Information processing methods, apparatus, and equipment

Abstract

Information processing methods, apparatus, and devices belonging to the field of communication technology. Among them, the information processing method includes obtaining performance monitoring accuracy according to at least two test sample points; determining monitoring event-related information regarding performance monitoring accuracy, which includes the definition of a monitoring event and the conditions for joining and / or leaving a monitoring event; and controlling the trigger operation of a performance monitoring report according to the performance monitoring accuracy and the monitoring event-related information.

Description

Technical Field

[0001] Cross-Reference to Related Applications The present disclosure claims priority to the Chinese patent application filed with the China National Intellectual Property Administration on October 23, 2023, with application number 202311376082.5 and titled "Information Processing Method, Apparatus and Device", the entire content of which is incorporated into the present disclosure by reference. The present disclosure relates to the field of communication technologies, and in particular, to an information processing method, apparatus and device.

Background Art

[0002] In the related art, a manner of monitoring model performance has been proposed to ensure the application degree of the model, but it does not relate to the use (for example, reporting) of the monitoring result, and thus similarly cannot support accurate implementation of decision operations such as model switching, fallback, updating, activation and / or deactivation.

[0003] In view of the above, the information processing aspect for model monitoring in the related art has a problem that accurate implementation of model decision operations cannot be supported.

[0004] It should be noted that since model monitoring is to monitor the performance of the model and judge the performance of the model according to the monitoring result, "model monitoring" can be understood or replaced as "performance monitoring".

Summary of the Invention

Problems to be Solved by the Invention

[0005] An objective of the present disclosure is to provide an information processing method, apparatus and device for solving the problem that accurate implementation of model decision operations existing in the information processing aspect for model monitoring in the related art cannot be supported.

Means for Solving the Problems

[0006] To solve the above problems, an embodiment of the present disclosure is an information processing method applicable to a first device, The performance monitoring accuracy is obtained based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI restored by a model corresponding to the target CSI, and the performance monitoring accuracy is based on the statistical results of the at least two test sample points. To determine the monitoring event-related information regarding the performance monitoring accuracy, which includes the definition of the monitoring event and the joining and / or leaving conditions of the monitoring event. This includes controlling the trigger operation for performance monitoring reports according to the performance monitoring accuracy and monitoring event-related information, The first device is a network device or terminal, and the present invention provides an information processing method.

[0007] Selectively controlling the trigger operation of performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the aforementioned monitoring event, then the operation of the performance monitoring report is triggered. Alternatively, the monitoring event includes triggering a performance monitoring report operation if the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event.

[0008] Selectively, the above method, This further includes determining monitoring reporting conditions that include reporting thresholds and / or reporting time durations, If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation to perform a performance monitoring report will be triggered. The monitoring event is that the performance monitoring accuracy does not meet the usage conditions, and the first number of times the performance monitoring accuracy meets the subscription conditions for the monitoring event within the reporting time duration is greater than the reporting threshold, thereby triggering a performance monitoring report operation, wherein the content of the report is to indicate that the currently used model is invalid, and the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event, or the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event and the number of real-time times the performance monitoring accuracy meets the withdrawal conditions for the monitoring event, or the first number refers to the number of consecutive times the performance monitoring accuracy meets the subscription conditions for the monitoring event. And / or, if the aforementioned monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then the operation of the performance monitoring report is triggered. The monitoring event is that the performance monitoring accuracy satisfies the usage conditions, and the second number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event within the reporting time length is greater than the reporting threshold, the operation of a performance monitoring report is triggered, the content of which is to indicate that the currently used model is invalid, the second number being determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event, or the second number being determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event and the number of real-time times the performance monitoring accuracy satisfies the joining condition of the monitoring event, or the second number being a consecutive number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event.

[0009] Selectively, the model includes the currently used model and at least one candidate model. Depending on the two test sample points mentioned above, obtaining performance monitoring accuracy is possible. Obtain a first performance monitoring accuracy for the currently used model, depending on at least two test sample points corresponding to the currently used model, This includes obtaining a second performance monitoring accuracy for the candidate model, depending on at least two test sample points corresponding to the candidate model.

[0010] Selectively, the monitoring events include a first monitoring event for the currently used model and a second monitoring event for the candidate model. The reported threshold includes a first reported threshold for the currently used model and a second reported threshold for the candidate model. The reported timing time length includes the first reported timing time length for the currently used model and the second reported timing time length for the candidate model.

[0011] Selectively, if the aforementioned monitoring event is one in which the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then triggering the operation of the performance monitoring report is: If the first monitoring event indicates that the first performance monitoring accuracy does not meet the usage conditions, and the first condition is met, and the second monitoring event indicates that the second performance monitoring accuracy of any candidate model meets the usage conditions, and the second condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch, and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not meet the usage conditions and the third condition is met, then a performance monitoring report operation is triggered, the content of which is to instruct a model switch and includes carrying information of the candidate model. Among these, the first condition includes the third number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event, or the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the third number refers to the consecutive number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event. The second condition includes that the fourth number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the fourth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The third condition includes that the fifth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fifth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the fifth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the fifth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event.

[0012] Selectively, if the aforementioned monitoring event is one in which the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then triggering the operation of the performance monitoring report is: If the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model satisfies the usage conditions and the fifth condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not satisfy the usage conditions and the sixth condition is met, then a performance monitoring report operation is triggered, and the content of the report is to instruct a model switch and includes carrying information of the candidate model. Among these, the fourth condition includes that the sixth number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event, or the sixth number refers to the consecutive number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event. The fifth condition includes that the seventh instance in which the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the seventh instance refers to the consecutive number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The sixth condition includes that the eighth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the eighth number refers to the consecutive number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event.

[0013] Selectively controlling the trigger operation of performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: The monitoring event further includes triggering a performance monitoring report operation if the performance monitoring accuracy of the currently used model is lower than that of any candidate model. The report will contain instructions to switch models and will carry information about the aforementioned candidate models.

[0014] Selectively, the above method, The method further includes obtaining the at least two test sample points using a time-domain sliding window of first length.

[0015] Selectively, depending on at least two test sample points, before obtaining performance monitoring accuracy, The method further includes pre-processing the at least two test sample points, Of these, the aforementioned preprocessing includes at least one of the following: weighting and anomaly exclusion.

[0016] Selectively, the joining and / or leaving conditions are determined according to the hysteresis parameters and the corresponding threshold values.

[0017] Selectively, the monitoring event-related information is determined according to the protocol's predefined and / or network-side configuration. And / or, the monitoring reporting conditions are determined according to the protocol's predefined and / or network-side configuration. and / or, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration. And / or, the hysteresis parameters are determined according to the protocol's predefined and / or network-side configuration. And / or, the threshold threshold is determined according to the protocol's predefined and / or network-side configuration.

[0018] The embodiments of this disclosure are information processing methods applicable to a first device, This includes obtaining at least two test sample points using a time-domain sliding window of first length, Of these, the test sample points are determined according to the target CSI and the CSI restored by the model that corresponds to the target CSI. The present invention further provides an information processing method in which the first device is a network device or a terminal.

[0019] Selectively, the above method, The method further includes pre-processing the at least two test sample points, Of these, the aforementioned preprocessing includes at least one of the following: weighting and anomaly exclusion.

[0020] Selectively, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration.

[0021] An embodiment of the present disclosure is an information processing device referred to as the first device, which includes a memory, a transceiver, and a processor. Memory is for storing computer programs, and transceivers are for sending and receiving data under the control of the processor, and the processor reads the computer programs in the memory, The performance monitoring accuracy is obtained based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI restored by a model corresponding to the target CSI, and the performance monitoring accuracy is based on the statistical results of the at least two test sample points. To determine the monitoring event-related information regarding the performance monitoring accuracy, which includes the definition of the monitoring event and the joining and / or leaving conditions of the monitoring event. This is for controlling the trigger operation of performance monitoring reports according to the performance monitoring accuracy and monitoring event-related information, The first device further provides an information processing device which is a network device or terminal.

[0022] Selectively controlling the trigger operation of performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the aforementioned monitoring event, then the operation of the performance monitoring report is triggered. Alternatively, the monitoring event includes triggering a performance monitoring report operation if the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event.

[0023] Selectively, the above operation is performed. This further includes determining monitoring reporting conditions that include reporting thresholds and / or reporting time durations, If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation to perform a performance monitoring report will be triggered. The monitoring event is that the performance monitoring accuracy does not meet the usage conditions, and the first number of times the performance monitoring accuracy meets the subscription conditions for the monitoring event within the reporting time duration is greater than the reporting threshold, thereby triggering a performance monitoring report operation, wherein the content of the report is to indicate that the currently used model is invalid, and the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event, or the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event and the number of real-time times the performance monitoring accuracy meets the withdrawal conditions for the monitoring event, or the first number refers to the number of consecutive times the performance monitoring accuracy meets the subscription conditions for the monitoring event. And / or, if the aforementioned monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then the operation of the performance monitoring report is triggered. The monitoring event is that the performance monitoring accuracy satisfies the usage conditions, and the second number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event within the reporting time length is greater than the reporting threshold, the operation of a performance monitoring report is triggered, the content of which is to indicate that the currently used model is invalid, the second number being determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event, or the second number being determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event and the number of real-time times the performance monitoring accuracy satisfies the joining condition of the monitoring event, or the second number being a consecutive number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event.

[0024] Selectively, the model includes the currently used model and at least one candidate model. Depending on the two test sample points mentioned above, obtaining performance monitoring accuracy is possible. Obtain a first performance monitoring accuracy for the currently used model, depending on at least two test sample points corresponding to the currently used model, This includes obtaining a second performance monitoring accuracy for the candidate model, depending on at least two test sample points corresponding to the candidate model.

[0025] Selectively, the monitoring events include a first monitoring event for the currently used model and a second monitoring event for the candidate model. The reported threshold includes a first reported threshold for the currently used model and a second reported threshold for the candidate model. The reported timing time length includes the first reported timing time length for the currently used model and the second reported timing time length for the candidate model.

[0026] Selectively, if the aforementioned monitoring event is one in which the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then triggering the operation of the performance monitoring report is: If the first monitoring event indicates that the first performance monitoring accuracy does not meet the usage conditions, and the first condition is met, and the second monitoring event indicates that the second performance monitoring accuracy of any candidate model meets the usage conditions, and the second condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch, and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not meet the usage conditions and the third condition is met, then a performance monitoring report operation is triggered, the content of which is to instruct a model switch and includes carrying information of the candidate model. Among these, the first condition includes the third number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event, or the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the third number refers to the consecutive number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event. The second condition includes that the fourth number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the fourth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The third condition includes that the fifth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fifth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the fifth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the fifth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event.

[0027] Selectively, if the aforementioned monitoring event is one in which the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then triggering the operation of the performance monitoring report is: If the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model satisfies the usage conditions and the fifth condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not satisfy the usage conditions and the sixth condition is met, then a performance monitoring report operation is triggered, and the content of the report is to instruct a model switch and includes carrying information of the candidate model. Among these, the fourth condition includes that the sixth number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event, or the sixth number refers to the consecutive number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event. The fifth condition includes that the seventh instance in which the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the seventh instance refers to the consecutive number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The sixth condition includes that the eighth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the eighth number refers to the consecutive number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event.

[0028] Selectively controlling the trigger operation of performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: The monitoring event further includes triggering a performance monitoring report operation if the performance monitoring accuracy of the currently used model is lower than that of any candidate model. The report will contain instructions to switch models and will carry information about the aforementioned candidate models.

[0029] Selectively, the above operation is performed. The method further includes obtaining the at least two test sample points using a time-domain sliding window of first length.

[0030] Selectively, the above operation is performed. The method further includes pre-processing at least two test sample points before obtaining performance monitoring accuracy, depending on at least two test sample points. Of these, the aforementioned preprocessing includes at least one of the following: weighting and anomaly exclusion.

[0031] Selectively, the joining and / or leaving conditions are determined according to the hysteresis parameters and the corresponding threshold values.

[0032] Selectively, the monitoring event-related information is determined according to the protocol's predefined and / or network-side configuration. And / or, the monitoring reporting conditions are determined according to the protocol's predefined and / or network-side configuration. and / or, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration. And / or, the hysteresis parameters are determined according to the protocol's predefined and / or network-side configuration. And / or, the threshold threshold is determined according to the protocol's predefined and / or network-side configuration.

[0033] An embodiment of the present disclosure is an information processing device referred to as the first device, which includes a memory, a transceiver, and a processor. Memory is for storing computer programs, and transceivers are for sending and receiving data under the control of the processor, and the processor reads the computer programs in the memory, This is for performing the operation of obtaining at least two test sample points using a time-domain sliding window of first length. Of these, the test sample points are determined according to the target CSI and the CSI restored by the model that corresponds to the target CSI. The first device further provides an information processing device which is a network device or a terminal.

[0034] Selectively, the above operation is performed. The method further includes pre-processing the at least two test sample points, Of these, the aforementioned preprocessing includes at least one of the following: weighting and anomaly exclusion.

[0035] Selectively, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration.

[0036] The embodiments of this disclosure are information processing devices applicable to the first device, A first acquisition unit for obtaining performance monitoring accuracy based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI restored by a model corresponding to the target CSI, and the performance monitoring accuracy is based on statistical results of the at least two test sample points. A first determination unit for determining monitoring event-related information relating to the performance monitoring accuracy, which includes a definition of a monitoring event and the joining and / or leaving conditions of the monitoring event, Includes a first control unit for controlling the trigger operation of performance monitoring reports according to the performance monitoring accuracy and monitoring event-related information, The first device further provides an information processing device, which is a network device or terminal.

[0037] Selectively controlling the trigger operation of performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the aforementioned monitoring event, then the operation of the performance monitoring report is triggered. Alternatively, the monitoring event includes triggering a performance monitoring report operation if the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event.

[0038] Selectively, the apparatus, It further includes a second determination unit for determining monitoring reporting conditions, which include reporting thresholds and / or reporting time durations. If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation to perform a performance monitoring report will be triggered. The monitoring event is that the performance monitoring accuracy does not meet the usage conditions, and the first number of times the performance monitoring accuracy meets the subscription conditions for the monitoring event within the reporting time duration is greater than the reporting threshold, thereby triggering a performance monitoring report operation, wherein the content of the report is to indicate that the currently used model is invalid, and the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event, or the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event and the number of real-time times the performance monitoring accuracy meets the withdrawal conditions for the monitoring event, or the first number refers to the number of consecutive times the performance monitoring accuracy meets the subscription conditions for the monitoring event. And / or, if the aforementioned monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then the operation of the performance monitoring report is triggered. The monitoring event is that the performance monitoring accuracy satisfies the usage conditions, and the second number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event within the reporting time length is greater than the reporting threshold, the operation of a performance monitoring report is triggered, the content of which is to indicate that the currently used model is invalid, the second number being determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event, or the second number being determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event and the number of real-time times the performance monitoring accuracy satisfies the joining condition of the monitoring event, or the second number being a consecutive number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event.

[0039] Selectively, the model includes the currently used model and at least one candidate model. Depending on the two test sample points mentioned above, obtaining performance monitoring accuracy is possible. Obtain a first performance monitoring accuracy for the currently used model, depending on at least two test sample points corresponding to the currently used model, This includes obtaining a second performance monitoring accuracy for the candidate model, depending on at least two test sample points corresponding to the candidate model.

[0040] Selectively, the monitoring events include a first monitoring event for the currently used model and a second monitoring event for the candidate model. The reported threshold includes a first reported threshold for the currently used model and a second reported threshold for the candidate model. The reported timing time length includes the first reported timing time length for the currently used model and the second reported timing time length for the candidate model.

[0041] Selectively, if the aforementioned monitoring event is one in which the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then triggering the operation of the performance monitoring report is: If the first monitoring event indicates that the first performance monitoring accuracy does not meet the usage conditions, and the first condition is met, and the second monitoring event indicates that the second performance monitoring accuracy of any candidate model meets the usage conditions, and the second condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch, and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not meet the usage conditions and the third condition is met, then a performance monitoring report operation is triggered, the content of which is to instruct a model switch and includes carrying information of the candidate model. Among these, the first condition includes the third number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event, or the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the third number refers to the consecutive number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event. The second condition includes that the fourth number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the fourth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The third condition includes that the fifth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fifth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the fifth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the fifth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event.

[0042] Selectively, if the aforementioned monitoring event is one in which the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then triggering the operation of the performance monitoring report is: If the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model satisfies the usage conditions and the fifth condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not satisfy the usage conditions and the sixth condition is met, then a performance monitoring report operation is triggered, and the content of the report is to instruct a model switch and includes carrying information of the candidate model. Among these, the fourth condition includes that the sixth number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event, or the sixth number refers to the consecutive number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event. The fifth condition includes that the seventh instance in which the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the seventh instance refers to the consecutive number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The sixth condition includes that the eighth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the eighth number refers to the consecutive number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event.

[0043] Selectively controlling the trigger operation of performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: The monitoring event further includes triggering a performance monitoring report operation if the performance monitoring accuracy of the currently used model is lower than that of any candidate model. The report will contain instructions to switch models and will carry information about the aforementioned candidate models.

[0044] Selectively, the apparatus, The system further includes a second acquisition unit for acquiring the at least two test sample points using a time-domain sliding window of first length.

[0045] Selectively, the apparatus, Depending on at least two test sample points, the system further includes a first processing unit for pre-processing the at least two test sample points before obtaining performance monitoring accuracy, Of these, the aforementioned preprocessing includes at least one of the following: weighting and anomaly exclusion.

[0046] Selectively, the joining and / or leaving conditions are determined according to the hysteresis parameters and the corresponding threshold values.

[0047] Selectively, the monitoring event-related information is determined according to the protocol's predefined and / or network-side configuration. And / or, the monitoring reporting conditions are determined according to the protocol's predefined and / or network-side configuration. and / or, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration. And / or, the hysteresis parameters are determined according to the protocol's predefined and / or network-side configuration. And / or, the threshold threshold is determined according to the protocol's predefined and / or network-side configuration.

[0048] The embodiments of this disclosure are information processing devices applicable to the first device, It includes a third acquisition unit for acquiring at least two test sample points using a time-domain sliding window of first length, Of these, the test sample points are determined according to the target CSI and the CSI restored by the model that corresponds to the target CSI. The first device further provides an information processing device which is a network device or terminal.

[0049] Selectively, the apparatus, The system further includes a second processing unit for pre-processing the at least two test sample points, Of these, the aforementioned preprocessing includes at least one of the following: weighting and anomaly exclusion.

[0050] Selectively, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration.

