal device described in the above embodiments. The network-side device can be network-side device 12 as shown in FIG1, and the terminal device can be terminal device 11 as shown in FIG1.

[0805] It is understood that the implementation process of each implementation method mentioned in this embodiment can refer to the relevant description of the method embodiment and achieve the same or corresponding technical effect. To avoid repetition, it will not be described again here.

[0806] This application also provides a readable storage medium storing a program or instructions. When executed by a processor, the program or instructions can implement the various processes of the above-described model monitoring method embodiments and achieve the same technical effects. To avoid repetition, further details are omitted here. The readable storage medium may include a computer-readable storage medium, such as a computer read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk.

[0807] This application also provides a chip that may include a processor and a communication interface. The communication interface is coupled to the processor. The processor can be used to run programs or instructions to implement the various processes of the above-described model monitoring method embodiments and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0808] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0809] This application also provides a computer program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the above-described model monitoring method embodiments, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0810] It should be noted that the scope of the methods and apparatus in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions in a substantially simultaneous manner or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. In addition, features described with reference to certain examples may be combined in other examples.

[0811] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When these computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center integrating one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital video discs (DVDs)), or semiconductor media (e.g., solid-state drives (SSDs)).

[0812] Through the above description of the embodiments, those skilled in the art will clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0813] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0814] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0815] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0816] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as flash memory, portable hard disk, read-only memory, random access memory, magnetic disk, or optical disk.

[0817] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A model monitoring method characterized by, The method includes: When the terminal determines that the performance state of the first model is in the first state, it sends first information to the network-side device. The first information is used by the network-side device to determine the performance state of the first model, and the first model is used for feedback of channel state information (CSI).

2. The method of claim 1, wherein, The first state includes: abnormal performance status.

3. The method of claim 2, wherein, The first information is sent by the terminal when it determines that the performance state of the first model is the first state and receives the first indication information. The first indication information is used to instruct the reporting of information used to determine the performance state of the first model.

4. The method according to any one of claims 1 to 3, characterized in that, The first model includes a second model and a third model, wherein the second model is used to generate CSI feedback and the third model is used to recover CSI.

5. The method of claim 4, wherein, The method further includes: The terminal determines second information, which is used to indicate the performance status of the first model as determined by the terminal.

6. The method of claim 5, wherein, The second information includes any one of the following: The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI. The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

7. The method of claim 6, wherein, The second piece of information was obtained through monitoring using a monitoring model; The input information of the monitoring model is any one of the following: the output information of the second model, or the CSI feedback in the output information of the second model; The output information of the monitoring model is any one of the following: the CSI recovered from the CSI feedback in the output information of the second model, or the first performance index.

8. The method according to claim 6 or 7, characterized in that, The second piece of information was obtained through monitoring using a monitoring model; The first performance metric is used to indicate the similarity or difference between the output information of the monitoring model and the input information of the second model.

9. The method of claim 6, wherein, The first performance metric includes K performance metrics, where K is a positive integer; If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or, If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state. or, If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

10. The method of claim 9, wherein, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the Radio Resource Control (RRC) signaling configuration.

11. The method of claim 4, wherein, The first information includes at least the input information of the second model.

12. The method of claim 11, wherein, The input information for the second model includes any one of the following: CSI measured at the current moment; Predict the CSI at at least one time point; The CSI measured at the current moment and the CSI feature information output by the second model at the previous moment.

13. The method of claim 5, wherein, The method further includes: The terminal sends the second information to the network-side device.

14. The method according to claim 13, characterized in that, The second information is sent by the terminal when the performance state indicated by the second information is the first state.

15. The method according to claim 13, characterized in that, The second information includes a first performance indicator, which is reported either in a non-quantized manner or in a Q-bit quantized manner, where Q is a positive integer.

16. The method according to claim 15, characterized in that, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

17. The method according to claim 13, characterized in that, The candidate transmission resource for the second information is the first resource, and the candidate transmission resource for the first information is the second resource; The first resource or the second resource is any one of the following: predefined by the protocol, configured by the CSI reporting configuration, configured by the RRC signaling, or pre-configured.

18. The method according to claim 17, characterized in that, The first resource or the second resource is pre-configured when the terminal activates the CSI feedback function based on artificial intelligence (AI) or machine learning (ML).

19. The method according to claim 17 or 18, characterized in that, The method of reporting the configuration of the first or second resource via CSI includes any of the following: A CSI reporting configuration includes a first resource and a second resource. A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources. A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

20. The method according to any one of claims 1 to 19, characterized in that, The method further includes: The terminal receives a second indication information sent by the network-side device, the second indication information being used to indicate an update to the monitoring model used by the terminal; The terminal updates the monitoring model it uses based on the second indication information.

