Communication method and communication device

Abstract

The embodiment of the invention provides a communication method and a communication device. The method comprises the step of sending a report for indicating artificial intelligence (AI) model switching or mode switching when an entry condition or an exit condition of an event is met. According to the method provided by the invention, a plurality of events and entry and exit conditions corresponding to the events are defined, and when the conditions are met, active model or mode switching can be carried out.

Description

[0001] This application claims priority to U.S. Provisional Patent Application No. 63 / 507,790, entitled “A Method of AI Model Monitor,” filed on June 13, 2023.

[0002] All the public information disclosed in the above application is incorporated herein by reference. Technical Field

[0003] This application relates to the field of communications, and more specifically, to a communication method and a communication device. Background Technology

[0004] AI-based algorithms have been introduced into modern wireless communications to solve wireless problems such as channel estimation, scheduling, channel state information (CSI) compression (from user equipment to base station), multiple-in-multiple-out (MIMO) beamforming, and localization. As a data-driven approach, AI-based algorithms inevitably suffer from low generalization ability. The performance of artificial intelligence (AI) models depends heavily on the quality of the data used to train them. Even if an AI model is trained on a large dataset, it may lack the necessary knowledge to perform effectively in other environments, especially in wireless communications where channel information changes rapidly.

[0005] Due to generalization issues, user equipment (UE) or base station (BS) needs to detect its generalization performance and then switch to an appropriate AI model. For example, there are multiple AI models for various scenarios, including indoor urban environments, outdoor urban environments, rural areas, and high-speed rail. When the surrounding environment changes, the BS or UE needs to switch to another model. In existing technologies, when inference performance degrades, the UE or BS switches its AI model or falls back to a non-AI mode. This is a passive solution.

[0006] Therefore, how to switch AI models in advance is a technical problem that urgently needs to be solved. Summary of the Invention

[0007] This application provides a communication method and a communication device. The technical solution of this application defines multiple events and corresponding entry and exit conditions for these events, enabling proactive model or mode switching when any one of these conditions is met.

[0008] According to a first aspect, embodiments of this application provide a communication method, including: sending a report indicating artificial intelligence (AI) model switching or mode switching when an event entry condition or exit condition is met.

[0009] The method provided in this application defines multiple events and corresponding entry and exit conditions for these events. When these conditions are met, active model or mode switching can be performed.

[0010] The event can be any of the following: switching from an AI model to a non-AI model, switching from the first AI model to the second AI model, switching from a non-AI model to an AI model, or switching from the third AI model to the fourth AI model. The third AI model can be the same as the second AI model, and the fourth AI model can be the same as the first AI model.

[0011] In one possible implementation, the event is an operation that switches from AI mode to non-AI mode. The entry condition is determined by Diff11 and Thresh11, or the exit condition is determined by Diff11 and Thresh12. Diff11 is a first difference, Thresh11 is a first threshold, and Thresh12 is a second threshold. The first difference is the difference between the first data and the first anchor point, which includes one or more reference data. The first threshold and the second threshold are predefined or configured thresholds corresponding to the event.

[0012] The first data includes monitoring or measurement data from user equipment or network equipment. Further, the first data is monitoring or measurement data related to the AI ​​model. The first data includes any one or more of the following: perception data, measurement data, channel data, neuron data of the AI ​​model, and potential output data of the AI ​​model. The network equipment can be a browser (BS).

[0013] The first anchor point can be indicated by the BS (Background Checker). For example, the first anchor point is the data anchor point of the current AI / machine learning (ML) model. Alternatively, the first anchor point is the data anchor point closest to the first data point; that is, the first anchor point is the data anchor point with the smallest difference from the first data point among multiple anchor points.

[0014] The event associated with switching from AI mode to non-AI mode can be called event K1. Event K1 indicates that the current or best AI / ML model may be outdated, and the UE or BS should switch to non-AI mode.

[0015] The method provided in this application defines an event K1 and the corresponding entry and exit conditions for event K1. When any one of these conditions is met, the mode can be actively switched.

[0016] In a possible implementation, the entry condition is Diff11–Hys11>Thresh11, or the exit condition is Diff11+Hys12<Thresh12.

[0017] Hys11 is the first hysteresis parameter, and Hys12 is the second hysteresis parameter. The first hysteresis parameter and the second hysteresis parameter are predefined or configured parameters corresponding to event K1.

[0018] The units of Thresh11, Thresh12, Hys11, and Hys12 are the same as the unit of Diff11.

[0019] Hys11 and Hys12 can be the same or different. For example, they can be similar. Thresh11 and Thresh12 can be the same or different. For example, they can be similar.

[0020] In a possible implementation, the entry condition is Diff11+offset11–Hys11>Thresh11, or the exit condition is Diff11+offset12+Hys12<Thresh12, where offset11 and offset12 are predefined or configured offsets.

[0021] offset11 and offset12 can be the same or different. The values of offset11 and offset12 can be positive or negative.

[0022] The method provided in this application defines event K1 and the entry and exit conditions corresponding to event K1. When any one of the conditions is met, the active switching of the mode can be achieved.

[0023] In a possible implementation, the event is an operation of switching from the first AI model to the second AI model. The entry condition is determined by Diff21, Diff22, Thresh21, and Thresh22, or the exit condition is determined by Diff21, Diff22, Thresh23, and Thresh24. Diff21 is the second difference, Diff22 is the third difference, Thresh21 is the third threshold, Thresh22 is the fourth threshold, Thresh23 is the fifth threshold, Thresh24 is the sixth threshold. The second difference is the difference between the first data and the second anchor point, where the second anchor point is the data anchor point associated with the first AI model. The third difference is the difference between the first data and the third anchor point, where the third anchor point is the data anchor point associated with the second AI model. The third threshold and the fifth threshold are predefined or configured thresholds corresponding to the first AI model, and the fourth threshold and the sixth threshold are predefined or configured thresholds corresponding to the second AI model.

[0024] The event related to the operation of switching from the first AI model to the second AI model can be called event K2. Event K2 indicates that the current AI / ML model (referred to as model 1) may be outdated, and the UE should switch to another model (model 2).

[0025] In event K2, the current model becomes worse than Thresh21, while the other model becomes better than Thresh22. Becoming worse than Thresh21 means that the second difference between the first data and the second anchor point is greater than Thresh21. Becoming better than Thresh22 means that the third difference between the first data and the third anchor point is less than Thresh22. Thresh21 and Thresh22 can be configured to be the same or different.

[0026] The method provided in this application defines event K2 and the corresponding entry and exit conditions for event K2. When any one of these conditions is met, the active switching of the model can be achieved.

[0027] In a possible implementation, the entry condition is Diff21 – Hys21 > Thresh21 and Diff22 + Hys22 < Thresh22, or the exit condition is Diff21 + Hys23 < Thresh23 or Diff22 – Hys24 > Thresh24.

[0028] Hys21 is the third hysteresis parameter, Hys22 is the fourth hysteresis parameter, Hys23 is the fifth hysteresis parameter, Hys24 is the sixth hysteresis parameter. The third hysteresis parameter, the fourth hysteresis parameter, the fifth hysteresis parameter, and the sixth hysteresis parameter are predefined or configured parameters corresponding to event K2.

[0029] Hys21 and Hys23 can be the same or different. For example, they can be similar. Hys22 and Hys24 can be the same or different. For example, they can be similar. Thresh21 and Thresh23 can be the same or different. For example, they can be similar. Thresh22 and Thresh24 can be the same or different. For example, they can be similar.

[0030] In a possible implementation, the entry condition is Diff21 + offset21 – Hys21 > Thresh21 and Diff22 + offset22 + Hys22 < Thresh22, and the exit condition is Diff21 + offset23 + Hys23 < Thresh23 or Diff22 + offset24 – Hys24 > Thresh24, where offset21, offset22, offset23, and offset24 are predefined or configured offsets.

[0031] offset21 and offset23 can be the same or different. For example, they can be similar. offset22 and offset24 can be the same or different. For example, they can be similar. The values of offset21, offset 22, offset23, and offset24 can be positive or negative.

[0032] The method provided in this application defines event K2 and the corresponding entry and exit conditions for event K2. When any of these conditions is met, the active switching of the model can be achieved.

[0033] In a possible implementation, the event is an operation of switching from the non-AI mode to the AI mode. The entry condition is determined by Diff31 and Thresh31, or the exit condition is determined by Diff31 and Thresh32. Diff31 is the fourth difference, Thresh31 is the seventh threshold, Thresh32 is the eighth threshold. The fourth difference is the difference between the first data and the fourth anchor point, and the fourth anchor point includes one or more reference data. The seventh threshold and the eighth threshold are predefined or configured thresholds corresponding to the event.

[0034] The event related to the switching from the non-AI mode to the AI mode can be called event K3. Event K3 indicates that the AI / ML model becomes better than the threshold, and the UE or BS should switch from the non-AI mode to the AI mode.

[0035] In event K3, the fourth difference between the first data and the fourth anchor point is less than the threshold. The fourth anchor point can be indicated by the BS, or the fourth anchor point is the data anchor point closest to the first data, that is, the fourth anchor point is the data anchor point with the smallest difference from the first data among multiple anchor points.