[0051] Embodiments of the present disclosure further provide a non-temporary, readable storage medium storing a computer program, wherein the computer program causes a processor to execute the above-described method. [Effects of the Invention]

[0052] The beneficial effects of the above-described technical aspects of this disclosure are as follows: In the above embodiment, the information processing method acquires performance monitoring accuracy according to at least two test sample points, the test sample points are determined according to target channel status information (CSI) and CSI restored by the model corresponding to the target CSI, the performance monitoring accuracy is based on statistical results of the at least two test sample points, monitoring event-related information for the performance monitoring accuracy is determined, which includes the definition of a monitoring event and the join and / or leave conditions of the monitoring event, and the trigger operation for performance monitoring reporting is controlled according to the performance monitoring accuracy and monitoring event-related information, the first device being a network device or terminal. This supports the realization of control over the trigger for performance monitoring reporting according to performance monitoring accuracy (i.e., monitoring results), thereby supporting the accurate realization of decision operations such as model switching, fallback, update, activation and / or deactivation, and suitably solves the problem of not being able to accurately realize model decision operations that exist in the information processing methods for performance monitoring in related technologies. [Brief explanation of the drawing]

[0053] [Figure 1] Figure 1 is a schematic diagram of the architecture of a wireless communication system according to an embodiment of the present disclosure. [Figure 2] Figure 2 is a schematic flowchart of the network-side model monitoring according to an embodiment of the present disclosure. [Figure 3] Figure 3 is a schematic diagram of the UE-side model monitoring flow according to an embodiment of the present disclosure. [Figure 4] Figure 4 is a schematic diagram of the UE-side model monitoring flow according to an embodiment of the present disclosure. [Figure 5] Figure 5 is a schematic flowchart of the information processing method according to an embodiment of the present disclosure. [Figure 6] Figure 6 is a schematic flowchart of the information processing method according to an embodiment of the present disclosure. [Figure 7]Figure 7 is a schematic diagram of the time of acquisition of test sample points according to the embodiment of this disclosure. [Figure 8] Figure 8 is a schematic diagram of the structure of an information processing device according to an embodiment of the present disclosure. [Figure 9] Figure 9 is a schematic diagram of the structure of an information processing device according to an embodiment of the present disclosure. [Figure 10] Figure 10 is a schematic diagram of the structure of an information processing device according to an embodiment of the present disclosure. [Figure 11] Figure 11 is a schematic diagram of the structure of an information processing device according to an embodiment of the present disclosure. [Modes for carrying out the invention]

[0054] The following describes the technical aspects of the embodiments of this disclosure clearly and completely, with reference to the drawings of the embodiments of this disclosure. It is clear that the embodiments described are only a selection of the embodiments of this disclosure, not all of them. Any other embodiments that can be obtained by those skilled in the art without creative work based on the embodiments of this disclosure are all covered by this disclosure.

[0055] In the embodiments of this disclosure, the term "and / or" describes a correlation between related objects, indicating that there may be three possible relationships. For example, A and / or B could represent three cases: A exists alone, both A and B exist, and B exists alone. The symbol " / " generally indicates an "or" relationship between the related objects before and after it.

[0056] In the embodiments of this disclosure, the term "plural" refers to two or more, and the same applies to other quantifiers.

[0057] It should be noted that the technical embodiments described herein are applicable to a variety of systems, particularly fifth-generation (5G) mobile communication systems. For example, applicable systems may include global system of mobile communication (GSM®) systems, code division multiple access (CDMA) systems, wideband code division multiple access (WCDMA®) general packet radio service (GPRS) systems, long term evolution (LTE®) systems, LTE frequency division duplex (FDD) systems, LTE time division duplex (TDD) systems, long term evolution advanced (LTE-A) systems, universal mobile telecommunication systems (UMTS), worldwide interoperability for microwave access (WiMAX) systems, 5G New Radio (NR) systems, and the like. Each of these various systems includes terminal equipment and network equipment. The system may further include a core network component, such as an evolved packet system (EPS), a 5G system (5GS), etc.

[0058] Figure 1 shows a block diagram of a wireless communication system applicable to embodiments of the present disclosure. The wireless communication system includes terminal equipment (which may be abbreviated as terminals) and network equipment.

[0059] The terminal equipment in the embodiments of this disclosure may be equipment that provides voice and / or data connectivity to a user, a handheld device with wireless connectivity, or other processing equipment connected to a wireless modem. In different systems, the name of the terminal equipment may differ; for example, in a 5G system, the terminal equipment may be called User Equipment (UE). The wireless terminal equipment can communicate with one or more Core Networks (CN) via a Radio Access Network (RAN). The wireless terminal equipment may be mobile terminal equipment such as a mobile phone (also called a "cellular" phone), or a computer with mobile terminal equipment, and may be, for example, a portable, pocket-sized, handheld, computer-integrated, or in-vehicle mobile device, which exchanges voice and / or data with the Radio Access Network. For example, the devices may include Personal Communication Service (PCS) telephones, cordless telephones, Session Initiated Protocol (SIP) telephones, Wireless Local Loop (WLL) stations, Personal Digital Assistants (PDAs), and the like. Wireless terminal devices may also be referred to as systems, subscriber units, subscriber stations, mobile stations, mobile consoles, remote stations, access points, remote terminals, access terminals, user terminals, user agents, and user devices, but are not limited to these in the embodiments of this disclosure.

[0060] The network equipment relating to the embodiments of this disclosure may be a base station, which may include multiple cells providing services to terminals. Depending on different specific application scenarios, the base station may be called an access point, or a device that communicates with wireless terminal equipment via one or more sectors on a wireless interface in an access network, or by other names. The network equipment may be used to convert received wireless frames and Internet Protocol (IP) packets to each other and act as a router between wireless terminal equipment and the rest of the access network, where the rest of the access network may include an Internet Protocol (IP) communication network. The network equipment may coordinate attribute management for the wireless interface. For example, the network equipment relating to the embodiments of this disclosure may be a network device (Base Transceiver Station, BTS) in a Global System for Mobile communications (GSM) or Code Division Multiple Access (CDMA), a network device (NodeB) in a Wide-band Code Division Multiple Access (WCDMA), an evolutionary network device (eNB or e-NodeB) in a long-term evolution (LTE) system, a 5G base station (the next generation Node B, gNB) in a 5G network architecture (next generation system), a home evolved Node B (HeNB), a relay node, a home base station (femto), a pico base station (pico), etc., but the embodiments of this disclosure are not limited to these.In some network configurations, network equipment may include centralized unit (CU) nodes and distributed unit (DU) nodes, and the centralized and distributed units may be geographically separated.

[0061] Network devices and terminal devices may each use one or more antennas to perform multi-input multi-output (MIMO) transmission, and the MIMO transmission may be single-user MIMO (SU-MIMO) or multiple-user MIMO (MU-MIMO). Depending on the combination and number of antennas, the MIMO transmission may be 2-dimensional MIMO (2D-MIMO), 3-dimensional MIMO (3D-MIMO), full-dimensional MIMO (FD-MIMO), or massive-MIMO, and may also be diversity transmission, pre-encoding transmission, beamforming transmission, etc.

[0062] The following describes the aspects of the embodiments of this disclosure.

[0063] Monitoring of artificial intelligence (AI) / machine learning (ML) models (i.e., AI and / or ML models) is a new problem that has arisen since the introduction of AI / ML algorithms in 5G NR. Each AI / ML model has its own scope of application, and AI / ML model monitoring refers to the need to monitor the model's performance to determine its applicability when the use case of the AI / ML model differs from its training case, and the intrinsic relationship between input and output changes. It should be explained that model monitoring involves monitoring the model's performance and determining its performance based on the monitoring results; therefore, "model monitoring" in the embodiments of this disclosure can be understood or replaced with "performance monitoring." For ease of description and understanding, the specific embodiments and accompanying drawings of this disclosure will primarily use the expression "model monitoring."

[0064] In real-world systems, the performance of a current model in AI / ML model monitoring should not be judged based on a single result, because actions based on a single evaluation result may lead to frequent switching, fallback, updating, activation, and / or deactivation of AI / ML models, i.e., a so-called ping-pong effect.

[0065] In existing configurations, the conclusion reached to improve the stability of AI / ML model monitoring for compressed Channel State Information (CSI) is to introduce a single dataset with K test sample points, and the AI / ML model monitoring accuracy is determined by the statistical results of performance evaluations at these K sample points. Based on this conclusion, the next issues to be resolved are: (1) how to obtain these K test sample point data within time-series data, and (2) how to instruct the network side to make decisions regarding activation, deactivation, update, fallback, and / or switching of the AI / ML model based on the AI / ML model monitoring accuracy obtained from the statistical results of performance evaluations at these K sample points. Furthermore, it is also necessary to address the potential ping-pong effect caused by frequent changes to the AI / ML model.

[0066] Based on the above, the following provides a detailed explanation of the related content.

[0067] 1. Regarding AI / ML-based CSI compression, the following AI / ML model monitoring metrics are summarized in existing configurations. 1. Monitor intermediate key performance indicators (KPIs) (e.g., Squared Generalized Cosine Similarity (SGCS)), i.e., monitor them based on inference accuracy. 2. The final (Eventual) KPI is used as a monitoring metric (e.g., throughput, hypothetical throughput, block error rate (BLER), hypothetical BLER, negative acknowledgment (NACK), or acknowledgment (ACK)), i.e., it is monitored based on system performance. 3. Model monitoring based on conventional CSI: This is a method based on a separate conventional CSI report. 4. Monitoring based on input / output data: For example, monitoring based on data drift between the training dataset and the observed dataset, or monitoring based on the output distribution.

[0068] 2. The primary AI / ML model monitoring entity for CSI compression may be the terminal (i.e., model monitoring on the terminal (User Equipment, UE) side) or the network (i.e., model monitoring on the network side). In existing configurations, the following conclusions have been reached regarding AI / ML model monitoring using Intermediate KPIs as monitoring metrics. 1. Network-side AI / ML model monitoring may be based on the target CSI reported from the UE side (which is acquired by the UE side and refers to the input of the CSI generation portion of the AI / ML model, also called ground truth CSI), and the CSI generation portion of the AI / ML model may achieve CSI compression and specifically include CSI generation inference and quantization. 2. Monitoring of AI / ML models on the UE side (1) The reconstructed CSI sent back from the network (which is obtained on the network side and refers to the output of the CSI reconstruction part of the AI / ML model) may be used as the basis, and the CSI reconstruction part of the AI / ML model may be capable of realizing CSI reconstruction and may specifically include inverse quantization and CSI reconstruction inference. (2) The AI / ML model of the network-side CSI restoration portion present on the UE side may be used as an alternative model. 3. Other possibilities include, for example, the UE side being based on a model that directly outputs intermediate KPIs, and further, for example, the network side being based on the target CSI (Time Division Duplex (TDD) mode) measured by the UE side's Channel Sounding Reference Signal (SRS).

[0069] 3. Based on the above description, the network-side AI / ML model monitoring flow based on the target CSI reported from the UE side may include the following, as shown in Figure 2. The UE acquires the target CSI based on the received CSI-related reference signal, then uses the AI / ML model (CSI generation portion) to obtain the compressed CSI, and subsequently performs a compressed CSI feedback report from the UE, followed by a target CSI report from the UE. The network inputs the received compressed CSI into an AI / ML model (CSI reconstruction part), obtains the reconstructed CSI from the AI / ML model, then monitors the network based on that and the target CSI reported from the UE obtained by the network, and calculates an Intermediate KPI based on the CSI reconstructed by the AI / ML model and the target CSI reported from the UE obtained by the network.

[0070] IV. Based on the above description, the flow of AI / ML model monitoring on the UE side based on the restored CSI returned from the network side may include the following, as shown in Figure 3. The UE acquires the target CSI based on the received CSI-related reference signal, then uses the AI / ML model (CSI generation portion) to obtain the compressed CSI, and subsequently performs a compressed CSI feedback report from the UE side. The network inputs the received compressed CSI into the AI / ML model (CSI restoration part), obtains the restored CSI from the AI / ML model, and then performs the return of the restored CSI from the network side (i.e., the return of the CSI restored by the AI / ML model). The UE calculates an Intermediate KPI based on the recovered CSI returned from the network side acquired by the UE and the target CSI, thereby enabling monitoring on the UE side.

[0071] 5. Based on the above description, the AI / ML model monitoring flow on the UE side, based on the alternative model existing on the UE side (i.e., based on the alternative model of the AI / ML model for the network-side CSI restoration part that exists on the UE side), may include the following, as shown in Figure 4. The UE acquires the target CSI based on the received CSI-related reference signal, then uses the AI / ML model (CSI generation portion) to obtain the compressed CSI, and subsequently performs a compressed CSI feedback report from the UE side. The network inputs the received compressed CSI into an AI / ML model (CSI reconstruction part), and obtains the reconstructed CSI from the AI / ML model. The UE inputs the compressed CSI obtained by the AI / ML model into an AI / ML model (an alternative model for the CSI restoration part), obtains the restored CSI using the alternative model on the UE side, and then the UE calculates an Intermediate KPI based on the target CSI and the CSI restored using the alternative model on the UE side, thereby enabling monitoring on the UE side.

[0072] 6. In existing configurations, the following conclusions have been reached regarding AI / ML model monitoring methods in CSI compression use cases based on intermediate KPIs for model accuracy. 1. When evaluating the performance of monitoring CSI compression based on intermdediate KPIs, the following model monitoring method should be used. Step 1: Generate a test dataset containing K test sample points. Step 2: For each of the K test sample points, monitor the intermediate KPI deviation KPI. Diff The formula KPI Diff =f(KPI Actual , KPI Genie ) is calculated using, and of which, KPI Actual This is the actual intermediate KPI, and KPI GenieThis is an ideal intermediate KPI (genie-aided intermediate KPI), Step 3: KPIs at K test sample points Diff The statistical results are calculated, and these results represent the performance of the monitoring accuracy. Of these, (1) KPI Genie This was introduced for evaluation and comparison purposes and may not be available in actual networks. (2) The complexity, overhead, and delay of the monitoring methods will be reported by each company.

[0073] The above conclusions indicate that a single test dataset for AI / ML model monitoring, containing K test sample points, is introduced. The statistical results obtained from calculations on this dataset for the K test sample points can represent the monitoring accuracy performance. Therefore, the impact of a single test data point on monitoring accuracy is limited, and the stability of the evaluation results for model monitoring performance can be guaranteed to some extent.

[0074] In existing implementations, a proposed method to improve the stability of AI / ML model monitoring with CSI compression involves introducing a single dataset with K test sample points, where the AI / ML model monitoring accuracy is the statistical result of performance evaluation at these K sample points. This method has a limited impact on monitoring accuracy from a single test data point, and can guarantee a certain degree of stability in the evaluation results of model monitoring performance. However, existing implementations still suffer from the following drawbacks. (1) Existing embodiments do not include specific examples that describe how to obtain a dataset of K test sample points, or when the model monitoring entity should calculate the monitoring accuracy according to the dataset of K test sample points. (2) After calculating the monitoring accuracy according to the statistical results of the dataset of K test sample points, existing embodiments do not have a specific embodiment that describes how to use the acquired monitoring accuracy to trigger AI / ML model monitoring reports, and the design of such trigger mechanism needs to be able to suppress the potential ping-pong effect caused by frequent changes to the AI / ML model.

[0075] Based on the foregoing, the embodiments of this disclosure provide information processing methods, apparatuses, and devices for solving the problem of being unable to support the accurate realization of model determination operations in information processing methods for model monitoring in related technologies. Of these, the methods, apparatuses, and devices are based on the same application concept, and since the principles for solving the problem are similar, the embodiments of the methods, apparatuses, and devices are referential to one another, and overlapping parts will not be repeated.

[0076] The information processing method according to the embodiment of this disclosure is applicable to a first device, and as shown in Figure 5, the method includes the following steps 51 to 53: Step 51 is to obtain model monitoring accuracy based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI restored by the model that corresponds to the target CSI, and the model monitoring accuracy is based on the statistical results of the at least two test sample points. Step 52 is to determine the monitoring event-related information for the model monitoring accuracy, which includes the definition of the monitoring event (indicating what kind of event it is) and the join and / or leave conditions for the monitoring event (indicating how to determine the join and / or leave of the monitoring event). Step 53 involves controlling the trigger operation for the model monitoring report according to the model monitoring accuracy and monitoring event-related information, wherein the first device is a network device or terminal.

[0077] Depending on whether the monitoring is on the network side or the terminal side (i.e., whether the first device is a network device or a terminal), the target channel state information (i.e., target CSI) can be obtained in several ways, such as (1) by a CSI-related reference signal (e.g., the target CSI in Figures 3 and 4 above), or (2) obtained by a CSI-related reference signal and reported by the terminal (i.e., the terminal's report is quantized and / or compressed again, resulting in new errors, for example, the target CSI obtained on the network side and reported by the UE side in Figure 2 above). Of course, other possibilities may also be included and are not limited here.

[0078] Regarding the acquisition of the "CSI restored by the model that corresponds to the target CSI," there are several possibilities, such as (1) it being the output of the CSI restoration part of the AI / ML model, as shown in Figure 2 above; (2) it being acquired when the output of the CSI restoration part of the AI / ML model is sent back to the terminal via the network, as shown in Figure 3 above; or (3) it being acquired by restoring it using an alternative model on the UE side, as shown in Figure 4 above. Of course, other possibilities may also be included, and we are not limited to them here.

[0079] The information processing method according to the embodiment of this disclosure acquires model monitoring accuracy according to at least two test sample points, the test sample points are determined according to target channel status information (CSI) and CSI restored by the model corresponding to the target CSI, the model monitoring accuracy is based on statistical results of the at least two test sample points, monitoring event-related information for the model monitoring accuracy is determined, which includes the definition of a monitoring event and the join and / or leave conditions of the monitoring event, and the trigger operation of the model monitoring report is controlled according to the model monitoring accuracy and the monitoring event-related information, the first device being a network device or terminal. This supports the realization of control over the trigger of the model monitoring report according to the model monitoring accuracy (i.e., monitoring results), thereby supporting the accurate realization of decision operations such as model switching, fallback, update, activation and / or deactivation, and suitably solves the problem of not being able to accurately realize model decision operations that exist in the information processing methods for model monitoring in related technologies.

[0080] In addition, controlling the trigger operation of the model monitoring report according to the aforementioned model monitoring accuracy and monitoring event-related information includes triggering the model monitoring report operation if the monitoring event is such that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the joining conditions for the monitoring event, or triggering the model monitoring report operation if the monitoring event is such that the model monitoring accuracy meets the usage conditions and the model monitoring accuracy meets the leaving conditions for the monitoring event.

[0081] In this way, concrete control over the trigger operation of model monitoring reports can be realized. Furthermore, a special case can be specifically realized in which a model monitoring report can be triggered by joining or leaving a certain monitoring event, but this is not the only such case.