21. The method according to any one of claims 1 to 19, characterized in that, The method further includes: The terminal receives a third indication message sent by the network-side device. The third indication message is used to indicate any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback. The terminal performs the operation corresponding to the third instruction information.

22. The method according to any one of claims 1 to 21, characterized in that, The method further includes: The terminal sends its capability information to the network-side device, the capability information indicating at least one of the following: Whether the terminal has a first capability, wherein the first capability is a model monitoring capability or the ability to deploy a monitoring model; Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner? 23. A model monitoring method, characterized in that, The method includes: The network-side device receives second information sent by the terminal, the second information being used to indicate the performance status of a first model determined by the terminal, the first model being used for feedback of Channel State Information (CSI); When the performance status indicated by the second information is the first status, the network-side device receives the first information sent by the terminal. The first information is used by the network-side device to determine the performance status of the first model.

24. The method according to claim 23, characterized in that, The method further includes: When the performance status indicated by the second information is the first status, the network-side device sends a first indication information to the terminal. The first indication information is used to indicate the reporting of information used to determine the performance status of the first model.

25. The method according to claim 23 or 24, characterized in that, The first model includes a second model and a third model, wherein the second model is used to generate CSI feedback and the third model is used to recover CSI.

26. The method according to claim 25, characterized in that, The second information includes any one of the following: The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI. The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

27. The method according to claim 26, characterized in that, The first performance metric includes K performance metrics, where K is a positive integer; If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or, If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state. or, If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

28. The method according to claim 27, characterized in that, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the Radio Resource Control (RRC) signaling configuration.

29. The method according to claim 25, characterized in that, The first information includes at least the input information of the second model.

30. The method according to claim 29, characterized in that, The input information for the second model includes any one of the following: CSI measured at the current moment; Predict the CSI at at least one time point; The CSI measured at the current moment and the CSI feature information output by the second model at the previous moment.

31. The method according to claim 23, characterized in that, The second information includes a first performance indicator, which is reported either in a non-quantized manner or in a Q-bit quantized manner, where Q is a positive integer.

32. The method according to claim 31, characterized in that, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

33. The method according to claim 23, characterized in that, The method further includes: If the number of times the performance state of the first model determined by the network-side device differs from the performance state of the first model determined by the terminal reaches a preset number, the network-side device sends a second indication message to the terminal, the second indication message being used to indicate that the monitoring model used by the terminal should be updated.

34. The method according to claim 23, characterized in that, The method further includes: When the network-side device determines that the performance state of the first model is the first state, the network-side device performs a first operation, the first operation including at least one of the following: Send a third indication message to the terminal, the third indication message being used to indicate any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback; Update the model used to recover CSI; Switch the model used to recover CSI; Discontinue the use of the model used to recover CSI.

35. The method according to claim 23, characterized in that, The candidate transmission resource for the second information configured by the network-side device is the first resource, and the candidate transmission resource for the first information configured by the network-side device is the second resource; The first resource or the second resource is configured through any of the following: CSI reporting configuration, RRC signaling configuration, or pre-configuration.

36. The method according to claim 35, characterized in that, The first resource or the second resource is pre-configured by the network-side device when the terminal activates the CSI feedback function based on artificial intelligence (AI) or machine learning (ML).

37. The method according to claim 35 or 36, characterized in that, The method of reporting the configuration of the first or second resource via CSI includes any of the following: A CSI reporting configuration includes a first resource and a second resource. A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources. A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

38. The method according to any one of claims 23 to 37, characterized in that, The method further includes: The network-side device receives capability information of the terminal sent by the terminal, the capability information being used to indicate at least one of the following: Whether the terminal has a primary capability, wherein the primary capability is the ability to monitor a model or the ability to deploy a monitoring model; Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner? 39. A model monitoring device, characterized in that, The device includes: a transmitting module; The sending module is used to send first information to the network-side device when the terminal determines that the performance state of the first model is a first state. The first information is used by the network-side device to determine the performance state of the first model, and the first model is used for feedback of channel state information (CSI).

40. The apparatus according to claim 39, characterized in that, The first state includes: abnormal performance status.

41. The apparatus according to claim 40, characterized in that, The first information is sent by the terminal when it determines that the performance state of the first model is the first state and receives the first indication information. The first indication information is used to instruct the reporting of information used to determine the performance state of the first model.