[0036] The method provided in this application defines event K3 and the corresponding entry and exit conditions for event K3. When any of these conditions is met, active switching of the mode can be achieved.

[0037] In a possible implementation, the entry condition is Diff31 + Hys31 < Thresh31, or the exit condition is Diff31 – Hys32 > Thresh32.

[0038] Hys31 is the seventh hysteresis parameter, and Hys32 is the eighth hysteresis parameter. The seventh hysteresis parameter and the eighth hysteresis parameter are predefined or configured parameters corresponding to event K3.

[0039] The units of Thresh31, Thresh32, Hys31, and Hys32 are the same as the unit of Diff31.

[0040] Hys31 and Hys32 can be the same or different. For example, they can be similar. Thresh31 and Thresh32 can be the same or different. For example, they can be similar.

[0041] In a possible implementation, the entry condition is Diff31 + offset31 + Hys31 < Thresh31, and the exit condition is Diff31 + offset32 – Hys32 > Thresh32, where offset31 and offset32 are predefined or configured offsets.

[0042] offset31 and offset32 can be the same or different. The values of offset31 and offset32 can be positive or negative.

[0043] The method provided in this application defines event K3 and the corresponding entry and exit conditions for event K3. When any of these conditions is met, active switching of the mode can be achieved.

[0044] In a possible implementation, the event is an operation of switching from the third AI model to the fourth AI model. The entry condition is determined by Diff41 and Diff42, or the exit condition is determined by Diff41 and Diff42. Diff41 is the fifth difference, Diff42 is the sixth difference. The fifth difference is the difference between the first data and the fifth anchor point, and the fifth anchor point is the data anchor point associated with the third AI model. The sixth difference is the difference between the first data and the sixth anchor point, and the sixth anchor point is the data anchor point associated with the fourth AI model.

[0045] The event related to the operation of switching from the third AI model to the fourth AI model can be called event K4. In event K4, another model (model 2) is better than the current model (model 1), and the UE or BS should switch to model 2.

[0046] The method provided in this application defines event K4 and the corresponding entry and exit conditions for event K4. When any one of these conditions is met, the active switching of the model can be achieved.

[0047] In a possible implementation, the entry condition is Diff42 + Hys41 < Diff41, and the exit condition is Diff42 – Hys42 > Diff41.

[0048] Hys41 is the ninth hysteresis parameter, and Hys42 is the tenth hysteresis parameter. The ninth hysteresis parameter and the tenth hysteresis parameter are predefined or configured parameters corresponding to event K4.

[0049] Hys41 and Hys42 can be the same or different. For example, they can be similar.

[0050] In a possible implementation, the entry condition is Diff42 + offset42 + Hys41 < Diff41 + offset41, and the exit condition is Diff42 + offset42 – Hys42 > Diff41 + offset41. offset41 and offset42 are predefined or configured offsets.

[0051] offset41 and offset42 can be the same or different. The values of offset41 and offset42 can be positive or negative.

[0052] The method provided in this application defines event K4 and the corresponding entry and exit conditions for event K4. When any one of these conditions is met, the active switching of the model can be achieved.

[0053] In a possible implementation, the first data includes the monitoring data or measurement data of the user equipment or the network equipment.

[0054] In one possible implementation, the first data includes monitoring data or measurement data related to the AI model of the user equipment or the network equipment.

[0055] In one possible implementation, the first data includes any one or more of the following: sensing data, measurement data, channel data, neuron data of the AI model, and potential output data of the AI model.

[0056] In one possible implementation, the method further includes: communicating according to the report.

[0057] In one possible implementation, the method is executed by the user equipment or the network equipment.

[0058] According to a second aspect, the present application provides a communication device, including: a sending module, configured to send a report indicating a switch of an artificial intelligence (AI) model or a mode switch when an entry condition or a departure condition of an event is satisfied.

[0059] In one possible implementation, the event is an operation of switching from the AI mode to the non-AI mode, the entry condition is determined by Diff11 and Thresh11, or the departure condition is determined by Diff11 and Thresh12. Diff11 is a first difference, Thresh11 is a first threshold, Thresh12 is a second threshold, the first difference is the difference between the first data and a first anchor point, the first anchor point includes one or more reference data, and the first threshold and the second threshold are predefined or configured thresholds corresponding to the event.

[0060] In one possible implementation, the entry condition is Diff11 – Hys11 > Thresh11, or the departure condition is Diff11 + Hys12 < Thresh12.

[0061] In one possible implementation, the entry condition is Diff11 + offset11 – Hys11 > Thresh11, or the departure condition is Diff11 + offset12 + Hys12 < Thresh12, where offset11 and offset12 are predefined or configured offsets.

[0062] In one possible implementation, Hys11 is a first hysteresis parameter, Hys12 is a second hysteresis parameter, and the first hysteresis parameter and the second hysteresis parameter are predefined or configured parameters corresponding to the event.

[0063] In a possible implementation, the event is an operation of switching from the first AI model to the second AI model. The entry condition is determined by Diff21, Diff22, Thresh21, and Thresh22, or the exit condition is determined by Diff21, Diff22, Thresh23, and Thresh24. Diff21 is the second difference, Diff22 is the third difference, Thresh21 is the third threshold, Thresh22 is the fourth threshold, Thresh23 is the fifth threshold, and Thresh24 is the sixth threshold. The second difference is the difference between the first data and the second anchor point, where the second anchor point is the data anchor point associated with the first AI model. The third difference is the difference between the first data and the third anchor point, where the third anchor point is the data anchor point associated with the second AI model. The third threshold and the fifth threshold are predefined or configured thresholds corresponding to the first AI model, and the fourth threshold and the sixth threshold are predefined or configured thresholds corresponding to the second AI model.

[0064] In a possible implementation, the entry condition is Diff21 – Hys21 > Thresh21 and Diff22 + Hys22 < Thresh22, or the exit condition is Diff21 + Hys23 < Thresh23 or Diff22 – Hys24 > Thresh24.

[0065] In a possible implementation, the entry condition is Diff21 + offset21 – Hys21 > Thresh21 and Diff22 + offset22 + Hys22 < Thresh22, and the exit condition is Diff21 + offset23 + Hys23 < Thresh23 or Diff22 + offset24 – Hys24 > Thresh24, where offset21, offset22, offset23, and offset24 are predefined or configured offsets.

[0066] In a possible implementation, Hys21 is the third hysteresis parameter, Hys22 is the fourth hysteresis parameter, Hys23 is the fifth hysteresis parameter, and Hys24 is the sixth hysteresis parameter. The third hysteresis parameter, the fourth hysteresis parameter, the fifth hysteresis parameter, and the sixth hysteresis parameter are predefined or configured parameters corresponding to the event.

[0067] In a possible implementation, the event is an operation of switching from a non-AI mode to an AI mode. The entry condition is determined by Diff31 and Thresh31, or the exit condition is determined by Diff31 and Thresh32. Diff31 is the fourth difference, Thresh31 is the seventh threshold, Thresh32 is the eighth threshold. The fourth difference is the difference between the first data and the fourth anchor point, and the fourth anchor point includes one or more reference data. The seventh threshold and the eighth threshold are predefined or configured thresholds corresponding to the event.

[0068] In a possible implementation, the entry condition is Diff31 + Hys31 < Thresh31, or the exit condition is Diff31 – Hys32 > Thresh32.

[0069] In a possible implementation, the entry condition is Diff31 + offset31 + Hys31 < Thresh31, and the exit condition is Diff31 + offset32 – Hys32 > Thresh32. offset31 and offset32 are predefined or configured offsets.

[0070] In a possible implementation, Hys31 is the seventh hysteresis parameter, Hys32 is the eighth hysteresis parameter, and the seventh hysteresis parameter and the eighth hysteresis parameter are predefined or configured parameters corresponding to the event.

[0071] In a possible implementation, the event is an operation of switching from the third AI model to the fourth AI model. The entry condition is determined by Diff41 and Diff42, or the exit condition is determined by Diff41 and Diff42. Diff41 is the fifth difference, Diff42 is the sixth difference. The fifth difference is the difference between the first data and the fifth anchor point, and the fifth anchor point is the data anchor point associated with the third AI model. The sixth difference is the difference between the first data and the sixth anchor point, and the sixth anchor point is the data anchor point associated with the fourth AI model.

[0072] In a possible implementation, the entry condition is Diff42 + Hys41 < Diff41, and the exit condition is Diff42 – Hys42 > Diff41.

[0073] In a possible implementation, the entry condition is Diff42 + offset42 + Hys41 < Diff41 + offset41, and the exit condition is Diff42 + offset42 – Hys42 > Diff41 + offset41. offset41 and offset42 are predefined or configured offsets.

[0074] In one possible implementation, Hys41 is the ninth hysteresis parameter and Hys42 is the tenth hysteresis parameter. The ninth and tenth hysteresis parameters are predefined or configured parameters corresponding to the event.

[0075] In one possible implementation, the first data includes monitoring or measurement data from user equipment or network equipment.

[0076] In one possible implementation, the first data includes monitoring or measurement data related to the AI ​​model of the user device or network device.

[0077] In one possible implementation, the first data includes any one or more of the following: perception data, measurement data, channel data, neuron data of the AI ​​model, and potential output data of the AI ​​model.