[0082] Furthermore, the information processing method further includes determining monitoring reporting conditions which include a reporting threshold and / or reporting time length, (1) triggering a model monitoring report operation if the monitoring event described above is such that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the enrollment conditions for the monitoring event, and (2) triggering a model monitoring report operation if the monitoring event is such that the model monitoring accuracy does not meet the usage conditions (e.g., the model monitoring accuracy is less than a certain threshold) and the first number of times the model monitoring accuracy has met the enrollment conditions for the monitoring event within the reporting time length is greater than the reporting threshold, wherein the content of the report is to indicate that the currently used model is invalid, and the first number is determined according to the real-time number of times the model monitoring accuracy has met the enrollment conditions for the monitoring event (e.g., if the enrollment condition is met once, the first number is increased by 1), or the first number is determined by the model monitoring accuracy meeting the enrollment conditions for the monitoring event The number of real-time occurrences and the number of real-time occurrences in which the model monitoring accuracy satisfies the withdrawal condition of the monitoring event are determined accordingly (for example, if the join condition is met once, the first count is increased by 1, and if the withdrawal condition is met once, the first count is decreased by 1), or the first count refers to the number of consecutive times in which the model monitoring accuracy satisfies the join condition of the monitoring event, and / or (2) if the monitoring event described above is one in which the model monitoring accuracy satisfies the usage condition and the model monitoring accuracy satisfies the withdrawal condition of the monitoring event, the operation of the model monitoring report is triggered if the monitoring event is one in which the model monitoring accuracy satisfies the usage condition (for example, the model monitoring accuracy is greater than a certain threshold) and the second number of times in which the model monitoring accuracy satisfies the withdrawal condition of the monitoring event within the reporting time length is greater than the reporting threshold, wherein the content of the report is to indicate that the currently used model is invalid, and the second number isThe model monitoring accuracy is determined according to the number of real-time instances in which the exit condition of the monitoring event is met (for example, if the exit condition is met once, the second count is increased by 1), or the second count is determined according to the number of real-time instances in which the model monitoring accuracy met the exit condition of the monitoring event and the number of real-time instances in which the model monitoring accuracy met the joining condition of the monitoring event (for example, if the joining condition is met once, the second count is decreased by 1, and if the exit condition is met once, the second count is increased by 1), or the second count refers to the number of consecutive instances in which the model monitoring accuracy met the exit condition of the monitoring event.

[0083] In this way, the trigger operation for model monitoring reports can be controlled more precisely. Specifically, in this embodiment, there may be two layers as follows: (1) The first layer involves comparing the monitoring accuracy calculated based on K test sample points (corresponding to at least two of the above-mentioned test sample points) within the current sliding window (i.e., the time-domain sliding window) with monitoring event-related information (definition of the monitoring event, join conditions and / or exit conditions for the monitoring event) to obtain the monitoring event status based on the current sliding window. Therefore, the first layer determines the current state of a single monitoring event (i.e., whether a certain monitoring event is currently joining or leaving) by comparing the monitoring accuracy calculated using K test sample points in a certain window with the monitoring event state threshold (including the threshold values ​​for join conditions and / or exit conditions). The first layer may also be understood as determining the state of a single monitoring event based on the monitoring accuracy. (2) In the second layer, after sliding window processing, the monitoring accuracy calculated from K test sample points in multiple windows allows us to obtain the monitoring event state (joining and / or leaving a certain monitoring event) at corresponding multiple time points. We then compare the cumulative effect of the monitoring event state over a certain period (defined by a timer) (e.g., the current number of joins and / or leaves of a certain monitoring event) with the threshold for triggering a report by the monitoring event (e.g., the maximum threshold for the number of joins of a certain monitoring event, the maximum threshold for the number of leaves of a certain monitoring event, the maximum threshold for consecutive occurrences of a certain monitoring event, or the maximum threshold for consecutive leaves of a certain monitoring event, etc.) to determine whether to trigger a model monitoring report. The second layer may also be understood as determining whether a monitoring event triggers a model monitoring report based on the cumulative effect of multiple monitoring event states within a certain period.

[0084] Based on the two layers described above, the monitoring accuracy calculated from K test sample points in a single window determines only the monitoring event state of the window at the current time. To determine whether to trigger a model monitoring report, it is necessary to base the decision on the cumulative effect of the monitoring event states of multiple sliding windows. Furthermore, in the two layers described above, the decision on whether to trigger a model monitoring report is based on the cumulative effect of the monitoring event states observed in multiple sliding windows within a certain period. This minimizes the possibility of false alarms caused by a single disturbance in the results of one or more sliding windows, and consequently reduces the likelihood of ping-pong effects occurring due to frequent model switching.

[0085] In addition, this embodiment may represent a special case, for example, where the determination of whether to trigger a report is based solely on the monitoring accuracy calculated for K test sample points in a single window, without relying on the second hierarchy described above (it may be understood that joining or leaving a certain monitoring event triggers a model monitoring report).

[0086] Of these, the reported timing time length can correspond to the "inactive timer" described below, and specifically, it can be set to the length of one or more windows (i.e., time domain sliding windows), thus covering the first and second layers mentioned above.

[0087] In embodiments of the present disclosure, the model includes a currently used model and at least one candidate model, and obtaining model monitoring accuracy in accordance with at least two test sample points as described above includes obtaining a first model monitoring accuracy for the currently used model in accordance with at least two test sample points corresponding to the currently used model, and obtaining a second model monitoring accuracy for the candidate model in accordance with at least two test sample points corresponding to the candidate model.

[0088] This allows for simultaneous monitoring of the currently used model and at least one candidate model.

[0089] Among these, the monitoring event includes a first monitoring event for the currently used model and a second monitoring event for the candidate model, and / or the reporting threshold includes a first reporting threshold for the currently used model and a second reporting threshold for the candidate model, and / or the reporting timing length includes a first reporting timing length for the currently used model and a second reporting timing length for the candidate model.

[0090] In this way, it is possible to precisely control the trigger operation of model monitoring reports in response to monitoring of the currently used model and at least one candidate model.

[0091] In embodiments of this disclosure, the operation of a model monitoring report is triggered when the above-described monitoring event is that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the subscription conditions for the monitoring event, in which case (1) the first monitoring event is that the first model monitoring accuracy does not meet the usage conditions (for example, the first model monitoring accuracy is less than a certain threshold) and the first condition is met, and the second monitoring event is that the second model monitoring accuracy of any candidate model meets the usage conditions (for example, the second model monitoring accuracy is less than a certain threshold) If the first monitoring event is greater than the first model monitoring accuracy (for example, the first model monitoring accuracy is less than a certain threshold) and the first condition is met, and the second monitoring event is greater than the first reporting threshold, and the content of the report is to instruct a model switch and carries information about the candidate model, or (2) if the first monitoring event is greater than the first model monitoring accuracy (for example, the first model monitoring accuracy is less than a certain threshold) and the first condition is met, and the second monitoring event is greater than the first model monitoring accuracy (for example, the second model monitoring accuracy is less than a certain threshold) and the third condition is met, then a model monitoring report is triggered, and the content of the report is to instruct a model switch and carries information about the candidate model, of which (1) the first condition is greater than the first reporting threshold, and the third number is determined in real time according to the number of times the first model monitoring accuracy has met the joining condition of the first monitoring event. (For example, if the joining condition is met once, the count for the third time is increased by 1), or the third time is determined according to the number of real-time times the first model monitoring accuracy met the joining condition for the first monitoring event and the number of real-time times the first model monitoring accuracy met the withdrawal condition for the first monitoring event (for example, if the joining condition is met once, the count for the third time is increased by 1, and if the withdrawal condition is met once, the count for the third time is decreased by 1), or the third time refers to the number of consecutive times the first model monitoring accuracy met the joining condition for the first monitoring event,(2) The second condition includes that the fourth number of times the second model monitoring accuracy satisfies the enrollment condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fourth number is determined according to the number of real-time times the second model monitoring accuracy satisfies the enrollment condition of the second monitoring event (for example, if the enrollment condition is met once, the count of the fourth number is increased by 1), or the fourth number is determined according to the number of real-time times the second model monitoring accuracy satisfies the enrollment condition of the second monitoring event and the number of real-time times the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event (for example, if the enrollment condition is met once, the count of the fourth number is increased by 1, and if the withdrawal condition is met once, the count of the fourth number is decreased by 1), or the fourth number refers to the number of consecutive times the second model monitoring accuracy satisfies the enrollment condition of the second monitoring event. (3) The third condition means that the fifth number of times the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, and the fifth number is determined according to the real-time number of times the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event (for example, if the withdrawal condition is met once, the count of the fifth number is increased by 1), or the fifth number is determined according to the real-time number of times the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second model monitoring accuracy satisfies the joining condition of the second monitoring event (for example, if the joining condition is met once, the count of the fifth number is decreased by 1, and if the withdrawal condition is met once, the count of the fifth number is increased by 1), or the fifth number refers to the number of consecutive times the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event.

[0092] In this way, monitoring reports for model switching can be accurately implemented. The situation described above, "triggering the operation of model monitoring reporting, and the content of the report instructing a model switch," may be understood as the performance of the currently used model being too poor (the model monitoring reporting condition for the currently used model is triggered), and at the same time, the performance of at least one candidate model is good (the model monitoring reporting condition for the candidate model is triggered), and a switch is performed, but is not limited to this. The condition "the second model monitoring accuracy of any candidate model satisfies the usage conditions" may be realized as the second model monitoring accuracy of at least two candidate models satisfying the usage conditions, in which case the content of the report may carry information from at least one of the at least two candidate models, for example, information from a selected target candidate model, but is not limited to this. Similarly, the condition "the second model monitoring accuracy of any candidate model does not meet the usage conditions" may be defined as the second model monitoring accuracy of at least two candidate models not meeting the usage conditions. In this case, the report may, but is not limited to, carry information from at least one of the two candidate models, such as information from a selected target candidate model.

[0093] Of these, the operation of the model monitoring report is triggered when the aforementioned monitoring event satisfies the usage conditions of the model monitoring accuracy, and the model monitoring accuracy satisfies the exit conditions of the monitoring event, in which case (1) the first monitoring event satisfies the usage conditions of the first model monitoring accuracy (for example, the first model monitoring accuracy is greater than a certain threshold), and the fourth condition is met, and the second monitoring event satisfies the usage conditions of the second model monitoring accuracy of any candidate model (for example, the second model monitoring accuracy is greater than a certain threshold) If (1) the first monitoring event is such that the first model monitoring accuracy satisfies the usage conditions (e.g., the first model monitoring accuracy is greater than a certain threshold) and the fourth condition is met, and the second monitoring event is such that the second model monitoring accuracy of any candidate model does not satisfy the usage conditions (e.g., the second model monitoring accuracy is less than a certain threshold) If the sixth condition is met, the operation of a model monitoring report is triggered, the content of which is to instruct a model switch and carries information of the candidate model, of which (1) the fourth condition is that the sixth number of times the first model monitoring accuracy satisfies the exit condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, the sixth number of times is determined in real time by the number of times the first model monitoring accuracy satisfies the exit condition of the first monitoring event (for example, if the exit condition is 1 (1) When the condition is met multiple times, the count of the sixth count is increased by 1.) Or, the sixth count is determined according to the number of real-time times when the first model monitoring accuracy meets the exit condition of the first monitoring event and the number of real-time times when the first model monitoring accuracy meets the join condition of the first monitoring event (for example, when the join condition is met once, the count of the sixth count is decreased by 1, and when the exit condition is met once, the count of the sixth count is increased by 1). Or, the sixth count refers to the number of consecutive times when the first model monitoring accuracy meets the join condition of the first monitoring event, and (2) the fifth condition is,This includes the fact that within the second reporting time length, the seventh instance in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event is greater than the second reporting threshold, wherein the seventh instance is determined according to the real-time number of times the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event (for example, if the enrollment condition is met once, the count of the seventh instance is increased by 1), or the seventh instance is determined according to the real-time number of times the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event and the real-time number of times the second model monitoring accuracy satisfies the withdrawal condition for the second monitoring event (for example, if the enrollment condition is met once, the count of the seventh instance is increased by 1, and if the withdrawal condition is met once, the count of the seventh instance is decreased by 1), or the seventh instance refers to the consecutive number of times the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event, and (3) the The sixth condition includes that the eighth instance in which the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the eighth instance is determined according to the real-time number of times the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event (for example, if the withdrawal condition is met once, the count of the eighth instance is increased by 1), or the eighth instance is determined according to the real-time number of times the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second model monitoring accuracy satisfies the joining condition of the second monitoring event (for example, if the joining condition is met once, the count of the eighth instance is decreased by 1, and if the withdrawal condition is met once, the count of the eighth instance is increased by 1), or the eighth instance refers to the number of consecutive times the second model monitoring accuracy satisfies the withdrawal condition of the second monitoring event.

[0094] In this way, monitoring reports for model switching can be accurately implemented in a similar manner. The situation described above, "triggering the operation of model monitoring reporting, and the content of the report instructing a model switch," may be understood as the performance of the currently used model being too poor (the model monitoring reporting condition for the currently used model is triggered) and at the same time, the performance of at least one candidate model is good (the model monitoring reporting condition for the candidate model is triggered), and a switch is performed, but is not limited to this. The condition "the usage conditions are met by the second model monitoring accuracy of any candidate model" may be realized as the usage conditions being met by the second model monitoring accuracy of at least two candidate models, in which case the content of the report may carry information from at least one of the at least two candidate models, for example, information from a selected target candidate model, but is not limited to this. Similarly, the condition "the second model monitoring accuracy of any candidate model does not meet the usage conditions" may be defined as the second model monitoring accuracy of at least two candidate models not meeting the usage conditions. In this case, the report may, but is not limited to, carry information from at least one of the two candidate models, such as information from a selected target candidate model.

[0095] In embodiments of the present disclosure, controlling the trigger operation of a model monitoring report in accordance with the aforementioned model monitoring accuracy and monitoring event-related information further includes triggering a model monitoring report operation if the monitoring event is that the model monitoring accuracy of the currently used model is lower than that of any candidate model, the report instructing a model switch and carrying information about the candidate model.

[0096] In this way, it is directly possible to trigger a model monitoring report operation and switch models when the model monitoring accuracy of the currently used model is lower than that of any candidate model (for example, the new model has better monitoring accuracy than the original model, and a model switch occurs). The condition "the model monitoring accuracy of the currently used model is lower than that of any candidate model" may be implemented as the model monitoring accuracy of the currently used model being lower than that of at least two candidate models, in which case the report may, but is not limited to, carry information from at least one of the at least two candidate models, such as information from a selected target candidate model.

[0097] Furthermore, the information processing method further includes obtaining the at least two test sample points using a time-domain sliding window of a first length.

[0098] In this way, the aforementioned at least two test sample points can be accurately obtained.

[0099] In embodiments of the present disclosure, the preprocessing of at least two test sample points before obtaining model monitoring accuracy is further included, wherein the preprocessing includes at least one of weighting and anomaly exclusion (e.g., excluding any anomaly sample points).

[0100] In this way, the accuracy of the test sample points used can be improved, further enhancing the accuracy of the model monitoring accuracy achieved.

[0101] The aforementioned joining and / or leaving conditions are determined according to the hysteresis parameters and the corresponding threshold values.

[0102] In this way, the ping-pong effect caused by frequent switching of model monitoring can be avoided to some extent. Specifically, by introducing hysteresis parameters, the influence of observation disturbances on the judgment can be reduced, and the values ​​of the hysteresis parameters and / or threshold values ​​corresponding to different conditions may be different, but are not limited to this.

[0103] In embodiments of this disclosure, the monitoring event-related information is determined according to the protocol predefined and / or the network-side configuration, and / or the monitoring reporting conditions are determined according to the protocol predefined and / or the network-side configuration, and / or the first length is determined according to the protocol predefined and / or the network-side configuration, and / or the sliding step size of the time-domain sliding window is determined according to the protocol predefined and / or the network-side configuration, and / or the hysteresis parameters are determined according to the protocol predefined and / or the network-side configuration, and / or the threshold is determined according to the protocol predefined and / or the network-side configuration.

[0104] In this way, the flexibility of realizing the configuration can be increased.

[0105] The embodiments of this disclosure further provide an information processing method applicable to a first device, which, as shown in Figure 6, includes the following step 61: Step 61 is to obtain at least two test sample points using a time-domain sliding window of first length, the test sample points being determined according to a target CSI and a CSI restored by the model corresponding to the target CSI, and the first device being a network device or terminal.

[0106] Depending on whether the monitoring is on the network side or the terminal side (i.e., whether the first device is a network device or a terminal), the target channel state information (i.e., target CSI) can be obtained in several ways, such as (1) by a CSI-related reference signal (e.g., the target CSI in Figures 3 and 4 above), or (2) obtained by a CSI-related reference signal and reported by the terminal (i.e., the terminal's report is quantized and / or compressed again, resulting in new errors, for example, the target CSI obtained on the network side and reported by the UE side in Figure 2 above). Of course, other possibilities may also be included and are not limited here.

[0107] Regarding the acquisition of the "CSI restored by the model that corresponds to the target CSI," there are several possibilities, such as (1) it being the output of the CSI restoration part of the AI / ML model, as shown in Figure 2 above; (2) it being acquired when the output of the CSI restoration part of the AI / ML model is sent back to the terminal via the network, as shown in Figure 3 above; or (3) it being acquired by restoring it using an alternative model on the UE side, as shown in Figure 4 above. Of course, other possibilities may also be included, and we are not limited to them here.

[0108] The information processing method according to the embodiment of this disclosure acquires at least two test sample points using a time-domain sliding window of first length, the test sample points being determined according to a target CSI and a CSI restored by the model corresponding to the target CSI, and the first device is a network device or terminal. This supports the accurate acquisition of test sample points, thereby improving the accuracy of the acquired model monitoring accuracy, and based on this, it further supports the accurate realization of control over the trigger of model monitoring reports according to the model monitoring accuracy (i.e., monitoring results), and as a result supports the accurate realization of decision operations such as model switching, fallback, update, activation and / or deactivation, and suitably solves the problem of not being able to accurately realize model decision operations that exist in information processing methods for model monitoring in related technologies.

[0109] Furthermore, the information processing method further includes preprocessing the at least two test sample points, the preprocessing of which includes at least one of weighting and anomaly exclusion (for example, excluding some anomaly sample points).

[0110] In this way, the accuracy of the test sample points used can be improved, further enhancing the accuracy of the model monitoring accuracy achieved.

[0111] Of these, the first length is determined according to the protocol's predefined and / or the network-side configuration, and / or the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or the network-side configuration.

[0112] In this way, the flexibility of realizing the configuration can be increased.

[0113] The information processing method according to the embodiments of this disclosure will be described below with reference to an example.