42. The apparatus according to any one of claims 39 to 41, characterized in that, The first model includes a second model and a third model, wherein the second model is used to generate CSI feedback and the third model is used to recover CSI.

43. The apparatus according to claim 42, characterized in that, The device further includes: a determining module; The determining module is used to determine second information, which is used to indicate the performance status of the first model determined by the terminal.

44. The apparatus according to claim 43, characterized in that, The second information includes any one of the following: The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI. The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

45. The method according to claim 44, characterized in that, The second piece of information was obtained through monitoring using a monitoring model; The input information of the monitoring model is any one of the following: the output information of the second model, or the CSI feedback in the output information of the second model; The output information of the monitoring model is any one of the following: the CSI recovered from the CSI feedback in the output information of the second model, or the first performance index.

46. ​​The method according to claim 44 or 45, characterized in that, The second piece of information was obtained through monitoring using a monitoring model; The first performance metric is used to indicate the similarity or difference between the output information of the monitoring model and the input information of the second model.

47. The apparatus according to claim 44, characterized in that, The first performance metric includes K performance metrics, where K is a positive integer; If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or, If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state. or, If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

48. The apparatus according to claim 47, characterized in that, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the Radio Resource Control (RRC) signaling configuration.

49. The apparatus according to claim 42, characterized in that, The first information includes at least the input information of the second model.

50. The apparatus according to claim 49, characterized in that, The input information for the second model includes any one of the following: CSI measured at the current moment; Predict the CSI at at least one time point; The CSI measured at the current moment and the CSI feature information output by the second model at the previous moment.

51. The apparatus according to claim 43, characterized in that, The sending module is also used to send the second information to the network-side device.

52. The apparatus according to claim 51, characterized in that, The second information is sent by the terminal when the performance state indicated by the second information is the first state.

53. The apparatus according to claim 51, characterized in that, The second information includes a first performance indicator, which is reported either in a non-quantized manner or in a Q-bit quantized manner, where Q is a positive integer.

54. The apparatus according to claim 53, characterized in that, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

55. The apparatus according to claim 51, characterized in that, The candidate transmission resource for the second information is the first resource, and the candidate transmission resource for the first information is the second resource; The first resource or the second resource is any one of the following: predefined by the protocol, configured by the CSI reporting configuration, configured by the RRC signaling, or pre-configured.

56. The apparatus according to claim 55, characterized in that, The first resource or the second resource is pre-configured when the terminal activates the CSI feedback function based on artificial intelligence (AI) or machine learning (ML).

57. The apparatus according to claim 55 or 56, characterized in that, The method of reporting the configuration of the first or second resource via CSI includes any of the following: A CSI reporting configuration includes a first resource and a second resource. A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources. A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

58. The apparatus according to any one of claims 39 to 57, characterized in that, The device further includes: a receiving module and a processing module; The receiving module is used to receive second indication information sent by the network-side device, the second indication information being used to indicate the updating of the monitoring model used by the terminal; The processing module is used to update the monitoring model used by the terminal based on the second indication information received by the receiving module.

59. The apparatus according to any one of claims 39 to 57, characterized in that, The device further includes: a receiving module and a processing module; The receiving module is configured to receive third indication information sent by the network-side device, the third indication information being used to indicate any of the following: updating the model used to generate CSI feedback, switching the model used to generate CSI feedback, or stopping the use of the model used to generate CSI feedback; The processing module is used to perform the operation corresponding to the third indication information received by the receiving module.

60. The apparatus according to any one of claims 39 to 59, characterized in that, The sending module is further configured to send the terminal's capability information to the network-side device, the capability information indicating at least one of the following: Whether the terminal has a first capability, wherein the first capability is a model monitoring capability or the ability to deploy a monitoring model; Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner? 61. A model monitoring device, characterized in that, The device includes: a receiving module; The receiving module is used to receive second information sent by the terminal, the second information being used to indicate the performance status of the first model determined by the terminal, the first model being used for feedback of channel state information (CSI); The receiving module is further configured to receive first information sent by the terminal when the performance state indicated by the second information is a first state, wherein the first information is used by the network-side device to determine the performance state of the first model.

62. The apparatus according to claim 61, characterized in that, The device further includes: a transmitting module; The sending module is configured to send first indication information to the terminal when the performance status indicated by the second information received by the receiving module is as described above. The first indication information is used to indicate the reporting of information used to determine the performance status of the first model.

63. The apparatus according to claim 61 or 62, characterized in that, The first model includes a second model and a third model, wherein the second model is used to generate CSI feedback and the third model is used to recover CSI.