[0078] In one possible implementation, the device also includes a processing module for communicating based on the report.

[0079] In one possible implementation, the device is located on a user equipment or network device.

[0080] According to a third aspect, a communication device is provided, including a processor and a memory. The processor is connected to the memory. The memory is used to store instructions, and the processor is used to execute the instructions. When the processor executes the instructions stored in the memory, the processor is capable of performing the methods in any possible implementation of the first aspect described above.

[0081] According to a fourth aspect, this application provides a computer-readable storage medium including instructions. When the instructions are executed on a processor, the processor is capable of performing the methods in any possible implementation of the first aspect described above.

[0082] According to a fifth aspect, this application provides a computer program product, including computer program code. When the computer program code is run on a computer, the computer is able to perform the methods in any possible implementation of the first aspect described above.

[0083] It should be noted that all or part of the aforementioned computer program code can be stored in the first storage medium. The first storage medium can be packaged together with the processor or packaged separately from the processor.

[0084] According to a sixth aspect, this application provides a chip system including a memory and a processor. The memory is used to store a computer program, and the processor is used to retrieve and run the computer program from the memory, causing an electronic device equipped with the chip system to perform the method in any possible implementation of the first aspect described above. Attached Figure Description

[0085] Figure 1 This is a schematic diagram of a communication system 100 according to an embodiment of this application.

[0086] Figure 2 This is a schematic diagram of a communication system 100 according to an embodiment of this application.

[0087] Figure 3 This is a schematic diagram of an electronic device (ED) 110 and base stations 170a, 170b and / or 170c according to an embodiment of this application.

[0088] Figure 4 This is a schematic diagram of a unit or module in the device according to an embodiment of this application.

[0089] Figure 5 This is a schematic diagram of an AI-based communication device.

[0090] Figure 6 This is a schematic diagram of device 500 receiving reference data samples from device 600 in an embodiment of this application.

[0091] Figure 7 This is a schematic diagram of a reference data sample consisting of multiple groups according to an embodiment of this application.

[0092] Figure 8 This is a schematic diagram of an approximation based on DNN in an embodiment of this application.

[0093] Figure 9 This is a schematic diagram of an embodiment of the communication method of this application.

[0094] Figure 10 This is a schematic diagram of an embodiment of the communication method of this application.

[0095] Figure 11 This is a schematic diagram of projecting a high-dimensional signal onto a low-dimensional signal according to an embodiment of this application.

[0096] Figure 12 This is a flowchart of a communication method according to an embodiment of this application.

[0097] Figure 13 This is a schematic diagram of the determination matrix U in an embodiment of this application.

[0098] Figure 14 This is a schematic diagram of the first sampling matrix P1 in an embodiment of this application.

[0099] Figure 15 This is a schematic diagram of the sampling matrix compression matrix U in an embodiment of this application.

[0100] Figure 16 This is a schematic diagram of the scoring distance in the low-frequency space according to an embodiment of this application.

[0101] Figure 17 This is a flowchart of a communication method according to an embodiment of this application.

[0102] Figure 18 This is a schematic block diagram of a communication device according to an embodiment of this application.

[0103] Figure 19 This is a schematic block diagram of another communication device according to an embodiment of this application. Detailed Implementation

[0104] The technical solutions in this application are described below with reference to the accompanying drawings.

[0105] The technical solutions in this application will now be described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this application.

[0106] This application will provide aspects, embodiments, or features relating to systems comprising multiple devices, components, modules, etc. It should be understood and recognized that individual systems may include additional devices, components, modules, etc., and / or may exclude all devices, components, modules, etc. discussed in conjunction with the accompanying drawings. Furthermore, combinations of these options may also be used.

[0107] Furthermore, in the embodiments of this application, the terms "exemplarily" and the phrase "for example" are used to indicate illustration or description, etc. Any embodiment or design described as "exemplarily" in this application should not be construed as superior to or more advantageous than other embodiments or designs. Rather, the term "example" is used to present the concept in a particular manner.

[0108] Unless otherwise specified, the phrases "in some possible embodiments," "in some possible application scenarios," etc., appearing in different places in this specification do not necessarily refer to the same embodiment, but rather to "one or more, but not all, embodiments." Unless otherwise specifically specified, the terms "comprising," "including," "having," and variations thereof mean "including but not limited to."

[0109] In this application, "at least one" means one or more, and "more than" means two or more. "And / or" describes the association of related objects, indicating that three relationships can exist. For example, A and / or B can mean only A, both A and B, or only B, where A and B can be singular or plural. The character " / " typically indicates that the preceding and following related objects are in an "OR" relationship.

[0110] The application scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. Those skilled in the art will understand that, with the evolution of system architecture and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0111] The technical solutions of this application embodiment can be applied to various communication systems, such as Global System for Mobile Communications (GSM), Code Division Multiple Access (CDMA) system, Wideband Code Division Multiple Access (WCDMA) system, General Packet Radio Service (GPRS) system, Long Term Evolution (LTE) system, LTE Frequency Division Duplex (FDD) system, LTE Time Division Duplex (TDD) system, Universal Mobile Telecommunications System (UMTS), Worldwide Interoperability for Microwave Access (WiMAX) system, Wireless Local Area Network (WLAN), Fifth Generation (5G) wireless communication system, New Radio (NR) wireless communication system, Sixth Generation (6G) wireless communication system, or other evolved communication systems.

[0112] To better describe the solutions of the embodiments of this application, the concepts and terms that may be involved in this application will be described below.

[0113] (1) Data collection Data is a crucial component of artificial intelligence (AI) and machine learning (ML) technologies. Data collection refers to the process by which network nodes, management entities, or user experiences (UX) gather data for AI / ML model training, data analysis, and inference.

[0114] (2) AI / ML model training AI / ML model training refers to the process of training an AI / ML model by learning the input / output relationship in a data-driven manner and then using the trained AI / ML model for inference.

[0115] (3) AI / ML model inference The process of using a trained AI / ML model to produce a set of outputs from a set of inputs.

[0116] (4) AI / ML model validation As a sub-process of training, validation is used to evaluate the quality of the AI / ML model using a different dataset than the one used for model training. Validation can help in selecting model parameters that generalize well beyond the dataset used for model training. The trained model parameters can be further tuned through the validation process.

[0117] (5) AI / ML model testing Similar to validation, testing is also a sub-process of training. It is used to evaluate the performance of the final AI / ML model using a different dataset than the datasets used for model training and validation. Unlike AI / ML model validation, testing does not assume subsequent model tuning.

[0118] (6) Online training Online training refers to an AI / ML training process in which the model used for inference is typically trained continuously in (near) real-time as new training samples arrive.

[0119] (7) Offline training Offline training refers to an AI / ML training process in which a model is trained on a collected dataset and then used or delivered for inference.

[0120] (8) AI / ML model delivery / transfer AI / ML model delivery / transfer is a general term referring to the delivery of an AI / ML model from one entity to another in any way. Delivering an AI / ML model over an air interface includes providing parameters of the model structure known to the receiving end, as well as providing a new model with parameters. Delivery can include a complete model or a partial model.

[0121] (9) Life cycle management (LCM) When training and / or inferring AI / ML models on a device, the entire AI / ML process needs to be monitored and managed to ensure the performance gains achieved through AI / ML technology. For example, due to the randomness of wireless channels and the mobility of UEs, the propagation environment of wireless signals changes frequently. However, it is difficult for AI / ML models to maintain optimal performance in all scenarios, and performance may even degrade sharply in some scenarios. Therefore, lifecycle management (LCM) of AI / ML models is crucial for the sustainable operation of AI / ML over the NR air interface. Lifecycle management covers the entire process of applying AI / ML technology on one or more nodes. Specifically, lifecycle management includes at least one of the following sub-processes: data collection, model training, model identification, model registration, model deployment, model configuration, model inference, model selection, model activation, deactivation, model switching, model rollback, model monitoring, model update, model transmission / delivery, and UE capability reporting. Model monitoring can be based on inference accuracy, including metrics related to key performance indicators (KPIs), or on system performance, including metrics related to system performance KPIs, such as accuracy and relevance, overhead, complexity (computational and memory costs), latency (timeliness of monitoring results, from model failure to recovery), and power consumption. Furthermore, data distribution may change after deployment due to environmental variations; therefore, models based on input or output data distribution should also be considered.

[0122] (10) Supervised learning The goal of supervised learning algorithms is to train a model that maps feature vectors (inputs) to labels (outputs) based on training data that includes example feature-label pairs. Supervised learning analyzes the training data and generates an inference function that can be used to map inference data. Supervised learning can be further divided into two types: classification and regression. Classification is used when the output of the AI / ML model is categorical data, i.e., it has two or more classes. Regression is used when the output of the AI / ML model is real numbers or continuous values.

[0123] (11) Unsupervised learning Unlike supervised learning, where AI / ML models learn to map inputs to target outputs, unsupervised methods learn concise representations of input data without labeled data. These representations can be used for data exploration, analysis, or the generation of new data. A typical example of unsupervised learning is clustering, which explores the hidden structure of the input data and provides classification results.