[0114] To address the above challenges, the embodiments of this disclosure provide an information processing method, which can be specifically implemented as an event-triggered information reporting model monitoring method. This method primarily addresses the dashed-line framed portion in Figures 2 to 4, solving the following problems: (1) how to acquire these K test sample point data within the time-series data (which may be understood as acquiring K intermediate KPI sample points); and (2) how to trigger an AI / ML model monitoring report based on the AI / ML model monitoring accuracy represented by the statistical results of the performance evaluation at these K sample points, and request the network side to make a decision regarding the activation, deactivation, update, fallback, and / or switching of the AI / ML model. Furthermore, it can also solve the problem of the potential ping-pong effect caused by frequent changes to the AI / ML model.

[0115] Furthermore, this embodiment is applicable to both network-side AI / ML model monitoring and UE-side AI / ML model monitoring, although there are some differences in specific details. For example, in the case of UE-side AI / ML model monitoring, the network side may configure several parameters for the UE side, whereas in the case of network-side AI / ML model monitoring, the above parameter configuration issue does not exist; however, it is necessary to consider the design of trigger signaling for the network side to trigger the execution of "UE-side target CSI reporting" by the UE. However, it may be considered that the network side's determination of when to trigger the UE's reporting falls under the specific implementation of the base station, and therefore it is not necessary to limit it to that. Accordingly, when describing this embodiment, we will describe it as much as possible using a general AI / ML model monitoring method (applicable to both network-side and UE-side), and further subdivide it according to AI / ML model monitoring on the network side and UE side, respectively.

[0116] This embodiment mainly includes the following: 1. The AI / ML model monitoring entity (which may belong to the network side or the UE side, corresponding to the above-mentioned first device being a network device or a terminal) establishes one time-domain sliding window with a length K (corresponding to the above-mentioned first length) for acquiring a test dataset of K test sample point data, and the monitoring accuracy (corresponding to the above-mentioned model monitoring accuracy) is based on the statistical results of these K test sample points (corresponding to the above-mentioned model monitoring accuracy being based on the statistical results of the at least two test sample points). Optionally, the configuration of the window length K is predefined by the protocol, configured by the network side, or a combination of protocol predefinition and network side configuration (corresponding to the above-mentioned first length being determined according to the protocol predefinition and / or the network side configuration). Optionally, for the sliding window (i.e., the time-domain sliding window), the configuration of its sliding step size Nstep in the time domain is predefined by the protocol, configured by the network side, or a combination of protocol predefinition and network side configuration (corresponding to the above-mentioned sliding step size of the time-domain sliding window being determined according to the protocol predefinition and / or the network side configuration). Optionally, the AI / ML model monitoring entity may perform preprocessing (including related operations for improving the accuracy of calculation results) on the K test sample points. For example, the preprocessing includes, but is not limited to, taking a subset containing K' (<K) test sample points to calculate the monitoring accuracy, or weighting the test sample points when calculating the monitoring accuracy based on the statistical results of the K test sample points (such as data cleaning), which corresponds to performing preprocessing on the above-mentioned at least two test sample points, wherein the preprocessing includes at least one of a weighting process and an anomaly elimination process.

[0117] 2. Define monitoring event-related information for the AI / ML model monitoring entity (network side or UE side). Monitoring event-related information may include the definition of the monitoring event, the joining conditions for the event, and the leaving conditions for the event. If the monitoring accuracy calculated based on a time-domain sliding window of length K satisfies the joining conditions, it is considered that the monitoring event has joined. If the monitoring accuracy calculated based on a time-domain sliding window of length K satisfies the leaving conditions, it is considered that the monitoring event has left. Selectively, the join and / or leave conditions include threshold parameters (corresponding to the threshold values ​​described above), the configuration of which these threshold parameters are determined by the protocol predefined, the network configuration, or a combination of the protocol predefined and the network configuration (corresponding to the fact that the threshold values ​​described above are determined according to the protocol predefined and / or the network configuration).

[0118] 3. Define the conditions for AI / ML model monitoring reports triggered by monitoring events from the AI / ML model monitoring entity (network side or UE side) (corresponding to the monitoring report conditions described above). Selectively, the configuration of the threshold parameters for AI / ML model monitoring reports triggered by monitoring events of the monitoring entity (corresponding to the reporting thresholds mentioned above) is determined by the protocol predefined, the network configuration, or a combination of the protocol predefined and the network configuration (corresponding to the fact that the monitoring reporting conditions mentioned above are determined according to the protocol predefined and / or the network configuration).

[0119] IV. The AI / ML model monitoring entity (network side or UE side) uses the monitoring accuracy calculated in a time-domain sliding window of length K and the defined monitoring events to determine the joining and leaving of monitoring events. Furthermore, it determines whether to trigger an AI / ML model monitoring report based on the determination status of the monitoring event and the conditions for AI / ML model monitoring reporting triggered by the monitoring event (corresponding to controlling the trigger operation of the model monitoring report according to the model monitoring accuracy, monitoring event-related information, and monitoring report conditions).

[0120] The present embodiment will be described in detail below, along with the flowcharts shown in Figures 2 and 3 above.

[0121] Method 1: Network-side AI / ML model monitoring based on target CSI reported from the UE side (corresponding to Figure 2 above, where the first device is a network device). Part 1: The network side establishes one sliding window of length K (where the length K refers to including K test sample points) for obtaining a test dataset of K test sample point data, and the monitoring accuracy is based on the statistical results of these K test sample points (corresponding to using the aforementioned time-domain sliding window of the first length to obtain the at least two test sample points, and the model monitoring accuracy being based on the statistical results of the at least two test sample points). The sliding window length K is predefined by the protocol, configured by the network side itself, or a combination of protocol predefined and network side self-configuration (corresponding to the aforementioned first length being determined according to the protocol predefined definition and / or the configuration on the network side). The configuration of the sliding step size Nstep of the sliding window is predefined by the protocol, configured by the network side itself, or a combination of protocol predefined and network side self-configuration (corresponding to the sliding step size of the aforementioned time-domain sliding window being determined according to the protocol predefined definition and / or the configuration on the network side). Optionally, the network side may perform preprocessing on the test data sample points in the sliding window, for example, the network side may discard some sample points that are considered inaccurate or non-representative (corresponding to the aforementioned abnormality exclusion processing), that is, select a subset with a cardinality of K' (K'<K) for subsequent calculation, or the network side may perform weighting on each data sample point (corresponding to the aforementioned weighting processing), so that its influence on the final statistical result can be changed.

[0122] Using Figure 2 as an example, one test sample point refers to one intermediate KPI calculated based on one CSI reconstructed by one AI / ML model (the CSI reconstructed by the AI / ML model is the output of the compressed CSI received on the network side and reported as feedback from the UE side, which has passed through the CSI reconstruction portion of the AI / ML model on the network side) and the corresponding target CSI reported by the UE side acquired on the network side (for example, the UE side's target CSI report may be based on the Physical Uplink Shared Channel (PUSCH), on a conventional Extended Type II Codebook CSI report (e.g., the configuration of Parameter Configuration (PC) 6 and / or PC8), on a similar Extended Type II Codebook CSI report (a more detailed codebook configuration), or on an SRS transmitted from the UE side). Specifically, when viewed on a time axis, the acquisition time of each intermediate KPI is a discrete point on the time axis, the reporting methods for compressed CSI and / or target CSI vary (periodic, semi-periodic, aperiodic), and the reporting policies are flexible and diverse (e.g., reporting a target CSI once each time a compressed CSI is feedback-reported, or reporting all target CSIs corresponding to the previous compressed CSI report simultaneously after multiple feedback-reported compressed CSIs). Therefore, the interval between any two intermediate KPI acquisition times can be any value greater than or equal to zero, and as shown in Figure 7 (Example of intermediate KPI acquisition times and sliding window on the time axis), we assume that the window length K=3 and sliding step size Nstep=3 (corresponding to the solid line frame in the figure), or the window length K=4 and sliding step size Nstep=2 (corresponding to the dashed line frame in the figure).It is important to explain here that the discrete points (i.e., intermediate KPI acquisition times) mentioned above are not necessarily at equal intervals, and one discrete point can correspond to multiple intermediate KPI acquisition times. For example, acquisition times 4-7 in Figure 7 all correspond to the fourth discrete point from left to right in the same figure (i.e., the intermediate KPI acquisition times for the 4th to 7th times are the same, corresponding to the situation described above where "target CSIs corresponding to the previous compressed CSI report are reported simultaneously," meaning that the 4th to 7th times are reported together after multiple CSIs have been collected, and the corresponding intermediate KPI acquisition times are the same).

[0123] In this embodiment, it is assumed that the compressed CSI feedback reporting method on the UE side and the target CSI reporting method on the UE side are both already determined and their correspondence is already clear. Considering the time delay when the AI / ML model monitoring entity calculates the intermediate KPI, the time of each intermediate KPI acquired on the time axis can be considered as = max[the time of acquisition of the CSI restored by the AI / ML model (the output obtained by the CSI restoration part of the AI / ML model in accordance with the compressed CSI reported as feedback from the UE side), the target CSI reporting time on the UE side corresponding to that (i.e., the time of acquisition of the CSI restored by the AI / ML model)] + the calculation time delay of the intermediate KPI, but is not limited to this.

[0124] Part 2: The network side defines AI / ML model monitoring events: Event1 (for example, when the network's AI / ML model monitoring accuracy is greater than a certain threshold, and the joining conditions are met, this event may be understood as occurring) and Event2 (for example, when the network's AI / ML model monitoring accuracy is less than a certain threshold).

[0125] (1) Define an event Event1 on the network side (the AI / ML model monitoring accuracy on the network side is greater than a certain threshold). The network should operate as follows. 1) When the following predetermined condition Condition1-1 is satisfied, it is considered that the joining condition for the event Event1 is satisfied, and the condition Condition1-1 is as follows. 2) When the following predetermined condition Condition1-2 is satisfied, it is considered that the leaving condition for the event Event1 is satisfied, and the condition Condition1-2 is as follows. The condition Condition1-1 (joining condition) is M NW -H ys >Thresh1, which corresponds to that the above joining condition is determined according to a hysteresis parameter and the corresponding threshold value. The condition Condition1-2 (leaving condition) is M NW +H ys <Thresh1, which corresponds to that the above leaving condition is determined according to a hysteresis parameter and the corresponding threshold value. The definitions of the variables in the above formulas are as follows. M NW refers to the AI / ML model monitoring accuracy calculated on the network side (corresponding to the aforementioned model monitoring accuracy), and the calculated value does not include any offset value. H ysrefers to the hysteresis parameter of event Event1, and the configuration of this parameter is predefined by the protocol, configured by the network side itself, or a mixture of the protocol predefined configuration and the network side configuration. Among them, the hysteresis parameter may be understood as a margin or a statistical confidence interval (for example, 3σ). When the difference between the observed value (corresponding to the aforementioned model monitoring accuracy) and the threshold is more significant than the hysteresis parameter, the determination can be made. By introducing the hysteresis parameter, the influence of observation disturbance on the determination can be reduced. Thresh1 refers to the threshold parameter of Event1, and the configuration of this parameter is predefined by the protocol, configured by the network side itself, or a mixture of the protocol predefined configuration and the network side's own configuration. Among them, M NW , H ys and Thresh1 have the same dimension.

[0126] At the same time, (2) define the network side event Event2 (that the AI / ML model monitoring accuracy on the network side is smaller than a certain threshold). The network should operate as follows. 1) When the following predetermined condition Condition2-1 is satisfied, it is considered that the joining condition for the event Event2 is satisfied, and the condition Condition2-1 is as follows. 2) When the following predetermined condition Condition2-2 is satisfied, it is considered that the leaving condition for the event Event2 is satisfied, and the condition Condition2-2 is as follows. The condition Condition2-1 (joining condition) is M NW +H ys ' <Thresh2, which corresponds to the fact that the above joining condition is determined according to the hysteresis parameter and the corresponding threshold. The condition Condition2-2 (leaving condition) is M NW -H ys The inequality '>Thresh2' may also be used, and this corresponds to the fact that the above exit condition is determined according to the hysteresis parameter and the corresponding threshold value. The definitions of the variables in the above formula are as follows: M NW This refers to the AI / ML model monitoring accuracy calculated on the network side, and this calculated value does not include any offset values. H ys ' refers to the hysteresis parameter of event Event2, and the configuration of this parameter is determined by the protocol's predefined settings, the network's own configuration, or a combination of the protocol's predefined settings and the network's own configuration. Thresh2 refers to the threshold parameter of Event2, and the configuration of this parameter is determined by the protocol's predefined settings, the network's own configuration, or a combination of the protocol's predefined settings and the network's own configuration. Eventually, M NW H ys ' and Thresh2 have the same dimension.

[0127] Part 3 defines the conditions under which AI / ML model monitoring events Event1 and / or Event2 trigger AI / ML model monitoring reports (corresponding to the monitoring report conditions described above). The conditions for triggering AI / ML model monitoring reports relate to the join and leave status of Event1 (and / or Event2) monitored on the network side, the occurrence threshold parameters of the monitoring event Event1 (and / or Event2) configured on the network side (e.g., maximum or minimum number of occurrences of the event, maximum or minimum number of consecutive occurrences of the event, etc.), and the timer parameters corresponding to the monitoring event Event1 (and / or Event2) (corresponding to the reporting time duration described above). The timer parameters are determined by the protocol predefined, the network side configuration, or a combination of the protocol predefined and the network side configuration.

[0128] The following are examples of embodiments using this method.

[0129] Implementation Example 1a: Network-side AI / ML model monitoring, where the AI / ML model M1 currently in use is deemed invalid based on the number of joins and / or exits of monitoring events within one or more sliding windows exceeding the threshold (AI / ML model M1 here may refer to a pair of models for UE-side CSI generation and network-side CSI restoration). On the network side, the maximum number of occurrences N of monitoring event Event2 for AI / ML model M1 is... max_count_E2 ) (corresponding to the reporting threshold mentioned above) is defined, and the configuration of this parameter is determined by the protocol's predefined settings, by the network's own configuration, or by a mixture of the protocol's predefined settings and the network's own configuration. The network side defines the above N max_count_E2 An invalid timer associated with (corresponding to the reported timing time length mentioned above) is defined, and the configuration of this parameter is determined by the protocol's predefined settings, by the network's own configuration, or by a combination of the protocol's predefined settings and the network's own configuration. Of these, Nmax_count_E2 The configuration may be such that = 1, in which case whether a monitoring event triggers an AI / ML model monitoring report may be determined by the network-side monitoring accuracy calculated with K data sample points within a single sliding window, and the Expiration timer, i.e., an expired counter that statistically counts the number of occurrences of monitoring event Event2, resets to zero and restarts counting when the timer returns to zero (countdown) or reaches its upper limit (countup).

[0130] The network side monitors the number of occurrences of Event2 in the AI / ML model M1, and of those, M NW This is calculated according to the sliding window described in the preceding paragraph. When the network monitors that the join condition (Condition2-1) of Event2 has been met, based on the calculation in the current sliding window (i.e., the time-domain sliding window), the network records the current occurrence count of Event2 as +1. When the network monitors that the exit condition (Condition2-2) of Event2 has been met, based on the calculation in the current sliding window, the network records the current occurrence count of Event2 as -1. The current occurrence count of Event2 recorded by the network, along with whether the invalid timer has returned to zero (countdown) or has not reached its upper limit (countup), equals the maximum occurrence count N of the monitored event Event2. max_count_E2If it becomes greater than the threshold, the network triggers an AI / ML model monitoring report (which may be a report between internal modules on the network side, for example, a report from a monitoring module to a decision module) reporting that the currently used AI / ML model M1 is invalid (the above-mentioned monitoring event is that the model monitoring accuracy does not meet the usage conditions, and the first number of times the model monitoring accuracy meets the join conditions of the monitoring event within the reporting time length is greater than the reporting threshold, in which case the operation of the model monitoring report is triggered, the content of which is to indicate that the currently used model is invalid, and the first number is determined according to the real-time number of times the model monitoring accuracy meets the join conditions of the monitoring event and the real-time number of times the model monitoring accuracy meets the exit conditions of the monitoring event).

[0131] Implementation Example 1b: Network-side AI / ML model monitoring. Based on the fact that the number of subscriptions to monitoring events within one or more sliding windows exceeds the threshold, it is determined that the currently used AI / ML model M1 is invalid.

[0132] This is similar to implementation example 1a, but differs in that it does not consider the occurrence of the exit condition (Condition2-2) for Event2, and only considers the following: The network side calculates the number of occurrences or consecutive occurrences of the join condition (Condition2-1) for Event2 according to the current sliding window, and the maximum number of occurrences N for monitored event Event2. max_count_E2 Alternatively, the maximum number of consecutive occurrences of monitoring event Event2 is N. max_consec_count_E2When it is monitored that the threshold is greater than the threshold, the network triggers an AI / ML model monitoring report, reporting that the currently used AI / ML model M1 is invalid, and the operation of the model monitoring report is triggered when the aforementioned monitoring event is that the model monitoring accuracy does not meet the usage conditions and the first number of times the model monitoring accuracy has met the subscription conditions for the monitoring event within the reporting time duration is greater than the reporting threshold, the content of the report is to indicate that the currently used model is invalid, the first number is determined according to the real-time number of times the model monitoring accuracy has met the subscription conditions for the monitoring event, or the first number refers to the number of consecutive times the model monitoring accuracy has met the subscription conditions for the monitoring event.

[0133] Implementation Example 1c (can be used in combination with Implementation Example 1a or Implementation Example 1b): Network-side AI / ML model monitoring. Based on the number of times the monitoring event has joined one or more sliding windows exceeding the threshold, the currently used AI / ML model M1 is determined to be invalid, and the count of consecutive multiple joins to monitoring event Event2 is reset to zero.

[0134] This case is similar to Case 1a or Case 1b, but differs in the following respects: The network side, according to the monitoring accuracy calculated in the sliding window, N max_consec_count_E2out When the network detects that the conditions for escaping Event2 (Condition2-2) have been met for a certain number of consecutive times, it resets the recorded number of current Event2 occurrences to zero and either resets the invalid timer to its upper limit (countdown) or to zero (countup). Of these, the number of times the conditions for escaping Event2 have been met consecutively is N. max_consec_count_E2out This can be due to a predefined protocol, a configuration on the network side, or a combination of a predefined protocol and a configuration on the network side.

[0135] Of these, the above implementation examples 1a to 1c all involve AI / ML model monitoring on the network side, representing several possible forms of monitoring reporting based on event triggers.

[0136] Implementation Example 2: Network-side AI / ML model monitoring. Based on the number of joins and / or exits of monitoring events within one or more sliding windows exceeding the threshold, it is decided to switch the currently used network-side CSI recovery AI / ML model.

[0137] The network side monitors the currently used network-side CSI reconstruction AI / ML model C1, while simultaneously monitoring the unused potential network-side CSI reconstruction AI / ML model C2. (Since the network side also possesses CSI reconstruction model C2, it is possible to obtain a reconstruction CSI based on model C2 from the compressed CSI reported as feedback from the received UE side, and then calculate the intermediate KPI using that and the target CSI reported by the UE side.) It should be noted here that we are using one potential CSI reconstruction model C2 (corresponding to the candidate model mentioned above) as an example, but of course, this is also applicable to multiple potential CSI reconstruction models C3, C4, ..., CN, and is not limited to them here.