64. The apparatus according to claim 63, characterized in that, The second information includes any one of the following: The first performance metric is used to characterize the similarity or difference between the CSI and the recovered CSI. The first monitoring result is determined based on the relationship between the first performance index and the first threshold.

65. The apparatus according to claim 64, characterized in that, The first performance metric includes K performance metrics, where K is a positive integer; If K equals 1, and one performance indicator satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state; or, If K is greater than 1, and the mean of the K performance indicators satisfies a preset relationship with the first threshold, then the first monitoring result is used to indicate that the performance state of the first model is the first state. or, If K is greater than 1, and the number of performance indicators among the K performance indicators that satisfy a preset relationship with the first threshold is greater than or equal to M, then the first monitoring result is used to indicate that the performance state of the first model is the first state, where M is a positive integer less than or equal to K.

66. The apparatus according to claim 65, characterized in that, K, M, or the first threshold is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the Radio Resource Control (RRC) signaling configuration.

67. The apparatus according to claim 63, characterized in that, The first information includes at least the input information of the second model.

68. The apparatus according to claim 67, characterized in that, The input information for the second model includes any one of the following: CSI measured at the current moment; Predict the CSI at at least one time point; The CSI measured at the current moment and the CSI feature information output by the second model at the previous moment.

69. The apparatus according to claim 61, characterized in that, The second information includes a first performance indicator, which is reported either in a non-quantized manner or in a Q-bit quantized manner, where Q is a positive integer.

70. The apparatus according to claim 69, characterized in that, Q is any of the following: predefined by the protocol, configured by the CSI reporting configuration, indicated by an indication field added in the CSI reporting configuration, or activated in the candidate parameter set of the RRC signaling configuration.

71. The apparatus according to claim 61, characterized in that, The device further includes: a transmitting module; The sending module is configured to send a second indication message to the terminal when the number of times the performance status of the first model determined by the network-side device differs from the performance status of the first model determined by the terminal reaches a preset number. The second indication message is used to indicate that the monitoring model used by the terminal should be updated.

72. The apparatus according to claim 61, characterized in that, The device further includes: a processing module; The processing module is configured to perform a first operation when the network-side device determines that the performance state of the first model is the first state, the first operation including at least one of the following: Send a third indication message to the terminal, the third indication message being used to indicate any of the following: update the model used to generate CSI feedback, switch the model used to generate CSI feedback, or stop using the model used to generate CSI feedback; Update the model used to recover CSI; Switch the model used to recover CSI; Discontinue the use of the model used to recover CSI.

73. The apparatus according to claim 61, characterized in that, The candidate transmission resource for the second information configured by the network-side device is the first resource, and the candidate transmission resource for the first information configured by the network-side device is the second resource; The first resource or the second resource is configured through any of the following: CSI reporting configuration, RRC signaling configuration, or pre-configuration.

74. The apparatus according to claim 73, characterized in that, The first resource or the second resource is pre-configured by the network-side device when the terminal activates the CSI feedback function based on artificial intelligence (AI) or machine learning (ML).

75. The apparatus according to claim 73 or 74, characterized in that, The method of reporting the configuration of the first or second resource via CSI includes any of the following: A CSI reporting configuration includes a first resource and a second resource. A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures both the first and second resources. A first-class CSI reporting configuration configures the first resource, and a second-class CSI reporting configuration configures the second resource.

76. The apparatus according to any one of claims 61 to 75, characterized in that, The receiving module is further configured to receive capability information of the terminal sent by the terminal, the capability information being used to indicate at least one of the following: Whether the terminal has a primary capability, wherein the primary capability is the ability to monitor a model or the ability to deploy a monitoring model; Does the terminal have a second capability, which is the ability to report information in a configured codebook quantization manner? 77. A terminal device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the steps of the model monitoring method as described in any one of claims 1 to 22.

78. A network-side device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the steps of the model monitoring method as described in any one of claims 23 to 38.

79. A communication system, characterized in that, It includes the model monitoring device as described in any one of claims 39 to 60 and the model monitoring device as described in any one of claims 61 to 76; or, it includes the terminal device as described in claim 77 and the network-side device as described in claim 78.

80. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the model monitoring method as described in any one of claims 1 to 22, or implement the steps of the model monitoring method as described in any one of claims 23 to 38.

81. A computer program product stored in a storage medium, the computer program product being executed by at least one processor to implement the model monitoring method as claimed in any one of claims 1 to 22, or to implement the steps of the model monitoring method as claimed in any one of claims 23 to 38.

Images