[0124] (12) Reinforcement learning Reinforcement learning is used to solve sequential decision-making problems. It is the process of training an agent's actions based on inputs (states) and feedback signals (rewards) from the environment. In reinforcement learning, the agent interacts with the environment by performing actions to maximize cumulative rewards. Each time the agent performs an action, the current state of the environment may transition to a new state, which in turn brings a corresponding reward. The agent can then perform the next action based on the received reward and the new state in the environment. During the training phase, the agent interacts with the environment to accumulate experience. Because direct interaction with real systems is costly, the environment is typically simulated by a simulator. During the inference phase, the agent can use the optimal decision rules learned from the training phase to achieve the maximum cumulative reward.

[0125] (13) Federal Learning Federated learning (FL) is a machine learning technique used to train AI / ML models by a central node (e.g., a server) and multiple distributed edge nodes (e.g., UEs, next-generation NodeBs (gNBs)). According to wireless FL technology, the server can provide edge nodes with a set of model parameters (e.g., weights, biases, gradients) describing the global AI / ML model. Edge nodes can use the received global AI / ML model parameters to initialize their local AI / ML models. The edge nodes can then use local data samples to train their local AI / ML models, resulting in trained local AI / ML models. Subsequently, the edge nodes can provide the server with a set of AI / ML model parameters describing their local AI / ML models. After receiving multiple sets of AI / ML model parameters describing the corresponding local AI / ML models at multiple edge nodes, the server can aggregate the local AI / ML model parameters reported from multiple UEs and update the global AI / ML model based on this aggregation. Subsequent iterations proceed very similarly to the first iteration. The server can then send the aggregated global model to multiple edge nodes. The above process involves multiple iterations until a global AI / ML model is finally determined, for example, when the AI / ML model converges or meets the training stopping criterion. It's important to note that wireless FL technology does not involve the exchange of local data samples. In fact, local data samples are retained at the corresponding edge nodes.

[0126] AI-based algorithms have been introduced into modern wireless communications to solve wireless problems such as channel estimation, scheduling, channel state information (CSI) compression (from user equipment to base station), multiple-in-multiple-out (MIMO) beamforming, and localization. AI algorithms are a data-driven approach that tunes some predefined architecture using a set of data samples called a training dataset. Recent AI methods use the SGD algorithm to set up neurons to train DNN architectures (including CNNs, RNNs, Transformers, etc.).

[0127] AI technologies (including ML technologies) in communications encompass AI-based communication at the physical layer and / or the MAC layer. At the physical layer, AI communication can aim to optimize component design and / or improve algorithm performance. At the MAC layer, AI / ML-based communication can leverage AI / ML capabilities to learn, predict, and / or make decisions to solve complex optimization problems using potentially better strategies and / or optimal solutions. This includes optimizing features in the MAC layer such as intelligent TRP management, intelligent beam management, intelligent channel resource allocation, intelligent power control, intelligent spectrum utilization, intelligent modulation and coding scheme (MCS), intelligent hybrid automatic repeat request (HARQ) strategies, and intelligent transmit / receive (Tx / Rx) mode adaptation.

[0128] AI architectures can include multiple nodes, which may be organized in either a centralized or distributed mode, both of which can be deployed in access networks, core networks, edge computing systems, or third-party networks. Centralized training and computing architectures are limited by potentially high communication overhead and strict user data privacy. Distributed training and computing architectures can include several frameworks, such as distributed machine learning and federated learning. In some embodiments, the AI ​​architecture may include an intelligent controller, which can perform as a single agent or multiple agents based on joint optimization or individual optimization. New protocols and signaling mechanisms are needed to allow corresponding interface links to be personalized with custom parameters to meet specific needs, while minimizing signaling overhead and maximizing the overall system spectral efficiency through personalized AI technologies.

[0129] The new protocols and signaling mechanisms are provided to operate within and switch between different operating modes, including switching between AI and non-AI modes, and also to provide measurement and feedback to accommodate different possible measurements and information that may require feedback, depending on the implementation.

[0130] Nowadays, it's common for neural network models to become larger and deeper, easily requiring significantly more computing resources than just one or two computers. Most neural network models are trained on powerful computing clouds. Users with the desired neural network architecture, original training dataset, and training objectives may not have sufficient local computing resources to train their models locally. To access a powerful computing cloud, users must completely send all specifications of their neural network architecture, training dataset, and training objectives to the cloud. This requires users to trust the cloud and fully authorize it to manipulate their intellectual property (neural network architecture, training dataset, and training objectives).

[0131] As a data-driven approach, AI-based algorithms inevitably suffer from low generalization ability: if the test data samples are outliers in the training dataset, the neural network will be unable to make good inferences about the test data samples. Even if the AI ​​model is trained on a large dataset, it may not have the necessary knowledge to perform effectively in other environments, especially in wireless communications where channel information changes rapidly.

[0132] In this application, the AI ​​model is exemplified by DNN, i.e., a deep neural network or network. Specific AI models should not be construed as limiting this application.

[0133] Figure 1 This is a schematic diagram of a communication system 100 according to an embodiment of this application.

[0134] refer to Figure 1As a non-limiting illustrative example, a simplified schematic diagram of a communication system is provided. Communication system 100 includes a radio access network 120. Radio access network 120 may be a next-generation (e.g., sixth-generation, 6G, or later) radio access network, or a traditional (e.g., 5G, 4G, 3G, or 2G) radio access network. One or more electronic devices (EDs) 110a to 120j (generally referred to as 110) may be interconnected with each other or connected to one or more network nodes (170a, 170b, generally referred to as 170) in radio access network 120. Core network 130 may be part of the communication system and may depend on or be independent of the radio access technology used in communication system 100. Furthermore, communication system 100 includes a public switched telephone network (PSTN) 140, the Internet 150, and other networks 160.

[0135] Figure 2 This is a schematic diagram of a communication system 100 according to an embodiment of this application.

[0136] Figure 2 An example communication system 100 is illustrated. Generally, communication system 100 enables multiple wireless or wired components to transmit data and other content. The purpose of communication system 100 may be to provide content such as voice, data, video, and / or text via broadcast, multicast, and unicast. Communication system 100 can operate by sharing resources (e.g., carrier spectrum bandwidth) among its constituent units. Communication system 100 may include terrestrial communication systems and / or non-terrestrial communication systems. Communication system 100 can provide a wide range of communication services and applications (e.g., earth monitoring, remote sensing, passive sensing and positioning, navigation and tracking, automated delivery and mobility, etc.). Communication system 100 can provide high availability and robustness through the joint operation of terrestrial and non-terrestrial communication systems. For example, integrating a non-terrestrial communication system (or components thereof) into a terrestrial communication system can enable heterogeneous networks comprising multiple layers. Compared to traditional communication networks, heterogeneous networks can achieve better overall performance through efficient multi-link joint operation, more flexible function sharing, and faster physical layer link switching between terrestrial and non-terrestrial networks.

[0137] Terrestrial and non-terrestrial communication systems can be considered as subsystems of a communication system. In the example shown, communication system 100 includes electronic devices (EDs) 110a to 110d (generally referred to as ED 110), radio access networks (RANs) 120a and 120b, a non-terrestrial communication network 120c, a core network 130, a public switched telephone network (PSTN) 140, the Internet 150, and other networks 160. RANs 120a and RAN 120b include corresponding base stations (BSs) 170a and 170b, which can generally be referred to as terrestrial transmit and receive points (T-TRPs) 170a and 170b. The non-terrestrial communication network 120c includes access nodes 120c, which can generally be referred to as non-terrestrial transmit and receive points (NT-TRPs) 172.

[0138] Alternatively or additionally, any ED 110 can be used to connect to, access, or communicate with any other T-TRP 170a and 170b, NT-TRP 172, Internet 150, core network 130, PSTN 140, other network 160, or any combination thereof. In some examples, ED 110a can communicate uplink and / or downlink with T-TRP 170a via interface 190a. In some examples, ED 110a, ED 110b, and ED 110d can also communicate directly with each other via one or more side-channel air interfaces 190b. In some examples, ED 110d can communicate uplink and / or downlink with NT-TRP 172 via interface 190c.

[0139] Air interfaces 190a and 190b can use similar communication technologies, such as any applicable wireless access technology. For example, communication system 100 can implement one or more channel access methods in air interfaces 190a and 190b, such as code division multiple access (CDMA), time division multiple access (TDMA), frequency division multiple access (FDMA), orthogonal FDMA (OFDMA), or single-carrier FDMA (SC-FDMA). Air interfaces 190a and 190b can utilize other higher-dimensional signal spaces, which may involve combinations of orthogonal and / or non-orthogonal dimensions. Air interface 190c enables communication between ED 110d and one or more NT-TRPs 172 via a wireless link or simply through a link. In some examples, the link is a dedicated connection for unicast transmission, a connection for broadcast transmission, or a connection between a group of EDs and one or more NT-TRPs for multicast transmission.