[0138] The network configures the relevant parameters for Event1 and / or Event2 for C1 and C2, respectively. These parameters may be configured according to the protocol's predefined settings, the network's own configuration, or a combination of both. The network also determines the maximum number of occurrences of the monitoring event Event2 for C1, N. max_count_E2 The network side defines the minimum number of occurrences N of the monitoring event Event1 for C2. min_count_E1The above parameters are defined and their configuration is determined by the protocol's predefined settings, the network's own configuration, or a combination of the protocol's predefined settings and the network's own configuration. The network side is defined as above N max_count_E2 and N min_count_E1 An invalid timer (corresponding to the reporting time length mentioned above) associated with (corresponding to the reporting threshold mentioned above) is defined, and the configuration of the above parameters is determined by the protocol pre-definition, by the network's own configuration, or by a mixture of the protocol pre-definition and the network's own configuration. Of these, N max_count_E2 It may be configured such that = 1, in which case whether or not to trigger an AI / ML model monitoring report is determined by K data sample points within a single sliding window, and N min_count_E1 It may also be configured such that = 1, in which case K data sample points within a single sliding window determine whether or not to trigger an AI / ML model monitoring report.

[0139] The network side monitors the number of occurrences of Event2 related to C1, and of those, M NW This is calculated according to the sliding window described in the preceding paragraph. When the network monitors the join condition (Condition2-1) of Event2 according to the calculation in the current sliding window, the network records it as the current occurrence count of Event2 + 1. When the network monitors the exit condition (Condition2-2) of Event2 according to the calculation in the current sliding window, the network records it as the current occurrence count of Event2 - 1.

[0140] The network side monitors the number of occurrences of Event1 related to C2, and of those, M NWThis is calculated according to the sliding window described in the preceding paragraph. When the network monitors the join condition (Condition1-1) of Event1 according to the calculation in the current sliding window, the network records it as the current occurrence count of Event1 + 1. When the network monitors the exit condition (Condition1-2) of Event1 according to the calculation in the current sliding window, the network records it as the current occurrence count of Event1 - 1.

[0141] The invalid timer configured for C1 has not returned to zero (countdown) or reached its upper limit (countup), and the number of occurrences of Event2 currently recorded on the network side for C1 is equal to the N configured for it. max_count_E2 When the value is greater than the value, and the invalid timer configured on C2 has not returned to zero (countdown) or reached its upper limit (countup), and the number of occurrences of Event1 currently recorded on the network side for C2 is the value configured on it. min_count_E1When the value is greater than the specified value, the network triggers an AI / ML model monitoring report, reporting that the performance of the currently used CSI recovery AI / ML model C1 is worse than that of the CSI recovery AI / ML model C2, and reporting C2 as a candidate model for switching. If the first monitoring event described above is that the usage conditions are not met by the monitoring accuracy of the first model, and the first condition is met, and the second monitoring event is that the usage conditions are met by the monitoring accuracy of either candidate model's second model, and the second condition is met, then the network triggers a model monitoring report operation, the content of which is to instruct a model switch and carries information about the candidate model, of which the first condition is that the first model monitoring accuracy is within the first report timing period. The second condition includes the third number of times the enrollment condition for the first monitoring event is met being greater than the first reporting threshold, the third number being determined in real time by the number of times the first model monitoring accuracy met the enrollment condition for the first monitoring event and the number of times the first model monitoring accuracy met the withdrawal condition for the first monitoring event; the second condition includes the fourth number of times the second model monitoring accuracy met the enrollment condition for the second monitoring event within the second reporting time length being greater than the second reporting threshold, the fourth number being determined in real time by the number of times the second model monitoring accuracy met the enrollment condition for the second monitoring event and the number of times the second model monitoring accuracy met the withdrawal condition for the second monitoring event.

[0142] Of these, the above implementation example 2 may also have variations similar to those in implementation examples 1a to 1c. That is, a trigger for monitoring reports may be defined by using a combination of conditions such as the joining and / or leaving conditions for different events, the maximum or minimum (consecutive) occurrence count of an event, or the initialization or zeroing of a counter, but this will not be repeated here.

[0143] Method 2: AI / ML model monitoring on the UE side, based on the restored CSI sent back from the network side (corresponding to Figure 3 above, where the first device is the terminal). Part 1: The UE establishes a time-domain sliding window of length K (where length K means that it contains K test sample points) to obtain a test dataset of K test sample point data, and the monitoring accuracy is based on the statistical results of these K test sample points (this corresponds to obtaining at least two of the test sample points using the time-domain sliding window of the above first length, and the model monitoring accuracy being based on the statistical results of the at least two test sample points). The sliding window length K is determined by a protocol predefined, by the network configuration (e.g., configured based on RRC), or by a combination of a protocol predefined and a network configuration (this corresponds to the above first length being determined according to the protocol predefined and / or the network configuration). The configuration of the sliding step size Nstep in the time domain of the sliding window is determined by the protocol predefined, by the network configuration (e.g., configured based on RRC), or by a combination of the protocol predefined and the network configuration (corresponding to the fact that the sliding step size of the time domain sliding window is determined according to the protocol predefined and / or the network configuration).

[0144] Using Figure 3 as an example, one test sample point refers to one intermediate KPI calculated based on one target CSI and the reconstructed CSI returned from the network side acquired on the corresponding UE side (for example, the reconstructed CSI return from the network side may be based on a Physical Downlink Shared Channel (PDSCH) or a pre-coded Channel State Information Reference Signal (CSI-RS), etc.). The examples of the intermediate KPI acquisition time, sliding window length K, and sliding step size Nstep of the sliding window are similar to those in Method 1 and will not be repeated here.

[0145] Part 2: The UE side defines AI / ML model monitoring events: Event3 (for example, when the UE's AI / ML model monitoring accuracy is greater than a certain threshold) and Event4 (for example, when the UE's AI / ML model monitoring accuracy is less than a certain threshold).

[0146] (1) Define Event3 on the UE side (that the AI / ML model monitoring accuracy on the UE side is greater than a certain threshold). The UE should do the following: 1) If the following specified condition, Condition3-1, is met, it will be considered that the conditions for joining Event3 have been met. Condition3-1 is as follows: 2) If the following predetermined condition, Condition3-2, is met, it will be considered that the exit condition for Event3 has been met, and Condition3-2 is as follows: Condition 3-1 (Membership Conditions) is: M UE -H ys''>which may be an inequality of Thresh3, corresponding to that the above access condition is determined according to the hysteresis parameter and the corresponding threshold threshold value. The condition Condition3-2 (departure condition) is M UE +H ys ''<which may be an inequality of Thresh3, corresponding to that the above departure condition is determined according to the hysteresis parameter and the corresponding threshold threshold value. Among them, the definitions of the variables in the above formula are as follows. M UE refers to the AI / ML model monitoring accuracy calculated on the UE side (corresponding to the aforementioned model monitoring accuracy), and the calculated value does not include any offset value. H ys ''refers to the hysteresis parameter of event Event3, and the configuration of the parameter is pre-defined by the protocol, configured by the network side (for example, configured based on RRC), or a mixture of the protocol pre-definition and the network side configuration. Thresh3 refers to the threshold parameter of Event3, and the configuration of the parameter is pre-defined by the protocol, configured by the network side (for example, configured based on Radio Resource Control (RRC)), or a mixture of the protocol pre-definition and the network side configuration. Among them, M UE , H ys '' and Thresh3 have the same dimension.

[0147] At the same time, (2) define the event Event4 on the UE side (the AI / ML model monitoring accuracy on the UE side is smaller than a certain threshold). The UE shall operate as follows. 1) When the following predetermined condition Condition4-1 is satisfied, it is considered that the access condition of the event Event4 is satisfied, and the condition Condition4-1 is as follows. 2) When the following predetermined condition Condition4-2 is satisfied, it is considered that the exit condition of the event Event4 is satisfied, and the condition Condition4-2 is as follows. Condition Condition4-1 (entry condition) is M UE +H ys '''<Thresh4, which may be an inequality, corresponding to that the entry condition is determined according to the hysteresis parameter and the corresponding threshold. Condition Condition4-2 (exit condition) is M UE -H ys '''>Thresh4, which may be an inequality, corresponding to that the exit condition is determined according to the hysteresis parameter and the corresponding threshold. The definitions of the variables in the above formula are as follows. M UE refers to the AI / ML model monitoring accuracy calculated on the UE side, and the calculated value does not include any offset value. H ys ''' refers to the hysteresis parameter of event Event4, and the configuration of the parameter is predefined by the protocol, configured by the network side (for example, configured based on RRC), or a combination of protocol predefinition and network side configuration. Thresh4 refers to the threshold parameter of Event4, and the configuration of the parameter is predefined by the protocol, configured by the network side (for example, configured based on RRC), or a combination of protocol predefinition and network side configuration. Wherein M UE , H ys ''' and Thresh4 have the same dimension.

[0148] Part 3 defines the conditions under which AI / ML model monitoring events Event3 and / or Event4 trigger AI / ML model monitoring reports (corresponding to the monitoring report conditions described above). The conditions for triggering AI / ML model monitoring reports relate to the join and leave status of Event3 (and / or Event4) monitored on the UE side, the occurrence threshold parameters of the monitoring event Event3 (and / or Event4) configured on the network side (e.g., maximum or minimum number of occurrences of the event, maximum or minimum number of consecutive occurrences of the event, etc.), and the timer parameters corresponding to the monitoring event Event3 (and / or Event4) (corresponding to the reporting time duration described above).

[0149] The following are examples of embodiments using this method.

[0150] Implementation Example 3: Monitoring of the AI / ML model on the UE side (based on the restored CSI returned from the network side), and determining that the currently used AI / ML model M1 is invalid (here, AI / ML model M1 may refer to a pair of models for generating the UE side CSI and restoring the network side CSI). On the UE side, the maximum number of occurrences N of monitoring event 4 for AI / ML model M1. max_count_E4 (corresponding to the reporting threshold mentioned above) is defined, and the configuration of this parameter is determined by the protocol pre-definition, the network-side configuration (e.g., configured based on RRC), or a combination of the protocol pre-definition and the network-side configuration. The UE side is as described above N max_count_E4 An invalid timer associated with it (corresponding to the reported timing duration described above) is defined, and the configuration of this parameter is determined by the protocol predefined, by the network configuration (e.g., configured based on RRC), or by a combination of the protocol predefined and the network configuration.

[0151] The UE side monitors the number of occurrences of Event4 in the AI / ML model M1, and of those, M UEThis is calculated according to the sliding window described in the preceding paragraph. When the UE monitors that the join condition (Condition4-1) of Event4 has been met, based on the calculation in the current sliding window (i.e., the time-domain sliding window), the UE records the current occurrence count of Event4 as +1. When the UE monitors that the exit condition (Condition4-2) of Event4 has been met, based on the calculation in the current sliding window, the UE records the current occurrence count of Event4 as -1. The current occurrence count of Event4 recorded by the UE, along with whether the invalid timer has returned to zero (countdown) or not reached its upper limit (countup), equals the maximum occurrence count N of the monitored event Event4. max_count_E4 If it is greater than the threshold, the UE triggers an AI / ML model monitoring report, reporting that the currently used AI / ML model M1 is invalid and requesting the network to make a decision (the above-mentioned monitoring event is that the model monitoring accuracy does not meet the usage conditions, and the first number of times the model monitoring accuracy meets the join conditions for the monitoring event within the reporting time duration is greater than the reporting threshold, in which case the operation of the model monitoring report is triggered, the content of which is to indicate that the currently used model is invalid, and the first number is determined according to the real-time number of times the model monitoring accuracy meets the join conditions for the monitoring event and the real-time number of times the model monitoring accuracy meets the exit conditions for the monitoring event).

[0152] Of these, the above implementation example 3 may also have variations similar to those in implementation examples 1a to 1c. That is, a trigger for monitoring reports may be defined by using a combination of conditions such as the joining and / or leaving conditions for different events, the maximum or minimum (consecutive) occurrence count of an event, or the initialization or zeroing of a counter, but this will not be repeated here.

[0153] Method 3: UE-side AI / ML model monitoring, based on an alternative model of the AI / ML model for the network-side CSI restoration portion that exists on the UE side (corresponding to Figure 4 above, where the first device is the terminal). Part 1: The UE establishes a time-domain sliding window of length K (where length K means that it contains K test sample points) to obtain a test dataset of K test sample point data, and the monitoring accuracy is based on the statistical results of these K test sample points (this corresponds to obtaining at least two of the test sample points using the time-domain sliding window of the above first length, and the model monitoring accuracy being based on the statistical results of the at least two test sample points). The sliding window length K is determined by a protocol predefined, by the network configuration (e.g., configured based on RRC), or by a combination of a protocol predefined and a network configuration (this corresponds to the above first length being determined according to the protocol predefined and / or the network configuration). The configuration of the sliding step size Nstep in the time domain of the sliding window is determined by the protocol predefined, by the network configuration (e.g., configured based on RRC), or by a combination of the protocol predefined and the network configuration (corresponding to the fact that the sliding step size of the time domain sliding window is determined according to the protocol predefined and / or the network configuration).

[0154] Taking Figure 4 as an example, one test sample point refers to one intermediate KPI calculated based on one target CSI and the corresponding CSI reconstructed using an alternative model on the UE side. The examples of the intermediate KPI acquisition time, sliding window length K, and sliding step size Nstep of the sliding window are similar to those in Method 1 and will not be repeated here.

[0155] Part 2: The UE side defines AI / ML model monitoring events: Event 3 (for example, when the UE's AI / ML model monitoring accuracy is greater than a certain threshold) and Event 4 (for example, when the UE's AI / ML model monitoring accuracy is less than a certain threshold). The definitions of Event 3 and Event 4 are the same as those in Part 2 of Method Two and will not be repeated here, however, the values ​​of the configured parameters may differ.

[0156] Part 3 defines the conditions under which AI / ML model monitoring events Event3 and / or Event4 trigger AI / ML model monitoring reports (corresponding to the monitoring report conditions described above). The conditions for triggering AI / ML model monitoring reports relate to the join and leave status of Event3 (and / or Event4) monitored on the UE side, the occurrence threshold parameters of the monitoring event Event3 (and / or Event4) configured on the network side (e.g., maximum or minimum number of occurrences of the event, maximum or minimum number of consecutive occurrences of the event, etc.), and the timer parameters corresponding to the monitoring event Event3 (and / or Event4) (corresponding to the reporting time duration described above).

[0157] Implementation Example 4: Monitoring of the AI / ML model on the UE side (based on an alternative model of the AI / ML model for the network-side CSI restoration portion that exists on the UE side), and determining that the currently used AI / ML model M1 is invalid (here, AI / ML model M1 may refer to a pair of models for UE-side CSI generation and UE-side CSI restoration, i.e., the UE-side AI / ML model and an alternative model of the AI / ML model for the network-side CSI restoration portion that exists on the UE side). Of these, (1) the relevant parts of this implementation example are the same as those in Implementation Example 3 and will not be repeated here.

[0158] (2) The above implementation example 4 may also have variations similar to those in implementation examples 1a to 1c. That is, a trigger for monitoring reports may be defined by using a combination of conditions such as the joining and / or leaving conditions for different events, the maximum or minimum (consecutive) occurrence count of an event, and the initialization or zeroing of a counter, but this will not be repeated here.

[0159] Implementation Example 5a: Monitoring of the AI / ML model on the UE side (based on an alternative model of the AI / ML model for the network-side CSI restoration part that exists on the UE side), and a decision is made to switch the AI / ML model currently used for generating the UE-side CSI. The UE side monitors the currently used UE-side CSI generation AI / ML model D1, while simultaneously monitoring an unused, potential UE-side CSI generation AI / ML model D2. (Since the UE side simultaneously possesses CSI generation model D2, it is possible to obtain a compressed CSI based on D2 from the target CSI via model D2, and then obtain a new CSI from the compressed CSI of D2 via an alternative model of the CSI reconstruction part AI / ML model, and then calculate the intermediate KPI with that and the target CSI.) It should be noted that here we are using one potential CSI generation model D2 (corresponding to the candidate model mentioned above) as an example, but of course, this is also applicable to cases with multiple potential CSI generation models D3, D4, ..., DN.

[0160] The UE configures the relevant parameters for Event3 and / or Event4 for D1 and D2, respectively. These parameters are configured based on protocol pre-definitions, network configuration (e.g., based on RRC), or a combination of protocol pre-definitions and network configuration. The UE also determines the maximum number of occurrences of monitoring event Event4 for D1, N. max_count_E4 The UE defines the minimum number of occurrences N of the monitoring event Event3 for D2. min_count_E3The above parameters are defined and configured by the protocol predefined, by the network-side configuration (e.g., configured based on RRC), or by a combination of the protocol predefined and the network-side configuration. The UE side is as follows: max_count_E4 and N min_count_E3 An invalid timer (corresponding to the reporting time length described above) is defined associated with (corresponding to the reporting threshold described above), and the configuration of the above parameters is determined by the protocol pre-definition, by the network configuration (e.g., configured based on RRC), or by a combination of the protocol pre-definition and the network configuration.

[0161] The UE side monitors the number of occurrences of Event4 related to D1, and of those, M NW This is calculated according to the sliding window described in the preceding paragraph. When the UE monitors the joining condition (Condition4-1) of Event4 according to the calculation in the current sliding window, the UE records the current occurrence count of Event4 + 1. When the UE monitors the leaving condition (Condition4-2) of Event4 according to the calculation in the current sliding window, the UE records the current occurrence count of Event4 - 1.

[0162] The UE side monitors the number of occurrences of Event3 related to D2, and of those, M NW This is calculated according to the sliding window described in the preceding paragraph. When the UE monitors the joining condition (Condition3-1) of Event3 according to the calculation in the current sliding window, the UE records the current occurrence count of Event3 as +1. When the UE monitors the leaving condition (Condition3-2) of Event3 according to the calculation in the current sliding window, the UE records the current occurrence count of Event3 as -1.