[0140] RAN 120a and RAN 120b communicate with core network 130 to provide various services, such as voice, data, and other services, to ED 110a, ED 110b, and ED 110c. RAN 120a and RAN 120b and / or core network 130 may communicate directly or indirectly with one or more other RANs (not shown), which may or may not be directly served by core network 130, and may or may not use the same radio access technology as RAN 120a, RAN 120b, or both. Core network 130 may also serve as a gateway access between (i) RAN 120a and RAN 120b or ED 110a, ED 110b, and ED 110c, or both, and (ii) other networks (e.g., PSTN 140, Internet 150, and other networks 160). Additionally, some or all of ED 110a, ED 110b, and ED 110c may include the ability to communicate with different wireless networks via different wireless links using different wireless technologies and / or protocols. ED 110a, ED 110b, and ED 110c may communicate with a service provider or exchange (not shown) via a wired communication channel and with the Internet 150, rather than wirelessly (or also wirelessly). PSTN 140 may include a circuit-switched telephone network for providing plain old telephone service (POTS). The Internet 150 may include a network of computers and subnets (intranets) or both, incorporating protocols such as Internet Protocol (IP), Transmission Control Protocol (TCP), and User Datagram Protocol (UDP). ED 110a, ED 110b, and ED 110c may be multimode devices capable of operating according to multiple wireless access technologies and include multiple transceivers required to support these technologies.

[0141] Figure 3 This is a schematic diagram of ED 110 and base stations 170a, 170b and / or 170c according to an embodiment of this application.

[0142] Figure 3Another example of an ED 110 and base stations 170a, 170b, and / or 170c is shown. The ED 110 is used to connect people, objects, machines, etc. The ED 110 can be widely used in various scenarios, such as cellular communication, device-to-device (D2D), vehicle-to-everything (V2X), peer-to-peer (P2P), machine-to-machine (M2M), machine-type communication (MTC), internet of things (IoT), virtual reality (VR), augmented reality (AR), industrial control, autonomous driving, telemedicine, smart grids, smart furniture, smart offices, smart wearables, smart transportation, smart cities, drones, robots, remote sensing, passive sensing, positioning, navigation and tracking, automated delivery, and mobility.

[0143] Each ED 110 represents any suitable end-user equipment used for wireless operation, which may include (or be referred to as) user equipment / user device (UE), wireless transmit / receive unit (WTRU), mobile station, fixed or mobile subscriber unit, cellular phone, station (STA), machine type communication (MTC) device, personal digital assistant (PDA), smartphone, laptop, computer, tablet, wireless sensor, consumer electronics, smart book, vehicle, car, truck, bus, train, or IoT device, industrial equipment or apparatus of the above (e.g., communication module, modem, or chip), etc. Future generations of ED 110 may be referred to using other terms. Base stations 170a and 170b are T-TRPs and will be referred to as T-TRP 170 below. Similarly, Figure 3 As shown, NT-TRP is referred to below as NT-TRP 172. Each ED 110 connected to T-TRP 170 and / or NT-TRP 172 can be configured to be dynamically or semi-statically turned on (i.e., established, activated, or enabled), turned off (i.e., released, deactivated, or disabled), and / or in response to one or more of connection availability and connection necessity.

[0144] ED 110 includes a transmitter 201 and a receiver 203 coupled to one or more antennas 204. Only one antenna 204 is shown in the figure. One, part, or all of the antennas may also be panels. The transmitter 201 and receiver 203 may, for example, be integrated as a transceiver. The transceiver is used to modulate data or other content for transmission through at least one antenna 204 or a network interface controller (NIC). The transceiver is also used to demodulate data or other content received through at least one antenna 204. Each transceiver includes any suitable structure for generating signals for wireless or wired transmission and / or for processing signals received wirelessly or wiredly. Each antenna 204 includes any suitable structure for transmitting and / or receiving wireless or wired signals.

[0145] ED 110 includes at least one memory 208. Memory 208 stores instructions and data used, generated, or collected by ED 110. For example, memory 208 may store software instructions or modules for implementing some or all of the functions and / or embodiments described herein, and executed by one or more processing units 210. Each memory 208 includes any suitable one or more volatile and / or non-volatile storage and retrieval devices. Any suitable type of memory can be used, such as random access memory (RAM), read-only memory (ROM), hard disk, optical disk, subscriber identity module (SIM) card, memory stick, secure digital (SD) card, and processor cache, etc.

[0146] ED 110 may also include one or more input / output devices (not shown) or interfaces (e.g., to...). Figure 1 (Wired interface of Internet 150 in the network). Input / output devices support interaction with users or other devices on the network. Each input / output device includes any suitable structure for providing or receiving information from the user, such as a speaker, microphone, keypad, keyboard, display, or touchscreen, including network interface communication.

[0147] ED 110 also includes a processor 210 for performing various operations, including operations related to preparing for uplink transmission to NT-TRP 172 and / or T-TRP 170, operations related to processing downlink transmissions received from NT-TRP 172 and / or T-TRP 170, and operations related to processing sidelink transmissions to and from another ED 110. Processing operations related to preparing for uplink transmission may include operations such as encoding, modulation, transmission beamforming, and generating symbols for transmission. Processing operations related to processing downlink transmission may include operations such as receive beamforming, demodulation, and decoding of received symbols. According to an embodiment, receiver 203 may receive downlink transmissions (possibly using receive beamforming), and processor 210 may extract signaling from the downlink transmissions (e.g., by detecting and / or decoding signaling). For example, the signaling may be a reference signal transmitted by NT-TRP 172 and / or T-TRP 170. In some embodiments, processor 276 performs transmit beamforming and / or receive beamforming based on beam direction indications (e.g., beam angle information (BAI)) received from T-TRP 170. In some embodiments, processor 210 may perform operations related to network access (e.g., initial access) and / or downlink synchronization, such as operations related to detecting synchronization sequences, decoding, and acquiring system information. In some embodiments, processor 210 may perform channel estimation, for example, using reference signals received from NT-TRP 172 and / or T-TRP 170.

[0148] Although not shown, processor 210 may be part of transmitter 201 and / or receiver 203. Although not shown, memory 208 may be part of processor 210.

[0149] The processing components in processor 210, transmitter 201, and receiver 203 may be implemented by the same or different processors, which execute instructions stored in memory (e.g., memory 208). Alternatively, some or all of the processing components in processor 210, transmitter 201, and receiver 203 may be implemented using special-purpose circuitry such as a field-programmable gate array (FPGA), a graphics processing unit (GPU), or an application-specific integrated circuit (ASIC).

[0150] In some implementations, T-TRP 170 may be referred to by other names, such as base station, base transceiver station (BTS), wireless base station, network node, network device, network-side device, transmit / receive node, Node B, evolved NodeB (eNodeB or eNB), femtocell, next-generation NodeB (gNB), transmission point (TP), site controller, access point (AP) or wireless router, relay station, remote radio head, ground node, ground network device or ground base station, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), location node, and other possibilities. T-TRP 170 can be a macro BS, micro BS, relay node, host node, or a combination thereof. T-TRP 170 may refer to the aforementioned equipment or to a device within the aforementioned equipment (e.g., a communication module, modem, or chip).

[0151] In some embodiments, the various parts of T-TRP 170 may be distributed. For example, some modules of T-TRP 170 may be located remotely from the device housing the antenna of T-TRP 170 and may be coupled to the device housing the antenna via a communication link (not shown), sometimes referred to as a fronthaul, such as the Common Public Radio Interface (CPRI). Therefore, in some embodiments, the term T-TRP 170 may also refer to modules on the network side that perform processing operations such as ED 110 location determination, resource allocation (scheduling), message generation, and encoding / decoding; these modules are not necessarily part of the device housing the antenna of T-TRP 170. These modules may also be coupled to other T-TRPs. In some embodiments, T-TRP 170 may actually be multiple T-TRPs that work together, for example, through coordinated multicast transmissions, to serve ED 110.

[0152] T-TRP 170 includes at least one transmitter 252 and at least one receiver 254 coupled to one or more antennas 256. Only one antenna 256 is shown in the figure. One, part or all of the antennas may also be panels. The transmitter 252 and receiver 254 may be integrated as a transceiver. T-TRP 170 also includes a processor 260 for performing various operations, including operations related to: preparing transmissions for downlink transmissions to ED 110, processing uplink transmissions received from ED 110, preparing transmissions for backhaul transmissions to NT-TRP 172, and processing transmissions received from NT-TRP 172 via backhaul. Processing operations related to preparing transmissions for downlink or backhaul transmissions may include operations such as encoding, modulation, precoding (e.g., MIMO precoding), transmit beamforming, and generating symbols for transmission. Processing operations related to processing received transmissions in the uplink or backhaul may include operations such as receive beamforming, demodulation, and decoding of received symbols. Processor 260 can also perform operations related to network access (e.g., initial access) and / or downlink synchronization, such as generating the contents of a synchronization signal block (SSB), generating system information, etc. In some embodiments, processor 260 also generates beam direction indications, such as BAI, which can be scheduled for transmission by scheduler 253. Processor 260 performs other network-side processing operations described herein, such as determining the location of ED 110, determining the location for deploying NT-TRP 172, etc. In some embodiments, processor 260 can generate signaling, for example, for configuring one or more parameters of ED 110 and / or one or more parameters of NT-TRP 172. Any signaling generated by processor 260 is transmitted by transmitter 252. Note that the term "signaling" as used herein may also be referred to as control signaling. Dynamic signaling can be transmitted in control channels, such as the physical downlink control channel (PDCCH). Static or semi-static higher-layer signaling can be included in packets transmitted in data channels such as the physical downlink shared channel (PDSCH).

[0153] Scheduler 253 may be coupled to processor 260. Scheduler 253 may be included within or operate separately from T-TRP 170, and may schedule uplink, downlink, and / or backhaul transmissions, including issuing scheduling authorizations and / or configuring unscheduled (“configured authorization”) resources. T-TRP 170 also includes memory 258 for storing information and data. Memory 258 stores instructions and data used, generated, or collected by T-TRP 170. For example, memory 258 may store software instructions or modules executed by processor 260 for implementing some or all of the functions and / or embodiments described herein.

[0154] Although not shown, processor 260 may form part of transmitter 252 and / or receiver 254. Furthermore, processor 260 may implement scheduler 253, which is not shown in the figure. Although not shown, memory 258 may be part of processor 260.

[0155] The processing components in processor 260, scheduler 253, transmitter 252, and receiver 254 can be implemented by the same or different processors, which execute instructions stored in memory, such as instructions in memory 258. Alternatively, some or all of the processing components in processor 260, scheduler 253, transmitter 252, and receiver 254 can be implemented using dedicated circuitry, such as FPGA, GPU, or ASIC.

[0156] Although the NT-TRP 172 is shown as a drone only as an example, it can be implemented in any suitable non-terrestrial form. Furthermore, in some implementations, the NT-TRP 172 may be referred to by other names, such as a non-terrestrial node, a non-terrestrial network device, or a non-terrestrial base station. The NT-TRP 172 includes a transmitter 272 and a receiver 274 coupled to one or more antennas 280. Only one antenna 280 is shown in the figure. One, some, or all of the antennas may also be panels. The transmitter 272 and receiver 274 may be integrated as a transceiver. The NT-TRP 172 also includes a processor 276 for performing various operations, including operations related to: preparing transmissions for downlink transmissions to ED 110, processing uplink transmissions received from ED 110, preparing transmissions for backhaul transmissions to T-TRP 170, and processing transmissions received from T-TRP 170 via backhaul. Processing operations related to preparing for downlink or backhaul transmissions may include operations such as encoding, modulation, precoding (e.g., MIMO precoding), transmit beamforming, and generating symbols for transmission. Processing operations related to receiving transmissions in the uplink or backhaul may include operations such as receive beamforming, demodulation, and decoding of received symbols. In some embodiments, processor 276 performs transmit beamforming and / or receive beamforming based on beam direction information (e.g., BAI) received from T-TRP 170. In some embodiments, processor 276 may generate signaling, for example, to configure one or more parameters of ED 110. In some embodiments, NT-TRP 172 implements physical layer processing but does not implement higher-level functions such as medium access control (MAC) or radio link control (RLC) layer functions. Since this is only an example, NT-TRP 172 may generally implement higher-level functions in addition to physical layer processing.

[0157] The NT-TRP 172 also includes a memory 278 for storing information and data. Although not shown, a processor 276 may be part of the transmitter 272 and / or receiver 274. Although not shown, the memory 278 may be part of the processor 276.

[0158] The processing components in processor 276, transmitter 272, and receiver 274 may be implemented by the same or different processors, which execute instructions stored in memory, such as instructions in memory 278. Alternatively, some or all of the processing components in processor 276, transmitter 272, and receiver 274 may be implemented using dedicated circuitry, such as a programmed FPGA, GPU, or ASIC. In some embodiments, NT-TRP 172 may actually be multiple NT-TRPs that work together, for example, through coordinated multipoint transmissions, to serve ED 110.

[0159] T-TRP 170, NT-TRP 172 and / or ED 110 may include other components, but these components have been omitted for clarity.

[0160] Figure 4 This is a schematic diagram of a unit or module in the device according to an embodiment of this application.

[0161] according to Figure 4 One or more steps in the provided embodiment method can be executed by the corresponding unit or module. Figure 4 The diagram illustrates units or modules within a device, such as in ED 110, T-TRP 170, or NT-TRP 172. For example, signals may be transmitted by a transmitting unit or transmitting module. Signals may be received by a receiving unit or receiving module. Signals may be processed by a processing unit or processing module. Other steps may be performed by an artificial intelligence (AI) module or a machine learning (ML) module. The corresponding units or modules may be implemented using hardware, one or more components or devices executing software, or a combination thereof. For example, one or more of these units or modules may be integrated circuits, such as a programmed FPGA, GPU, or ASIC. It should be understood that if these modules are implemented by a processor using software, these modules may be retrieved by the processor, wholly or partially, individually or collectively, for processing, in one or more instances, and these modules themselves may include instructions for further deployment and instantiation.

[0162] Further details regarding ED 110, T-TRP 170, and NT-TRP 172 are known to those skilled in the art. Therefore, these details are omitted herein.

[0163] Figure 5 This is a schematic diagram of an AI-based communication device.

[0164] The wireless system includes multiple connected devices. Device 500 is a base station (BS) or user equipment (UE). Device 500 may have three systems: a sensing system 510, a communication system 520, and / or an AI system 530. The sensing system 510 senses and collects signals and data, the communication system 520 transmits and receives signals and data, and the AI ​​system 530 trains and infers AI implementations. An exemplary AI implementation is based on two cycles of deep learning: a training cycle and an inference cycle. In some possible application scenarios, the training cycle may also be called a learning cycle, and the inference cycle may also be called an inference cycle.

[0165] Deep learning consists of two cycles: training (or learning) and inference (or deduction). During the training cycle, the coefficients of neurons are learned from training data to achieve a specific training goal or target. During the inference cycle, input data samples are fed into the trained neural network, which then outputs a prediction.

[0166] During the training cycle, the AI ​​system 530 of device 500 can train one or more DNNs, wherein the perception system 510 of device 500 can generate signals and / or data. The communication system 520 of device 500 can receive signals or data from one or more other devices. During and / or after the AI ​​system 530 completes training, the device's communication can send the training results to one or more other devices.

[0167] During the inference cycle, the AI ​​system 530 of device 500 can utilize one or more DNNs to perform one or more inferences to complete one or more tasks. The perception system 510 of device 500 can generate signals and / or data, and the communication system 520 of device 500 can receive signals or data from one or more other devices. After the AI ​​system 530 of device 500 completes the inference, the communication system 520 of device 500 can send the inference results to one or more other devices.

[0168] AI implementations can switch between these two cycles, or remain in both cycles simultaneously. For example, the AI ​​system 530 of device 500 can train a second DNN, but still perform inference on the first DNN.

[0169] During the training cycle, the AI ​​system 530 of device 500 can operate in single-user mode. In this mode, the AI ​​system 530 trains one or more DNNs using data provided by the perception system 510 of device 500. Examples of data include local perception data and local channel data. Local perception data includes RGB data, light detection and ranging (LiDAR) data, temperature data, barometric pressure data, power outage data, etc. Local channel data includes channel state information (CSI), received signal strength indicator (RSSI), latency data, etc.

[0170] Optionally, the AI ​​system 530 of device 500 can operate in a collaborative mode. In this mode, the AI ​​system 530 trains one or more DNNs using data received by the communication system 520 of device 500. Example data includes perceptual data, channel data, neuron data, and potential output data. Perceptual data includes RGB data, LiDAR data, temperature data, barometric pressure data, power outage data, etc. Channel data includes CSI, RSSI, latency data, etc. Neuron data includes multiple neurons or multiple gradients. Potential output data includes multiple potential outputs.

[0171] Figure 6 This is a schematic diagram of device 500 receiving reference data samples from device 600 according to an embodiment of this application. In collaborative mode, the AI ​​system 530 of device 500 can use the following data: accumulating the perception data received by the communication system 520 of device 500 into a training dataset; accumulating the channel data received by the communication system 520 of device 500 into a training dataset; setting local neurons using neurons received by the communication system 520 of device 500, which is a typical joint learning scheme; and inputting the potential output received by the communication system 520 of device 500 into its DNN.

[0172] Alternatively, in collaborative mode, the AI ​​system 530 of device 500 can use the data received by the communication system 520 of device 500 and its local data, for example: mixing the local perception data provided by the perception system 510 of device 500 with the perception data received by the communication system 520 of device 500 into a training dataset; mixing the local channel data provided by the perception system 510 of device 500 with the channel data received by the communication system 520 of device 500 into a training dataset; averaging the local neurons owned by the AI ​​system 530 of device 500 with the neurons received by the communication system 520 of device 500, which is a typical joint learning scheme; averaging the local latent outputs owned by the AI ​​system 530 of device 500 and inputting them into its DNN.

[0173] Figure 7 This is a schematic diagram of reference data samples consisting of multiple groups, according to an embodiment of this application. During a training cycle, the communication system 520 of device 500 can receive some reference data samples in single-user or cooperative mode. Some devices transmit reference data samples in broadcast, multicast, or unicast channels. Other devices transmit one or more indicators about which layer(s) the reference data samples are associated with, where, for example, there are three groups of reference data samples: the first group of reference data samples is indicated to be associated with the input layer of the DNN, the second group of reference data samples is indicated to be associated with the output of a latent layer of the DNN, and the third group of reference data samples is indicated to be associated with the layer output of the DNN.