[0163] The invalid timer configured in D1 has not returned to zero (countdown) or has not reached its upper limit (countup), and the number of occurrences of Event4 in D1 recorded on the UE side is equal to the N configured in it. max_count_E4 When the value is greater than the value, and the invalid timer configured in D2 has not returned to zero (countdown) or reached its upper limit (countup), and the number of occurrences of the current Event3 in D2 recorded on the network side is the value configured in N. min_count_E3 When the value is greater than the specified value, the UE triggers an AI / ML model monitoring report, reporting that the performance of the currently used AI / ML model D1 for CSI generation is worse than that of the AI / ML model D2 for CSI generation, reporting D2 as a candidate model for switching, and requesting the network to make a decision. If the first monitoring event described above is that the monitoring accuracy of the first model does not meet the usage conditions and the first condition is met, and the second monitoring event is that the monitoring accuracy of the second model of any candidate model meets the usage conditions and the second condition is met, the UE triggers a model monitoring report operation, the content of which is to instruct a model switch and carries information about the candidate model, of which the first condition is within the first reporting time length. The second condition includes the third number of times the model monitoring accuracy satisfies the enrollment condition for the first monitoring event being greater than the first reporting threshold, wherein the third number is determined by the real-time number of times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event and the real-time number of times the first model monitoring accuracy satisfies the withdrawal condition for the first monitoring event; the second condition includes the fourth number of times the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event being greater than the second reporting threshold within the second reporting time length, wherein the fourth number is determined by the real-time number of times the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event and the real-time number of times the second model monitoring accuracy satisfies the withdrawal condition for the second monitoring event.

[0164] Of these, the above implementation example 5a may also have variations similar to those in implementation examples 1a to 1c. That is, a trigger for monitoring reports may be defined by using a combination of conditions such as the joining and / or leaving conditions for different events, the maximum or minimum (consecutive) occurrence count of an event, or the initialization or zeroing of a counter, but this will not be repeated here.

[0165] Implementation Example 5b: Monitoring of the AI / ML model on the UE side (based on an alternative model of the AI / ML model for the network-side CSI restoration portion that exists on the UE side), and based on joining events Event3 and Event4, it is decided to switch the AI / ML model currently used for generating the UE side CSI.

[0166] This implementation example is a special case of Implementation Example 5a, and is D1's N max_count_E4 and N of D2 min_count_E3 Configure both to be 1, and set the threshold for joining Event4 of D1. E4in And the threshold for joining Event 3 in D2 E3in They can be formed together, and in this case, the combination event "satisfies the joining conditions of Event4 and Event3" is "the monitoring accuracy of D2 is greater than the monitoring accuracy of D1 X=thresh E3in -thresh E4in This is equivalent to an event that is "only X greater". Using the above configuration, it is possible to set it up so that if a potential AI / ML model is X better than the current model monitoring accuracy, a monitoring report is triggered, reporting the monitoring accuracy of candidate model D2 and / or each model, and requesting the network to make a decision (the aforementioned monitoring event corresponds to the fact that the model monitoring accuracy of the currently used model is less than the model monitoring accuracy of any candidate model, which triggers the model monitoring report operation, of which the content of the report is to instruct a model switch and carries the information of the candidate model).

[0167] It should be noted here that the relevant details of the examples described above are mutually relatable, and any overlapping parts will not be repeated. Furthermore, the numerical settings for the thresholds and / or threshold values ​​mentioned above may be the same or different, and are not limited to those settings.

[0168] Based on the above, the embodiments of the present disclosure have the following advantages. 1. With AI / ML model monitoring based on a sliding window, the monitoring entity can continuously monitor the AI / ML model over time, and there is no need for the network side to trigger the initialization of AI / ML model monitoring via signaling (Media Access Control-Control Element (MAC CE) or Downlink Control Information (DCI)), thus saving signaling overhead. In particular, when the monitoring entity is on the UE side, one preferred example of this embodiment is that the network side initiates RRC configuration signaling to the UE to configure the relevant parameters of the sliding window (monitoring configuration), and when the UE receives the RRC configuration, it starts persistent, continuous AI / ML model monitoring based on a sliding window. 2. AI / ML model monitoring based on sliding windows allows for relatively flexible calculations, and different sliding window sizes and sliding window step sizes can be set for different UE capabilities (e.g., computing power, monitoring accuracy requirements, monitoring immediacy requirements, etc.). This prevents sliding window calculations from becoming too frequent, or makes the changes in monitoring accuracy calculated after sliding window processing more significant. In other words, the window length and sliding window step size can be flexibly configured and jointly optimized according to the demand and computing power of the terminal. 3. By defining monitoring event-related information, the AI / ML model monitoring entity can determine triggers for AI / ML model monitoring reports based on the joining and leaving of monitoring events, and threshold parameters related to monitoring events (e.g., maximum or minimum number of occurrences of monitoring events, maximum or minimum number of consecutive occurrences of monitoring events, etc.). This effectively reduces errors and misjudgments and resolves the ping-pong effect caused by frequent switching of AI / ML model monitoring.

[0169] The embodiments of this disclosure further provide an information processing device referred to as the first device, as shown in Figure 8, which includes a memory 81, a transceiver 82, and a processor 83. Memory 81 is for storing computer programs, and transceiver 82 is for sending and receiving data under the control of processor 83, and processor 83 reads the computer program in memory 81, The method involves obtaining model monitoring accuracy based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI restored by the model that corresponds to the target CSI, and the model monitoring accuracy is based on the statistical results of the at least two test sample points. To determine the monitoring event-related information for the aforementioned model monitoring accuracy, which includes the definition of the monitoring event and the joining and / or leaving conditions of the monitoring event, This is for controlling the trigger operation of the model monitoring report according to the model monitoring accuracy and monitoring event-related information, Of these, the first device is a network device or terminal.

[0170] The information processing device according to the embodiment of this disclosure acquires model monitoring accuracy in accordance with at least two test sample points, the test sample points are determined in accordance with target channel status information (CSI) and a CSI restored by the model that corresponds to the target CSI, the model monitoring accuracy is based on statistical results of the at least two test sample points, and monitoring event-related information for the model monitoring accuracy is determined, which includes the definition of a monitoring event and the join and / or leave conditions of the monitoring event, and the trigger operation of the model monitoring report is controlled in accordance with the model monitoring accuracy and the monitoring event-related information, the first device being a network device or terminal. This supports the realization of control over the trigger of the model monitoring report in accordance with the model monitoring accuracy (i.e., monitoring results), thereby supporting the accurate realization of decision operations such as model switching, fallback, update, activation and / or deactivation, and suitably solves the problem of not being able to accurately realize model decision operations that exist in the information processing methods for model monitoring in related technologies.

[0171] Specifically, the transceiver 82 is for receiving and transmitting data under the control of the processor 83.

[0172] In Figure 8, the bus architecture may include an arbitrary number of interconnected buses and bridges, specifically connecting various circuits between one or more processors, represented by processor 83, and memory, represented by memory 81. The bus architecture can also connect various other circuits, such as peripherals, voltage regulators, and power management circuits, but these are known in the art and will not be further described herein. The bus interface provides the interface. The transceiver 82 may consist of multiple elements, including a transmitter and a receiver, and provides a means for communicating with various other devices over a transmission medium, which may include wireless channels, wired channels, optical cables, etc. Processor 83 is responsible for managing the bus architecture and general processing, and memory 81 can store data used when operations are performed by processor 83.

[0173] The processor 83 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a complex programmable logic device (CPLD), and the processor may employ a multi-core architecture.

[0174] The processor is for executing any of the methods described in the embodiments of this disclosure by calling a computer program stored in memory, in accordance with the executable instructions obtained. The processor and memory may be physically separated.

[0175] It should be explained here that when the first device is implemented as a terminal, it may further include a bus interface and a user interface capable of communicating with it. For different user devices, the user interface may be an interface that can connect necessary devices externally or internally, and the connected devices include, but are not limited to, a small keyboard, display, speaker, microphone, joystick, etc.

[0176] In addition, controlling the trigger operation of the model monitoring report according to the aforementioned model monitoring accuracy and monitoring event-related information includes triggering the model monitoring report operation if the monitoring event is such that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the joining conditions for the monitoring event, or triggering the model monitoring report operation if the monitoring event is such that the model monitoring accuracy meets the usage conditions and the model monitoring accuracy meets the leaving conditions for the monitoring event.

[0177] Furthermore, the operation further includes determining monitoring reporting conditions which include a reporting threshold and / or reporting time length, and triggering a model monitoring report operation when the monitoring event described above is that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the enrollment conditions for the monitoring event, (1) triggering a model monitoring report operation when the monitoring event is that the model monitoring accuracy does not meet the usage conditions and the first number of times the model monitoring accuracy has met the enrollment conditions for the monitoring event within the reporting time length is greater than the reporting threshold, wherein the content of the report is to indicate that the currently used model is invalid, and the first number is determined according to the number of real-time times the model monitoring accuracy has met the enrollment conditions for the monitoring event, or the first number is determined according to the number of real-time times the model monitoring accuracy has met the enrollment conditions for the monitoring event and the number of real-time times the model monitoring accuracy has met the disengagement conditions for the monitoring event, or the first The number of times includes the number of consecutive times the model monitoring accuracy satisfies the enrollment condition of the monitoring event, and / or (2) if the monitoring event described above is one in which the model monitoring accuracy satisfies the usage condition and the model monitoring accuracy satisfies the exit condition of the monitoring event, the operation of the model monitoring report is triggered if the monitoring event is one in which the model monitoring accuracy satisfies the usage condition and the second number of times the model monitoring accuracy satisfies the exit condition of the monitoring event within the reporting time length is greater than the reporting threshold, wherein the content of the report is to indicate that the currently used model is invalid, and the second number is determined according to the number of real-time times the model monitoring accuracy satisfies the exit condition of the monitoring event, or the second number is determined according to the number of real-time times the model monitoring accuracy satisfies the exit condition of the monitoring event and the number of real-time times the model monitoring accuracy satisfies the enrollment condition of the monitoring event, or the second number isThis includes the fact that the model monitoring accuracy refers to the number of consecutive times the exit condition for the monitoring event is met.

[0178] In embodiments of the present disclosure, the model includes a currently used model and at least one candidate model, and obtaining model monitoring accuracy in accordance with at least two test sample points as described above includes obtaining a first model monitoring accuracy for the currently used model in accordance with at least two test sample points corresponding to the currently used model, and obtaining a second model monitoring accuracy for the candidate model in accordance with at least two test sample points corresponding to the candidate model.

[0179] Of these, the monitoring event includes a first monitoring event for the currently used model and a second monitoring event for the candidate model; the reporting threshold includes a first reporting threshold for the currently used model and a second reporting threshold for the candidate model; and the reporting timing length includes a first reporting timing length for the currently used model and a second reporting timing length for the candidate model.

[0180] In the embodiments of this disclosure, the operation of a model monitoring report is triggered when the above-described monitoring event is that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the subscription conditions for the monitoring event, in which case: (1) When the first monitoring event is that the first model monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second model monitoring accuracy of any candidate model meets the usage conditions and the second condition is met, the operation of a model monitoring report is triggered, and the content of the report is to instruct a model switch and carries information about the candidate model, or (2) When the first monitoring event is that the first model monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second model monitoring accuracy of any candidate model does not meet the usage conditions and the third condition is met, the operation of a model monitoring report is triggered, and the content of the report is to instruct a model switch and carries information about the candidate model The report includes carrying a report, of which (1) the first condition includes that the third number of times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event within the first reporting time length is greater than the first reporting threshold, the third number being determined according to the real-time number of times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event, or the third number being determined according to the real-time number of times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event and the real-time number of times the first model monitoring accuracy satisfies the withdrawal condition for the first monitoring event, or the third number being the number of consecutive times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event, (2) the second condition includes that the fourth number of times the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, the fourth number being determined according to the real-time number of times the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the fourth number being,The second model monitoring accuracy is determined according to the number of real-time instances in which the second monitoring event's enrollment condition is met and the number of real-time instances in which the second model monitoring accuracy is met and the number of real-time instances in which the second monitoring event's withdrawal condition is met, or the fourth instance refers to the number of consecutive instances in which the second model monitoring accuracy is met and the enrollment condition is met for the second monitoring event, and (3) the third condition refers to the fifth instance in which the second model monitoring accuracy is met and the withdrawal condition is met for the second monitoring event within the second reporting time length being greater than the second reporting threshold, and the fifth instance is determined according to the number of real-time instances in which the second model monitoring accuracy is met and the number of real-time instances in which the second model monitoring accuracy is met and the number of real-time instances in which the second model monitoring accuracy is met and the enrollment condition is met for the second monitoring event, or the fifth instance refers to the number of consecutive instances in which the second model monitoring accuracy is met and the withdrawal condition is met for the second monitoring event.

[0181] Among these, the above-mentioned monitoring event is that the model monitoring accuracy satisfies the usage conditions and the model monitoring accuracy satisfies the exit conditions for the monitoring event, and triggering the operation of a model monitoring report includes (1) the first monitoring event is that the first model monitoring accuracy satisfies the usage conditions and satisfies the fourth condition, and the second monitoring event is that the second model monitoring accuracy of any candidate model satisfies the usage conditions and satisfies the fifth condition, and the operation of a model monitoring report is triggered, and the content of the report is to instruct a model switch and carries the information of the candidate model, or (2) the first monitoring event is that the first model monitoring accuracy satisfies the usage conditions and satisfies the fourth condition, and the second monitoring event is that the second model monitoring accuracy of any candidate model does not satisfy the usage conditions and satisfies the sixth condition, and the operation of a model monitoring report is triggered, and the content of the report is to instruct a model switch and carries the information of the candidate model, and among these, (1) the fourth condition This includes the sixth number of times within the first reporting time length that the first model monitoring accuracy satisfies the withdrawal condition of the first monitoring event, and the sixth number is determined according to the number of real-time times that the first model monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the sixth number is determined according to the number of real-time times that the first model monitoring accuracy satisfies the withdrawal condition of the first monitoring event and the number of real-time times that the first model monitoring accuracy satisfies the joining condition of the first monitoring event, or The sixth count refers to the number of consecutive times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event, (2) the fifth condition includes that the seventh count in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, the seventh count is determined according to the real-time count in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the seventh count is determined according to the real-time count in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event,the second model monitoring accuracy is determined according to the real-time count of satisfying the exit condition of the second monitoring event, or the seventh count refers to the consecutive count of times that the second model monitoring accuracy satisfies the entry condition of the second monitoring event; (3) the sixth condition includes that an eighth count of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event within the second report timing duration is greater than a second report threshold, the eighth count is determined according to the real-time count of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event, or the eighth count is determined according to the real-time count of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event and the real-time count of times that the second model monitoring accuracy satisfies the entry condition of the second monitoring event, or the eighth count refers to the consecutive count of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event.

[0182] Further, controlling the trigger operation of the model monitoring report according to the aforementioned model monitoring accuracy and monitoring event related information further includes triggering a model monitoring report operation when the monitoring event is that the model monitoring accuracy of the currently used model is lower than the model monitoring accuracy of any candidate model, wherein the content of the report is for instructing to perform model switching, and carries information of the candidate model.

[0183] In the embodiments of the present disclosure, the operation further includes acquiring the at least two test sample points by using a time domain sliding window of a first length.

[0184] Further, the operation further includes performing preprocessing on the at least two test sample points before acquiring the model monitoring accuracy according to the at least two test sample points, wherein the preprocessing includes at least one of a weighting process and an anomaly exclusion process.

[0185] Wherein the access condition and / or the departure condition are determined according to a hysteresis parameter and a corresponding threshold value.

[0186] In the embodiments of the present disclosure, the monitoring event related information is determined according to protocol predefinition and / or configuration on the network side, and / or the monitoring reporting condition is determined according to protocol predefinition and / or configuration on the network side, and / or the first length is determined according to protocol predefinition and / or configuration on the network side, and / or the sliding step size of the time domain sliding window is determined according to protocol predefinition and / or configuration on the network side, and / or the hysteresis parameter is determined according to protocol predefinition and / or configuration on the network side, and / or the threshold value is determined according to protocol predefinition and / or configuration on the network side.

[0187] It should be noted that the above apparatus according to the embodiments of the present disclosure can implement all the method steps implemented by the method embodiments described above and achieve the same technical effects, and the same parts and beneficial effects as those of the method embodiments in this embodiment will not be described in detail repeatedly here.

[0188] The embodiments of the present disclosure further provide an information processing apparatus serving as a first apparatus. As shown in Figure 9, the information processing apparatus includes a memory 91, a transceiver 92, and a processor 93, The memory 91 is configured to store a computer program, the transceiver 92 is configured to transmit and receive data under the control of the processor 93, and the processor 93 is configured to read the computer program in the memory 91 to perform the operation of acquiring at least two test sample points by using a time domain sliding window of a first length, Wherein the test sample points are determined according to a target CSI and CSI restored by a model corresponding to the target CSI. The aforementioned first device is a network device or terminal.

[0189] The information processing device according to the embodiment of this disclosure acquires at least two test sample points using a time-domain sliding window of first length, the test sample points being determined according to a target CSI and a CSI restored by the model corresponding to the target CSI, and the first device is a network device or terminal. This supports the accurate acquisition of test sample points, thereby increasing the accuracy of the acquired model monitoring accuracy, and based on this, it further supports the accurate implementation of control over the trigger of model monitoring reports according to the model monitoring accuracy (i.e., monitoring results), and as a result supports the accurate implementation of decision operations such as model switching, fallback, update, activation and / or deactivation, and suitably solves the problem of not being able to accurately implement model decision operations that exist in the information processing modes for model monitoring in related technologies.

[0190] Specifically, the transceiver 92 is for receiving and transmitting data under the control of the processor 93.

[0191] In Figure 9, the bus architecture may include an arbitrary number of interconnected buses and bridges, specifically connecting various circuits between one or more processors, represented by processor 93, and memory, represented by memory 91. The bus architecture can also connect various other circuits, such as peripherals, voltage regulators, and power management circuits, but these are known in the art and will not be further described herein. The bus interface provides the interface. The transceiver 92 may consist of multiple elements, including a transmitter and a receiver, and provides a means for communicating with various other devices via a transmission medium, which includes wireless channels, wired channels, optical cables, etc. Processor 93 is responsible for managing the bus architecture and general processing, and memory 91 can store data used when operations are performed by processor 93.

[0192] The processor 93 may be a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or a Complex Programmable Logic Device (CPLD), and the processor may employ a multi-core architecture.

[0193] The processor is for executing any of the methods described in the embodiments of this disclosure by calling a computer program stored in memory, in accordance with the executable instructions obtained. The processor and memory may be physically separated.

[0194] It should be explained here that when the first device is implemented as a terminal, it may further include a bus interface and a user interface capable of communicating with it. For different user devices, the user interface may be an interface that can connect necessary devices externally or internally, and the connected devices include, but are not limited to, a small keyboard, display, speaker, microphone, joystick, etc.

[0195] Furthermore, the operation further includes pre-processing the at least two test sample points, the pre-processing of which includes at least one of weighting and anomaly exclusion.

[0196] In embodiments of the present disclosure, the first length is determined according to a protocol predefined and / or network-side configuration, and / or the sliding step size of the time-domain sliding window is determined according to a protocol predefined and / or network-side configuration.