[0174] The AI ​​system 530 of device 500 can measure the distance between its local data samples and reference data samples in groups. The AI ​​system 530 of device 500 can sample its local layer inputs, local latent layer outputs, and / or layer outputs randomly, non-randomly, uniformly, or non-uniformly. Then, the AI ​​system 530 of device 500 measures the distance between the local samples and the reference samples received by the communication system 520 of device 500. If the average distance across all groups is consistently below one or more predefined thresholds, the AI ​​system 530 of device 500 can indicate that the current training process is working as expected; otherwise, the AI ​​system 530 can indicate an anomaly.

[0175] Even when a device lacks an AI system but possesses a perception and communication system, its perception system can still measure the distance between its local data samples and reference data samples associated with the DNN layer inputs. If the average distance across the layer inputs is below a predefined threshold, the device's perception system can consider the data captured as "good" data; otherwise, it considers it as bad data. The device's communication system can either send only good data to other devices while omitting bad data, or it can label the perceived data with distances before sending it to other devices.

[0176] The UE can report information about its data to the BS, which then determines whether the data differs significantly from the training data. If the difference is too large, the BS can switch its operating mode from AI mode to non-AI mode, or switch to another AI model. However, the UE directly reporting raw data may be considered an infringement of user privacy. Transmitting raw data over the air is inefficient and may also violate privacy policies. Therefore, a pressing technical challenge is how to securely and efficiently transmit data status information.

[0177] To protect the original data and save bandwidth, a set of reference data samples is encoded or compressed into a space with a lower dimension than its original space. The encoder or compressor can be linear or nonlinear. Linear encoders can be implemented using standard bases such as Fourier bases, DCT, wavelets, or custom bases. These bases can consist of a unitary matrix (orthogonal normalized). Nonlinear encoders can be implemented using DNNs. Figure 8 This is a schematic diagram of an approximation based on DNN in an embodiment of this application.

[0178] Unlike traditional compression schemes built for reliable reconstruction, when data is compressed to a lower-dimensional space, the encoder intentionally avoids reliable reconstruction but preserves as much topological distance as possible. In other words, after encoding to a lower-dimensional space, the relative distance between two data samples in their original signal space is well preserved.

[0179] Figure 9 This is a flowchart of a communication method according to an embodiment of this application.

[0180] In 710, the first coefficient is sent.

[0181] The first coefficient is determined based on the first data and the reference basis, and the dimension of the first coefficient is smaller than the dimension of the first data.

[0182] The first data includes monitoring or measurement data from user equipment or network equipment. Further, the first data is monitoring or measurement data related to an AI model. In this embodiment, the network equipment can be a BS (Browser / Server). If the first data is data sent from the UE uplink to the BS, then the data is the UE's monitoring or measurement data. If the first data is data sent from the BS downlink to the UE, then the data is the BS's monitoring or measurement data.

[0183] One or more reference bases are predefined or configured. A reference base is one of a plurality of predefined or configured reference bases. For example, a reference base can be configured by the BS to the UE. A reference base can be an orthogonal basis, and any two columns of the reference base are completely orthogonal to each other. A typical orthogonal basis is a DFT basis.

[0184] In 720, communication is based on the first coefficient.

[0185] Figure 10 This is a flowchart of a communication method according to an embodiment of this application. To protect the original data and save bandwidth, a set of reference data samples is encoded or compressed into a space with a lower dimension than its original space. The encoder or compressor can be linear or nonlinear. A linear encoder can be implemented using some standard basis, such as Fourier basis, discrete cosine transform (DCT), or wavelet. Alternatively, a linear encoder can have some custom basis, which can consist of a unitary matrix (orthogonal normalized). A nonlinear encoder can be implemented using some DNN.

[0186] In the embodiments given below, the UE projects the high-dimensional signal onto the low-dimensional signal (coefficients) through a transformation (orthogonal normalized basis U). Reporting coefficients instead of raw data is more efficient and better for privacy.

[0187] In 810, one or more reference bases are configured or predefined.

[0188] The coefficient of reference basis indicator (CRBI) is used to indicate coefficients relative to a reference basis (e.g., an orthogonal basis). Suppose {u1, u2, …, ur} is an orthogonal set of vectors in a subspace Rn. This set forms the basis U of the subspace Rn. The elements represented by basis U in the subspace Rn can be written as a finite weighted linear combination of these basis elements. The coefficients of this weighted linear combination are called the components or coordinates of the vectors relative to basis U. ).

[0189] Figure 11 This is a schematic diagram illustrating the projection of a high-dimensional signal onto a low-dimensional signal according to an embodiment of this application. For example, ,in Let U be an n×1 primitive space, and U be an n×r orthogonal basis. Let n be an r×1 spectral subspace, where n is an integer greater than 1 and r < 1. <n。 This refers to the data that the UE needs to report, such as sensing data, measurement data, AI / machine learning (ML) data, channel data, and environmental data. U is both a reference basis and an orthogonal basis; any two columns of U are completely orthogonal to each other. Embodiments of this application can use columns as a basis, which can be easily applied to a basis matrix with rows as the basis, abbreviated as UH. A typical orthogonal basis is the Discrete Fourier Transform (DFT) basis. CRBI is the reference coefficient.

[0190] Denoted as an n×1 reference sample, and U is an n×r matrix. It can be represented by a weighted linear combination of each column of U: , where are r×1 spectral coefficients or weights. In the case where r << n, is equivalent low-dimensional space signal (vector). The matrix U is a unitary matrix, where , . Then, the matrix UH is an encoder or compressor that compresses the high-dimensional (n×1) reference sample into a low-dimensional (r×1) .

[0191] In a possible implementation scenario, multiple reference bases (UA, UB, UC...) are configured or predefined. The BS configures the reference base to be used, such as UX. The UE reports the CRBI according to UX. According to the formula , the UE knows U and , so the coefficient can be calculated.

[0192] In a possible implementation scenario, a reference matrix U is configured or predefined, and one or more pruning bases are indicated or predefined as reference bases. The reference matrix Y is a matrix with M rows and N columns. The pruning bases of the reference base are K columns of Y, such as the first K columns of Y, and K is configured, K ≤ N. Optionally, it can be specified which K columns of Y are selected as the pruning bases.

[0193] In 820, the UE determines the coefficients of its reference base.

[0194] Configure or predefine a reference base (U). The BS can configure one or more reference signals, and the UE can obtain the original data by measuring the reference signals . Optionally, the reference signals can also not be configured, and the UE can obtain the original data by sensing . The UE determines its own CRBI through . U is a unitary matrix, which satisfies that the conjugate transpose of the matrix is equal to the inverse of the matrix, that is, , and I is the identity matrix.

[0195] The UE can obtain one or more reported data from a single timeslot. Based on the observation interval in time (or without restriction), the UE should obtain the CRBI values ​​reported in the uplink timeslot. For example, the UE reports the CRBI value in uplink timeslot n. The UE can obtain one or more corresponding CRBI values ​​by measuring data within a configured time window n-5 to n-1. The UE can choose to report multiple CRBI values, or it can choose to report the average / maximum / minimum of multiple CRBI values.

[0196] In 830, the UE reports CRBI or CRBI index.

[0197] For example, the UE can acquire P reported data points within a time window from n-5 to n-1, which can be achieved through... Obtain P CRBI values ​​corresponding to P reported data points. The UE can choose to report the average, maximum, or minimum of the P CRBI values. The reported data includes the UE's monitoring data or measurement data.

[0198] The UE can directly report CRBI or report the index corresponding to the CRBI. The base station can configure a physical uplink control channel (PUCCH) or a physical uplink shared channel (PUSCH) for the UE to report CRBI. CRBI reporting supports periodic reporting, non-periodic reporting, and semi-static reporting.

[0199] In some potential application scenarios, the UE reports the index corresponding to the CRBI. In this scenario, one or more CRBI tables are predefined or configured. A reference base can be associated with one CRBI table or multiple CRBI tables. When a reference base is associated with multiple CRBI tables, the BS indicates which CRBI table to use.

[0200] The CRBI indexes of the CRBI table are reported by the UE. As shown in Table 1, the CRBI index is represented by 4 bits. Although all CRBI values ​​in Table 1 are represented as the same {c0, c1, …, cr}, each CRBI index corresponds to a different CRBI value. In some possible implementations, the value of {c0, c1, …, cr} is different in different rows of the CRBI table; for example, some are {c0, c1, …, c5}, and some are {c0, c1, …, c6}.

[0201] Table 1

[0202] In some possible implementations, a CRBI index may correspond to a CRBI range, and Table 1 should not be construed as a limitation of this application.

[0203] The communication method provided in this embodiment allows the UE to report its data to the BS with minimal air interface overhead. The BS then determines whether the data differs significantly from the training data, which improves the efficiency of data reporting while protecting data privacy.

[0204] Figure 12 This is a flowchart of a communication method according to an embodiment of this application. In this embodiment, differential CRBI indexing can be used for reporting.

[0205] In 910, the reference CRBI index is determined.

[0206] The reference CRBI index can be indicated by the BS, or it can be configured or predefined.

[0207] In 920, the offset is reported to the BS.