[0197] It should be explained here that the apparatus according to the embodiments of this disclosure can implement all the method steps realized by the embodiments of the method and achieve the same technical effects, but the same parts and beneficial effects as those of the embodiments of the method in these embodiments will not be described in detail here.

[0198] The embodiments of this disclosure further provide an information processing device applicable to the first device, as shown in Figure 10, the information processing device is A first acquisition unit 101 for obtaining model monitoring accuracy based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI restored by the model corresponding to the target CSI, and the model monitoring accuracy is based on the statistical results of the at least two test sample points. A first determination unit 102 for determining monitoring event-related information regarding the model monitoring accuracy, which includes a definition of a monitoring event and the joining and / or leaving conditions of the monitoring event, The system includes a first control unit 103 for controlling the trigger operation of the model monitoring report according to the model monitoring accuracy and monitoring event-related information, Of these, the first device is a network device or terminal.

[0199] The information processing device according to the embodiment of the present disclosure acquires model monitoring accuracy in accordance with at least two test sample points, the test sample points are determined in accordance with target channel status information (CSI) and a CSI restored by the model that corresponds to the target CSI, the model monitoring accuracy is based on statistical results of the at least two test sample points, and the device determines monitoring event-related information for the model monitoring accuracy, which includes the definition of a monitoring event and the join and / or leave conditions of the monitoring event, and controls the trigger operation of the model monitoring report in accordance with the model monitoring accuracy and the monitoring event-related information, wherein the first device is a network device or a terminal. This supports the realization of control over the trigger of the model monitoring report in accordance with the model monitoring accuracy (i.e., monitoring results), thereby supporting the accurate realization of decision operations such as model switching, fallback, update, activation and / or deactivation, and suitably solves the problem that exists in the information processing methods for model monitoring in related technologies that it is not possible to accurately realize the decision operations of the model.

[0200] In addition, controlling the trigger operation of the model monitoring report according to the aforementioned model monitoring accuracy and monitoring event-related information includes triggering the model monitoring report operation if the monitoring event is such that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the joining conditions for the monitoring event, or triggering the model monitoring report operation if the monitoring event is such that the model monitoring accuracy meets the usage conditions and the model monitoring accuracy meets the leaving conditions for the monitoring event.

[0201] Furthermore, the information processing device further includes a second determination unit for determining monitoring reporting conditions, which include a reporting threshold and / or reporting time length, and (1) if the monitoring event described above is such that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the enrollment conditions for the monitoring event, triggering a model monitoring report operation; if the monitoring event is such that the model monitoring accuracy does not meet the usage conditions and the first number of times the model monitoring accuracy has met the enrollment conditions for the monitoring event within the reporting time length is greater than the reporting threshold, the content of the report is to indicate that the currently used model is invalid, the first number of times is determined according to the number of real-time times the model monitoring accuracy has met the enrollment conditions for the monitoring event, or the first number of times is determined according to the number of real-time times the model monitoring accuracy has met the enrollment conditions for the monitoring event and the number of real-time times the model monitoring accuracy has met the disengagement conditions for the monitoring event. Or, the first count refers to the number of consecutive times the model monitoring accuracy satisfies the join condition of the monitoring event, and / or, (2) if the monitoring event described above is one in which the model monitoring accuracy satisfies the usage condition and the model monitoring accuracy satisfies the exit condition of the monitoring event, triggering a model monitoring report operation is if the monitoring event is one in which the model monitoring accuracy satisfies the usage condition and the second count within the reporting time length in which the model monitoring accuracy satisfies the exit condition of the monitoring event is greater than the reporting threshold, wherein the content of the report is to indicate that the currently used model is invalid, and the second count is determined according to the number of real-time times the model monitoring accuracy satisfies the exit condition of the monitoring event, or the second count is determined according to the number of real-time times the model monitoring accuracy satisfies the exit condition of the monitoring event and the number of real-time times the model monitoring accuracy satisfies the join condition of the monitoring event, or the second count is,This includes the fact that the model monitoring accuracy refers to the number of consecutive times the exit condition for the monitoring event is met.

[0202] The models include a currently used model and at least one candidate model, and obtaining model monitoring accuracy in accordance with at least two test sample points as described above includes obtaining a first model monitoring accuracy for the currently used model in accordance with at least two test sample points corresponding to the currently used model, and obtaining a second model monitoring accuracy for the candidate model in accordance with at least two test sample points corresponding to the candidate model.

[0203] In the embodiments of the present disclosure, the monitoring event includes a first monitoring event for the currently used model and a second monitoring event for the candidate model; the reporting threshold includes a first reporting threshold for the currently used model and a second reporting threshold for the candidate model; and the reporting timing length includes a first reporting timing length for the currently used model and a second reporting timing length for the candidate model.

[0204] Among these, the above-mentioned monitoring event is that the model monitoring accuracy does not meet the usage conditions and the model monitoring accuracy meets the subscription conditions for the monitoring event, and triggering the operation of a model monitoring report includes (1) if the first monitoring event is that the first model monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second model monitoring accuracy of any candidate model meets the usage conditions and the second condition is met, triggering the operation of a model monitoring report, the content of which is to instruct a model switch and carries the information of the candidate model, or (2) if the first monitoring event is that the first model monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second model monitoring accuracy of any candidate model does not meet the usage conditions and the third condition is met, triggering the operation of a model monitoring report, the content of which is to instruct a model switch and carries the information of the candidate model, and among these, (1) the One condition is that the third number of times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the third number is determined according to the real-time number of times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event, or the third number is determined according to the real-time number of times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event and the real-time number of times the first model monitoring accuracy satisfies the withdrawal condition for the first monitoring event, or (1) The third count refers to the number of consecutive times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event, (2) the second condition includes that the fourth count within the second reporting time length in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event is greater than the second reporting threshold, the fourth count is determined according to the real-time count in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the fourth count is determined according to the real-time count in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event,The second model monitoring accuracy is determined according to the number of real-time instances in which the withdrawal condition of the second monitoring event is met, or the fourth instance refers to the number of consecutive instances in which the second model monitoring accuracy meets the enrollment condition of the second monitoring event; (3) The third condition refers to the fifth instance in which the second model monitoring accuracy meets the withdrawal condition of the second monitoring event within the second reporting time length being greater than the second reporting threshold; the fifth instance is determined according to the number of real-time instances in which the second model monitoring accuracy meets the withdrawal condition of the second monitoring event, or the fifth instance is determined according to the number of real-time instances in which the second model monitoring accuracy meets the withdrawal condition of the second monitoring event and the number of real-time instances in which the second model monitoring accuracy meets the enrollment condition of the second monitoring event, or the fifth instance refers to the number of consecutive instances in which the second model monitoring accuracy meets the withdrawal condition of the second monitoring event.

[0205] In embodiments of the present disclosure, triggering a model monitoring report operation when the above-described monitoring event satisfies the usage conditions of the model monitoring accuracy and satisfies the exit conditions of the monitoring event includes (1) triggering a model monitoring report operation when the first monitoring event satisfies the usage conditions of the first model monitoring accuracy and satisfies the fourth condition, and the second monitoring event satisfies the usage conditions of the second model monitoring accuracy of any candidate model and satisfies the fifth condition, with the content of the report instructing a model switch and carrying information of the candidate model, or (2) triggering a model monitoring report operation when the first monitoring event satisfies the usage conditions of the first model monitoring accuracy and satisfies the fourth condition, and the second monitoring event does not satisfy the usage conditions of any candidate model and satisfies the sixth condition, with the content of the report instructing a model switch and carrying information of the candidate model, of which (1) the The fourth condition is that the sixth number of times the first model monitoring accuracy satisfies the withdrawal condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the sixth number is determined in real time according to the number of times the first model monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the sixth number is determined in real time according to the number of times the first model monitoring accuracy satisfies the withdrawal condition of the first monitoring event and the number of times the first model monitoring accuracy satisfies the joining condition of the first monitoring event, or (1) The sixth count refers to the number of consecutive times the first model monitoring accuracy satisfies the enrollment condition for the first monitoring event, (2) the fifth condition includes that the seventh count in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, the seventh count is determined according to the real-time count in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the seventh count is determined according to the real-time count in which the second model monitoring accuracy satisfies the enrollment condition for the second monitoring event,the seventh number of times is determined according to the real-time number of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event, or the seventh number of times refers to the consecutive number of times that the second model monitoring accuracy satisfies the entry condition of the second monitoring event; (3) the sixth condition includes that the eighth number of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event within the second reporting timing duration is greater than the second reporting threshold, and the eighth number of times is determined according to the real-time number of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event, or the eighth number of times is determined according to the real-time number of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event and the real-time number of times that the second model monitoring accuracy satisfies the entry condition of the second monitoring event, or the eighth number of times refers to the consecutive number of times that the second model monitoring accuracy satisfies the exit condition of the second monitoring event.

[0206] Further, controlling the trigger operation of the model monitoring report according to the aforementioned model monitoring accuracy and monitoring event related information further includes triggering the model monitoring report operation when the monitoring event is that the model monitoring accuracy of the currently used model is lower than the model monitoring accuracy of any candidate model, wherein the content of the report is used to instruct to perform model switching, and carries the information of the candidate model.

[0207] In the embodiments of the present disclosure, the information processing apparatus further includes a second acquisition unit configured to acquire the at least two test sample points by using a time domain sliding window of a first length.

[0208] Further, the information processing apparatus further includes a first processing unit configured to perform preprocessing on the at least two test sample points before acquiring the model monitoring accuracy according to the at least two test sample points, wherein the preprocessing includes at least one of a weighting process and an anomaly rejection process.

[0209] The aforementioned joining and / or leaving conditions are determined according to the hysteresis parameters and the corresponding threshold values.

[0210] In embodiments of this disclosure, the monitoring event-related information is determined according to the protocol predefined and / or the network-side configuration, and / or the monitoring reporting conditions are determined according to the protocol predefined and / or the network-side configuration, and / or the first length is determined according to the protocol predefined and / or the network-side configuration, and / or the sliding step size of the time-domain sliding window is determined according to the protocol predefined and / or the network-side configuration, and / or the hysteresis parameters are determined according to the protocol predefined and / or the network-side configuration, and / or the threshold is determined according to the protocol predefined and / or the network-side configuration.

[0211] What should be explained here is that the apparatus according to the embodiments of this disclosure can implement all the method steps realized by the embodiments of the method and can achieve the same technical effects, but the same parts and beneficial effects as those of the embodiments of the method in these embodiments will not be described in detail here.

[0212] The embodiments of this disclosure further provide an information processing device applicable to the first device, as shown in Figure 11, the information processing device is: It includes a third acquisition unit 111 for acquiring at least two test sample points using a time-domain sliding window of first length, Of these, the test sample points are determined according to the target CSI and the CSI restored by the model that corresponds to the target CSI. The aforementioned first device is a network device or terminal.

[0213] The information processing device according to the embodiment of the present disclosure acquires at least two test sample points using a time-domain sliding window of first length, the test sample points being determined according to a target CSI and a CSI restored by the model corresponding to the target CSI, and the first device is a network device or terminal. This supports the accurate acquisition of test sample points, thereby improving the accuracy of the acquired model monitoring accuracy, and based on this, it further supports the accurate realization of control over the trigger of model monitoring reports according to the model monitoring accuracy (i.e., monitoring results), and as a result supports the accurate realization of decision operations such as model switching, fallback, update, activation and / or deactivation, and suitably solves the problem of not being able to accurately realize model decision operations that exist in the information processing modes for model monitoring in related technologies.

[0214] Furthermore, the information processing device further includes a second processing unit for pre-processing the at least two test sample points, the pre-processing of which includes at least one of weighting processing and abnormality exclusion processing.

[0215] In embodiments of the present disclosure, the first length is determined according to a protocol predefined and / or network-side configuration, and / or the sliding step size of the time-domain sliding window is determined according to a protocol predefined and / or network-side configuration.

[0216] What should be explained here is that the apparatus according to the embodiments of this disclosure can implement all the method steps realized by the embodiments of the method and can achieve the same technical effects, but the same parts and beneficial effects as those of the embodiments of the method in these embodiments will not be described in detail here.

[0217] It should be noted that the division of modules in the embodiments of this disclosure is illustrative and merely a division based on logical functions; other division methods are possible in actual implementation. Furthermore, each functional module in each embodiment of this disclosure may be integrated into a single processing module, each module may exist physically separately, or two or more modules may be integrated into a single module. The integrated module may be implemented in hardware form or in the form of a software functional unit.

[0218] The integrated module may be implemented in the form of a software functional unit and, if sold or used as an independent product, may be stored in a single processor-readable storage medium. Based on this understanding, essential parts of the technical aspects of the present disclosure, or parts that contribute to the prior art, or all or part of such technical aspects, can be embodied in the form of a software product. Such computer software product is stored in a single storage medium and includes several instructions for causing a single computer device (which may be a personal computer, server, or network device, etc.) or processor to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage mediums include various media capable of storing program code, such as USB flash drives, portable hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0219] Embodiments of the present disclosure further provide a non-temporary, readable storage medium storing a computer program, wherein the computer program causes a processor to execute the above-described method.

[0220] The aforementioned non-temporary readable storage medium may be any available medium or data storage device accessible by the processor, and includes, but is not limited to, magnetic storage devices (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage devices (e.g., optical disks (Compact Disc, CD), Digital Video Disc (DVD), Blu-ray Disc (Blu-ray® Disc, BD), High-definition Versatile Disc (HVD), etc.), and semiconductor storage devices (e.g., ROM, Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Non-volatile Memory Device (NAND (Non-volatile Memory Device) FLASH), Solid State Drives (SSD)), etc.).

[0221] Of these, all of the above-mentioned embodiments of the information processing method are applicable to the embodiments of the non-temporary readable storage medium and can achieve the same technical effects.

[0222] Those skilled in the art will understand that embodiments of the present disclosure may be provided as methods, systems, or computer program products. Accordingly, the present disclosure may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware. Moreover, the present disclosure may take the form of computer program products implemented on one or more computer-compatible storage media (including, but not limited to, magnetic disk memory and optical memory) containing computer-compatible program code.

[0223] This disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of this disclosure. Each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, should be understood to be realized by computer-executable instructions. These computer-executable instructions are provided to a processor of a general-purpose computer, a dedicated computer, a embedded processor or other programmable data processing device to form a machine, and the instructions executed by the processor of the computer or other programmable data processing device form a device for realizing the functions specified in one or more flows of the flowchart and / or one or more blocks of the block diagram.

[0224] These processor-executable instructions may be stored in processor-readable memory that can guide a computer or other programmable data processing device to operate in a particular manner, and the instructions stored in said processor-readable memory form a product including an instruction unit. The instruction unit implements functions specified in one or more flows of a flowchart and / or one or more blocks of a block diagram.

[0225] These processor-executable instructions may be loaded into a computer or other programmable data processing device, and by executing a series of operational steps on the computer or other programmable device, they form the processing to be implemented on the computer, and the instructions executed on the computer or other programmable device provide steps to implement the functions specified in one or more flows of a flowchart and / or one or more blocks of a block diagram.

[0226] It should be explained that the above division into modules is merely a division in terms of logical function, and in actual implementation, all or some of the modules may be integrated into a single physical entity or may be physically separated. Furthermore, these modules may all be implemented by calling software via processing elements, or all by hardware, or some modules may be implemented by calling software via processing elements while others are implemented by hardware. For example, a deterministic module may be an independently provided processing element or may be implemented by integration into a chip of the above device. In addition, a deterministic module may be stored in the memory of the above device in the form of program code and executed by being called by a processing element of the above device. The implementation of other modules is similar to that of deterministic modules. Furthermore, these modules may all or some be integrated or implemented individually. The processing elements described herein may be integrated circuits having signal processing capabilities. In implementation, each step of the above method or each of the above modules may be implemented by using hardware integrated logic circuits within a processor element or by using software-form instructions.

[0227] For example, each module, unit, subunit, or submodule may be configured as one or more integrated circuits that implement the above method, such as one or more Application Specific Integrated Circuits (ASICs), one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs). As another example, if one of the above modules is implemented by calling program code via a processing element, the processing element may be a general-purpose processor such as a Central Processing Unit (CPU), or another processor capable of calling program code. As yet another example, these modules may be integrated together or implemented in the form of a system-on-a-chip (SOC).

[0228] The terms “first,” “second,” and so on in the specification and claims of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. The data used in this manner are interchangeable in appropriate contexts, and it should be understood that the embodiments of the Application described herein can be implemented in an order other than that illustrated or described herein. Furthermore, the terms “includes” and “have,” and all their variations, cover non-exclusive inclusion; for example, a procedure, method, system, product or apparatus including a set of steps or units is not limited to including only those steps or units explicitly listed, but may include other steps or units not explicitly listed, or other steps or units specific to those procedures, methods, products or apparatus. Furthermore, as used in the specification and claims, “and / or” refers to at least one of the connected entities, for example, “A and / or B and / or C” includes seven cases: A only exists, B only exists, C only exists, both A and B exist, both B and C exist, both A and C exist, and all of A, B and C exist. Similarly, as used in this specification and claims, “at least one of A and B” should be understood as “A only exists, B only exists, or both A and B exist.”

[0229] Clearly, a person skilled in the art can make various modifications and variations of this disclosure without departing from the spirit and scope of this disclosure. If such modifications and variations of this disclosure fall within the scope of the claims of this disclosure and the equivalent art, then such modifications and variations are also included in this disclosure.

Claims

1. An information processing method applicable to the first device, wherein the method is The performance monitoring accuracy is obtained based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI reconstructed by a model corresponding to the target CSI, and the performance monitoring accuracy is based on the statistical results of the at least two test sample points. To determine the monitoring event-related information regarding the performance monitoring accuracy, which includes the definition of the monitoring event and the joining and / or leaving conditions of the monitoring event. This includes controlling the trigger operation for performance monitoring reports according to the performance monitoring accuracy and monitoring event-related information, The first device is a network device or terminal, and the information processing method is as described above.

2. Controlling the trigger operation for performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the aforementioned monitoring event, then the operation of the performance monitoring report is triggered. Alternatively, the information processing method according to claim 1, which includes triggering a performance monitoring report operation when the monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event.

3. This further includes determining monitoring reporting conditions that include reporting thresholds and / or reporting time durations, If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation of the performance monitoring report will be triggered. The monitoring event is that the performance monitoring accuracy does not meet the usage conditions, and the first number of times the performance monitoring accuracy meets the subscription conditions for the monitoring event within the reporting time duration is greater than the reporting threshold, thereby triggering a performance monitoring report operation, wherein the content of the report is to indicate that the currently used model is invalid, and the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event, or the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event and the number of real-time times the performance monitoring accuracy meets the withdrawal conditions for the monitoring event, or the first number refers to the number of consecutive times the performance monitoring accuracy meets the subscription conditions for the monitoring event. And / or, if the aforementioned monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then triggering the operation of the performance monitoring report is: The information processing method according to claim 2, wherein the monitoring event is that the performance monitoring accuracy satisfies the usage conditions, and the operation of a performance monitoring report is triggered when the second number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event within the reporting time length is greater than the reporting threshold, the content of which is to indicate that the currently used model is invalid, the second number is determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event, or the second number is determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event and the number of real-time times the performance monitoring accuracy satisfies the joining condition of the monitoring event, or the second number refers to the number of consecutive times the performance monitoring accuracy satisfies the exit condition of the monitoring event.

4. The aforementioned model includes the currently used model and at least one candidate model. Depending on the two test sample points mentioned above, obtaining performance monitoring accuracy is possible. To obtain a first performance monitoring accuracy for the currently used model, based on at least two test sample points corresponding to the currently used model, The information processing method according to claim 3, further comprising obtaining a second performance monitoring accuracy for the candidate model in accordance with at least two test sample points corresponding to the candidate model.

5. The aforementioned monitoring events include a first monitoring event for the currently used model and a second monitoring event for the candidate model. The reported threshold includes a first reported threshold for the currently used model and a second reported threshold for the candidate model. The information processing method according to claim 4, wherein the reported timing time length includes a first reported timing time length for the currently used model and a second reported timing time length for the candidate model.

6. If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation of the performance monitoring report will be triggered. If the first monitoring event indicates that the first performance monitoring accuracy does not meet the usage conditions, and the first condition is met, and the second monitoring event indicates that the second performance monitoring accuracy of any candidate model meets the usage conditions, and the second condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch, and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not meet the usage conditions and the third condition is met, then a performance monitoring report operation is triggered, the content of which is to instruct a model switch and includes carrying information of the candidate model. Among these, the first condition includes the third number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event, or the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the third number refers to the consecutive number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event. The second condition includes that the fourth number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the fourth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The third condition includes that the fifth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fifth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the fifth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the fifth number refers to the consecutive number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, as described in claim 5.

7. The above-mentioned monitoring event is triggered when the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, which triggers the operation of the performance monitoring report. If the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model satisfies the usage conditions and the fifth condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not satisfy the usage conditions and the sixth condition is met, then a performance monitoring report operation is triggered, and the content of the report is to instruct a model switch and includes carrying information of the candidate model. Among these, the fourth condition includes that the sixth number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event, or the sixth number refers to the consecutive number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event. The fifth condition includes that the seventh instance in which the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the seventh instance refers to the consecutive number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The sixth condition includes that the eighth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the eighth number refers to the consecutive number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, as described in claim 5.

8. Controlling the trigger operation for performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: The monitoring event further includes triggering a performance monitoring report operation if the performance monitoring accuracy of the currently used model is lower than that of any candidate model. The information processing method according to claim 2, wherein the content of the report is intended to instruct a model switch and carries information about the candidate model.

9. The information processing method according to claim 1, further comprising obtaining the at least two test sample points using a time-domain sliding window of a first length.

10. Depending on at least two test sample points, before obtaining performance monitoring accuracy, The method further includes pre-processing the at least two test sample points, The information processing method according to claim 1, wherein the preprocessing includes at least one of weighting processing and abnormality exclusion processing.

11. The information processing method according to claim 1, wherein the aforementioned joining conditions and / or leaving conditions are determined according to a hysteresis parameter and a corresponding threshold value.

12. The aforementioned monitoring event-related information is determined according to the protocol's predefined settings and / or the network's configuration. And / or, the monitoring reporting conditions are determined according to the protocol's predefined and / or network-side configuration. and / or, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration. And / or, the hysteresis parameters are determined according to the protocol's predefined and / or network-side configuration. The information processing method according to any one of claims 1 to 11, wherein the threshold is determined according to the predefined protocol and / or the network configuration.

13. An information processing method applicable to the first device, wherein the method is This includes obtaining at least two test sample points using a time-domain sliding window of first length, Of these, the test sample points are determined according to the target CSI and the CSI restored by the model that corresponds to the target CSI. The first device is a network device or terminal, and the information processing method is also described.

14. The method further includes pre-processing the at least two test sample points, The information processing method according to claim 13, wherein the preprocessing includes at least one of weighting processing and abnormality exclusion processing.

15. The aforementioned first length is determined according to the protocol's predefined and / or network-side configuration. The information processing method according to any one of claims 13 to 14, wherein the sliding step size of the time-domain sliding window is determined according to the predefined protocol and / or the network configuration.

16. The first device is an information processing device, the information processing device includes a memory, a transceiver, and a processor. Memory is for storing computer programs, and transceivers are for sending and receiving data under the control of the processor, and the processor reads the computer programs in the memory, The performance monitoring accuracy is obtained based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI reconstructed by a model corresponding to the target CSI, and the performance monitoring accuracy is based on the statistical results of the at least two test sample points. To determine the monitoring event-related information regarding the performance monitoring accuracy, which includes the definition of the monitoring event and the joining and / or leaving conditions of the monitoring event. This is for controlling the trigger operation of performance monitoring reports according to the performance monitoring accuracy and monitoring event-related information, Among these, the first device is an information processing device, which is a network device or terminal.

17. Controlling the trigger operation for performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the aforementioned monitoring event, then the operation of the performance monitoring report is triggered. Alternatively, the information processing device according to claim 16, which includes triggering a performance monitoring report operation if the monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event.

18. The aforementioned operation is, This further includes determining monitoring reporting conditions that include reporting thresholds and / or reporting time durations, If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation of the performance monitoring report will be triggered. The monitoring event is that the performance monitoring accuracy does not meet the usage conditions, and the first number of times the performance monitoring accuracy meets the subscription conditions for the monitoring event within the reporting time duration is greater than the reporting threshold, thereby triggering a performance monitoring report operation, wherein the content of the report is to indicate that the currently used model is invalid, and the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event, or the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event and the number of real-time times the performance monitoring accuracy meets the withdrawal conditions for the monitoring event, or the first number refers to the number of consecutive times the performance monitoring accuracy meets the subscription conditions for the monitoring event. And / or, if the aforementioned monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then triggering the operation of the performance monitoring report is: The information processing device according to claim 17, wherein the monitoring event is that the performance monitoring accuracy satisfies the usage conditions, and the operation of a performance monitoring report is triggered when the second number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event within the reporting time length is greater than the reporting threshold, the content of which is to indicate that the currently used model is invalid, the second number is determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event, or the second number is determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event and the number of real-time times the performance monitoring accuracy satisfies the joining condition of the monitoring event, or the second number refers to the number of consecutive times the performance monitoring accuracy satisfies the exit condition of the monitoring event.

19. The aforementioned model includes the currently used model and at least one candidate model. Depending on the two test sample points mentioned above, obtaining performance monitoring accuracy is possible. To obtain a first performance monitoring accuracy for the currently used model, based on at least two test sample points corresponding to the currently used model, The information processing device according to claim 18, further comprising obtaining a second performance monitoring accuracy for the candidate model in accordance with at least two test sample points corresponding to the candidate model.

20. The aforementioned monitoring events include a first monitoring event for the currently used model and a second monitoring event for the candidate model. The reported threshold includes a first reported threshold for the currently used model and a second reported threshold for the candidate model. The information processing device according to claim 19, wherein the reported timing time length includes a first reported timing time length for the currently used model and a second reported timing time length for the candidate model.

21. If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation of the performance monitoring report will be triggered. If the first monitoring event indicates that the first performance monitoring accuracy does not meet the usage conditions, and the first condition is met, and the second monitoring event indicates that the second performance monitoring accuracy of any candidate model meets the usage conditions, and the second condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch, and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not meet the usage conditions and the third condition is met, then a performance monitoring report operation is triggered, the content of which is to instruct a model switch and includes carrying information of the candidate model. Among these, the first condition includes the third number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event, or the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the third number refers to the consecutive number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event. The second condition includes that the fourth number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the fourth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The third condition includes that the fifth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fifth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the fifth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the fifth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, as described in claim 20.

22. The above-mentioned monitoring event is triggered when the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, which triggers the operation of the performance monitoring report. If the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model satisfies the usage conditions and the fifth condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not satisfy the usage conditions and the sixth condition is met, then a performance monitoring report operation is triggered, and the content of the report is to instruct a model switch and includes carrying information of the candidate model. Among these, the fourth condition includes that the sixth number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event, or the sixth number refers to the consecutive number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event. The fifth condition includes that the seventh instance in which the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the seventh instance refers to the consecutive number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The sixth condition includes that the eighth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the eighth number refers to the consecutive number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, as described in claim 20.

23. Controlling the trigger operation for performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: The monitoring event further includes triggering a performance monitoring report operation if the performance monitoring accuracy of the currently used model is lower than that of any candidate model. The information processing device according to claim 17, wherein the content of the report is intended to instruct a model switch and carries information about the candidate model.

24. The aforementioned operation is, The information processing device according to claim 16, further comprising acquiring the at least two test sample points using a time-domain sliding window of a first length.

25. The aforementioned operation is, The method further includes pre-processing at least two test sample points before obtaining performance monitoring accuracy, depending on at least two test sample points. The information processing device according to claim 16, wherein the preprocessing includes at least one of weighting processing and abnormality exclusion processing.

26. The information processing device according to claim 16, wherein the aforementioned joining conditions and / or withdrawal conditions are determined according to a hysteresis parameter and a corresponding threshold value.

27. The aforementioned monitoring event-related information is determined according to the protocol's predefined settings and / or the network's configuration. And / or, the monitoring reporting conditions are determined according to the protocol's predefined and / or network-side configuration. and / or, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration. And / or, the hysteresis parameters are determined according to the protocol's predefined and / or network-side configuration. The information processing device according to any one of claims 16 to 26, wherein the threshold is determined according to the protocol's predefined and / or network-side configuration.

28. The first device is an information processing device, the information processing device includes a memory, a transceiver, and a processor. Memory is for storing computer programs, and transceivers are for sending and receiving data under the control of the processor, and the processor reads the computer programs in the memory, This is for performing the operation of obtaining at least two test sample points using a time-domain sliding window of first length. Of these, the test sample points are determined according to the target CSI and the CSI restored by the model that corresponds to the target CSI. The aforementioned first device is an information processing device, which is a network device or terminal.

29. The aforementioned operation is, The method further includes pre-processing the at least two test sample points, The information processing device according to claim 28, wherein the preprocessing includes at least one of weighting processing and abnormality exclusion processing.

30. The aforementioned first length is determined according to the protocol's predefined and / or network-side configuration. The information processing device according to any one of claims 28 to 29, wherein the sliding step size of the time-domain sliding window is determined according to the predefined protocol and / or the network-side configuration.

31. An information processing device applicable to the first device, wherein the information processing device is A first acquisition unit for obtaining performance monitoring accuracy based on at least two test sample points, wherein the test sample points are determined based on target channel status information (CSI) and a CSI restored by a model corresponding to the target CSI, and the performance monitoring accuracy is based on statistical results of the at least two test sample points. A first determination unit for determining monitoring event-related information relating to the performance monitoring accuracy, which includes a definition of a monitoring event and the joining and / or leaving conditions of the monitoring event, Includes a first control unit for controlling the trigger operation of performance monitoring reports according to the performance monitoring accuracy and monitoring event-related information, Among these, the first device is an information processing device, which is a network device or terminal.

32. Controlling the trigger operation for performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the aforementioned monitoring event, then the operation of the performance monitoring report is triggered. Alternatively, the information processing apparatus according to claim 31, which includes triggering a performance monitoring report operation if the monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event.

33. It further includes a second determination unit for determining monitoring reporting conditions, which include reporting thresholds and / or reporting time durations. If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation of the performance monitoring report will be triggered. The monitoring event is that the performance monitoring accuracy does not meet the usage conditions, and the first number of times the performance monitoring accuracy meets the subscription conditions for the monitoring event within the reporting time duration is greater than the reporting threshold, thereby triggering a performance monitoring report operation, wherein the content of the report is to indicate that the currently used model is invalid, and the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event, or the first number is determined according to the number of real-time times the performance monitoring accuracy meets the subscription conditions for the monitoring event and the number of real-time times the performance monitoring accuracy meets the withdrawal conditions for the monitoring event, or the first number refers to the number of consecutive times the performance monitoring accuracy meets the subscription conditions for the monitoring event. And / or, if the aforementioned monitoring event is that the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, then triggering the operation of the performance monitoring report is: The information processing apparatus according to claim 32, wherein the monitoring event is that the performance monitoring accuracy satisfies the usage conditions, and the operation of a performance monitoring report is triggered when the second number of times the performance monitoring accuracy satisfies the exit condition of the monitoring event within the reporting time length is greater than the reporting threshold, the content of which is to indicate that the currently used model is invalid, the second number is determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event, or the second number is determined according to the number of real-time times the performance monitoring accuracy satisfies the exit condition of the monitoring event and the number of real-time times the performance monitoring accuracy satisfies the joining condition of the monitoring event, or the second number refers to the number of consecutive times the performance monitoring accuracy satisfies the exit condition of the monitoring event.

34. The aforementioned model includes the currently used model and at least one candidate model. Depending on the two test sample points mentioned above, obtaining performance monitoring accuracy is possible. To obtain a first performance monitoring accuracy for the currently used model, based on at least two test sample points corresponding to the currently used model, The information processing apparatus according to claim 33, comprising obtaining a second performance monitoring accuracy for the candidate model in accordance with at least two test sample points corresponding to the candidate model.

35. The aforementioned monitoring events include a first monitoring event for the currently used model and a second monitoring event for the candidate model. The reported threshold includes a first reported threshold for the currently used model and a second reported threshold for the candidate model. The information processing apparatus according to claim 34, wherein the reported timing time length includes a first reported timing time length for the currently used model and a second reported timing time length for the candidate model.

36. If the aforementioned monitoring event occurs when the performance monitoring accuracy does not meet the usage conditions, and the performance monitoring accuracy meets the subscription conditions for the monitoring event, then the operation of the performance monitoring report will be triggered. If the first monitoring event indicates that the first performance monitoring accuracy does not meet the usage conditions, and the first condition is met, and the second monitoring event indicates that the second performance monitoring accuracy of any candidate model meets the usage conditions, and the second condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch, and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy does not meet the usage conditions and the first condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not meet the usage conditions and the third condition is met, then a performance monitoring report operation is triggered, the content of which is to instruct a model switch and includes carrying information of the candidate model. Among these, the first condition includes the third number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event, or the third number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the third number refers to the consecutive number of times the first performance monitoring accuracy satisfies the enrollment condition of the first monitoring event. The second condition includes that the fourth number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the fourth number is determined according to the number of real-time times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the number of real-time times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the fourth number refers to the number of consecutive times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The third condition includes that the fifth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the fifth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the fifth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the fifth number refers to the consecutive number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, as described in claim 35.

37. The above-mentioned monitoring event is triggered when the performance monitoring accuracy satisfies the usage conditions and the performance monitoring accuracy satisfies the exit conditions for the monitoring event, which triggers the operation of the performance monitoring report. If the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model satisfies the usage conditions and the fifth condition is met, then a performance monitoring report operation is triggered, and the content of the report is intended to instruct a model switch and carries the information of the candidate model, Alternatively, if the first monitoring event is that the first performance monitoring accuracy satisfies the usage conditions and the fourth condition is met, and the second monitoring event is that the second performance monitoring accuracy of any candidate model does not satisfy the usage conditions and the sixth condition is met, then a performance monitoring report operation is triggered, and the content of the report is to instruct a model switch and includes carrying information of the candidate model. Among these, the fourth condition includes that the sixth number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event within the first reporting time length is greater than the first reporting threshold, wherein the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event, or the sixth number is determined according to the real-time number of times the first performance monitoring accuracy satisfies the withdrawal condition of the first monitoring event and the real-time number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event, or the sixth number refers to the consecutive number of times the first performance monitoring accuracy satisfies the joining condition of the first monitoring event. The fifth condition includes that the seventh instance in which the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event, or the seventh instance is determined according to the real-time number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition for the second monitoring event, or the seventh instance refers to the consecutive number of times the second performance monitoring accuracy satisfies the enrollment condition for the second monitoring event. The sixth condition includes that the eighth number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event within the second reporting time length is greater than the second reporting threshold, wherein the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, or the eighth number is determined according to the real-time number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event and the real-time number of times the second performance monitoring accuracy satisfies the joining condition of the second monitoring event, or the eighth number refers to the consecutive number of times the second performance monitoring accuracy satisfies the withdrawal condition of the second monitoring event, as described in claim 35.

38. Controlling the trigger operation for performance monitoring reports according to the aforementioned performance monitoring accuracy and monitoring event-related information is: The monitoring event further includes triggering a performance monitoring report operation if the performance monitoring accuracy of the currently used model is lower than that of any candidate model. The information processing device according to claim 32, wherein the content of the report is intended to instruct a model switch and carries information about the candidate model.

39. The information processing apparatus according to claim 31, further comprising a second acquisition unit for acquiring the at least two test sample points using a time-domain sliding window of first length.

40. Depending on at least two test sample points, the system further includes a first processing unit for pre-processing the at least two test sample points before obtaining performance monitoring accuracy, The information processing apparatus according to claim 31, wherein the preprocessing includes at least one of weighting processing and abnormality exclusion processing.

41. The information processing apparatus according to claim 31, wherein the aforementioned joining conditions and / or leaving conditions are determined according to a hysteresis parameter and a corresponding threshold value.

42. The aforementioned monitoring event-related information is determined according to the protocol's predefined settings and / or the network's configuration. And / or, the monitoring reporting conditions are determined according to the protocol's predefined and / or network-side configuration. and / or, the first length is determined according to the protocol's predefined and / or network-side configuration. And / or, the sliding step size of the time-domain sliding window is determined according to the protocol's predefined and / or network-side configuration. And / or, the hysteresis parameters are determined according to the protocol's predefined and / or network-side configuration. The information processing apparatus according to any one of claims 31 to 41, wherein the threshold is determined according to the predefined protocol and / or the network configuration.

43. An information processing device applicable to the first device, wherein the information processing device is It includes a third acquisition unit for acquiring at least two test sample points using a time-domain sliding window of first length, Of these, the test sample points are determined according to the target CSI and the CSI restored by the model that corresponds to the target CSI. The aforementioned first device is an information processing device, which is a network device or terminal.

44. The system further includes a second processing unit for pre-processing the at least two test sample points, The information processing apparatus according to claim 43, wherein the preprocessing includes at least one of weighting processing and abnormality exclusion processing.

45. The aforementioned first length is determined according to the protocol's predefined and / or network-side configuration. The information processing apparatus according to any one of claims 43 to 44, wherein the sliding step size of the time-domain sliding window is determined according to the predefined protocol and / or the network configuration.

46. A non-temporary, readable storage medium storing a computer program, wherein the computer program causes a processor to execute the method according to any one of claims 1 to 15.

Images