[0208] The UE reports the offset to the BS. Based on the offset and the reference CRBI index, the BS can determine the CRBI index of the current data. For example, the differential CRBI can be obtained through equation (1).

[0209] Offset = Current data CRBI index - Reference CRBI index (1) The communication method provided in this embodiment allows the UE to report its data to the BS with minimal air interface overhead. The BS then determines whether the data differs significantly from the training data, which improves the efficiency of data reporting while protecting data privacy.

[0210] Furthermore, the communication method provided in this application can also be applied to downlink (DL) transmissions where the BS instructs the UE to CRBI or CRBI index, thereby indicating data information on the BS side. For specific implementation details, please refer to [link / reference]. Figures 9 to 12 The description of the subject matter will not be repeated in this application.

[0211] Figure 13 This is a schematic diagram of the determination matrix U in an embodiment of this application.

[0212] Each column of matrix U can be a standard basis such as a Fourier basis, DCT basis, or wavelet basis. Alternatively, the r columns of matrix U can be constructed on the distribution of a set of reference samples x. , An example of calculating matrix U on the distribution of ... can be shown below: Accumulate a sufficient number (M) of n×1 samples , ... , M << n; juxtapose them into an n×M matrix , the order of the data samples is not important; apply the descending singular value decomposition (SVD) to : , where U is an n×r unitary (orthogonal) matrix representing the commonality among all M reference samples .

[0213] Since a set of reference data samples corresponds to one layer output, each set of reference data samples has its own matrix U. The first set has matrix U1 and a compressed version , the second set has matrix U2 and a compressed version .

[0214] The communication system of the device receives the first matrix U1 and the first set of reference samples (after compression) , as well as the second matrix U2 and the second set of reference samples (after compression) .

[0215] Figure 14 is a schematic diagram of the first sampling matrix P1 in an embodiment of this application.

[0216] The first matrix U1 is n1×r1, and the second matrix U2 is n2×r2. If n1 and / or n2 are very large numbers, the first sampling matrix P1 can be applied to the first matrix U1, and the second sampling matrix P2 can be applied to the second matrix U2. The first sampling matrix P1 is m1×n1 (m1 << n1), and each row has only one "1", indicating the position to be sampled in . The second sampling matrix P2 is m2×n2 (m2 << n2), and each row has only one "1", indicating

[0217] Figure 15 is a schematic diagram of the sampling matrix compressing matrix U in an embodiment of this application. [[ID=,44]]

[0218] In one possible implementation, the device's communication system receives a first compact matrix θ1, a first sampling matrix P1, and a first set of compressed reference samples. The device's communication system receives the second compact matrix θ2, the second sampling matrix P2, and the second set of compressed reference samples. .

[0219] Alternatively, the device's communication system receives the left inverse of the first compact matrix. The first sampling matrix P1 and the first set of reference samples (compressed) The device's communication system receives the inverse of the second compact matrix. The second sampling matrix P2 and the second set of reference samples (compressed) .

[0220] Figure 16 This is a schematic diagram of the scoring distance in the low-frequency space according to an embodiment of this application.

[0221] The device's communication system can receive measurements of the first two samples. and The first scoring function for the distance between The device's communication system can receive measurements of two samples from the second group. and The second scoring function for the distance between The first rating function d1 and the second rating function d2 can be the same or different. The first rating function... Second scoring function This could be a dot product, inner product, Euclidean distance, etc. Alternatively, it could be the first scoring function. Second scoring function It can be based on DNN.

[0222] Alternatively, the device's communication system can receive measurements of the two distributions in the first group. and The first scoring function for the distance between The device's communication system can receive measurements from two distributions in the second group. and The second scoring function for the distance between The first rating function d1 and the second rating function d2 can be the same or different. The first rating function... Second scoring function This could be mutual information, the Hilbert-Schmidt independence criterion (HSIC) measure, KL divergence, graph edit distance, Wasserstein distance, Jensen-Shannon divergence (JSD), etc. Alternatively, a first scoring function could be used. Second scoring function It can be based on DNN.

[0223] Due to generalization issues, the UE or BS needs to detect its generalization performance and then switch to an appropriate AI model. For example, there are multiple AI models for various scenarios, including urban indoor, urban outdoor, rural, and high-speed rail scenarios. When the surrounding environment changes, the BS or UE needs to switch to another model. In existing technologies, when inference performance degrades, the UE or BS switches its AI model or falls back to a non-AI mode. This is a passive solution. This application provides various event-triggered designs to enable proactive model switching.

[0224] The BS configures multiple candidate AI / ML models for the UE, with each model having a model index. The configuration signal can be radio resource control (RRC), medium access control-control element (MAC-CE), or downlink control information (DCI), and can be broadcast, multicast, or unicast.

[0225] BS configures associated data anchors for each candidate AI / ML model. Anchors are sets of reference data, such as reference coefficients (…). The reference data are the reference coefficients of the reference base, which is configured or predefined. The set can be a vector, such as a one-dimensional array, where the vector has a size of r, which is predefined or configured. The set has a size of K (cj, j=1,2,…K), where K is predefined or configured. Each of the N anchors includes K reference coefficients.

[0226] The association between data anchors and candidate AI / ML models can be implicitly determined or explicitly configured. For example, the association between a data anchor and a candidate AI / ML model is implicitly determined by giving the anchor index the same value as the model index, i.e., {model index k, anchor index k}. For example, the association between a data anchor and a candidate AI / ML model is explicitly configured, where the BS configures a data anchor j associated with a candidate model k, i.e., {model index k, anchor index j}.

[0227] Reference data (e.g., coefficients) The coefficients are the reference coefficients of the reference basis (orthogonal normalized basis U). During the information exchange between the UE and the BS, the UE can project the high-dimensional signal onto the low-dimensional signal (coefficients) through the transformation (orthogonal normalized basis U). In ), the transformation equation is: ,in For the reported data, U is the reference base. is a reference coefficient. One column of U is one of the bases, which means that any two columns of U are perfectly orthogonal to each other.

[0228] Configure or predefine a reference base (U). For example, the BS can configure a reference signal with respect to the reference base. Optionally, this reference signal can also be sensed by the UE. The UE communicates with the UE via... Determine its coefficient of reference basis indicator (CRBI). U is a unitary matrix that satisfies the condition that the conjugate transpose of the matrix is ​​equal to the inverse of the matrix, i.e., I is the identity matrix.

[0229] UE calculates its data (e.g., coefficients) The difference between the reference data cj (j=1,2,…K) in the anchor point and the reference data cj in the anchor point can be calculated by a method or function indicated by BS or predefined. For example, the difference can be calculated by any one of equations (2), (3) and (4). It is the reference data in the reported data and anchor points. The difference between them. This is the data reported by the UE. It is the j-th reference data in the anchor point. <> represents the inner product. This represents the magnitude of the vector. This indicates other user-defined functions. , .

[0230] (2) (3) (4) Equations (2) to (4) are merely examples; the UE calculates its data (e.g., coefficients). The difference between the reference data cj (j=1,2,…K) in the anchor point and the reference data cj in the anchor point can also be calculated by dot product, Euclidean distance or DNN-based algorithm, etc. The above embodiments should not be construed as limiting the present application.

[0231] An anchor point is a set of reference data. The UE calculates the difference between its data and the anchor point according to a method that the BS can indicate or predefined, such as equation (4) or (5). It is the difference between the reported data and the anchor point. It can be reporting data K reference data in the anchor point The minimum value of the difference between them can also be the reported data. K reference data in the anchor point The average of the differences between them.

[0232] (5) (6) Alternatively, the difference between the reported data and the anchor point can be obtained through mutual information, the Hilbert-Schmidt independence criterion (HSIC) measure, Kullback-Leibler (KL) divergence, graph edit distance, Wasserstein distance, Jensen-Shannon divergence (JSD) distance, DNN-based algorithms, etc.

[0233] For AI / ML pattern monitors, the BS configures the monitor reporting method to be event-triggered. One of the multiple AI model monitoring events can be configured as event K1, event K2, event K3, event K4, etc.

[0234] Figure 17 This is a flowchart of a communication method according to an embodiment of this application.

[0235] In 1510, when the entry or exit conditions of an event are met, a report instructing the AI ​​model to switch or the mode to switch is sent.

[0236] This application defines the following events: Event K1: the operation of switching from an AI model to a non-AI model; Event K2: the operation of switching from one AI / ML model (Model 1) to another AI / ML model (Model 2); Event K3: the operation of switching from a non-AI mode to an AI mode; Event K4: the operation of switching from one AI / ML model (Model 2) to another AI / ML model (Model 1).

[0237] The method provided in this application defines multiple events and corresponding entry and exit conditions for these events. When these conditions are met, active model or mode switching can be performed.

[0238] In 1520, communication is conducted based on reports.

[0239] In event K1, the entry or exit condition requires determining whether the difference between the first data and the first anchor point is greater than a threshold. The first anchor point can be indicated by the BS, for example, the first anchor point is the data anchor point of the UE's current AI / ML model. Alternatively, the first anchor point is the data anchor point closest to the UE's first data. The first data includes the UE's monitoring data or measurement data. In one possible implementation, the first data includes any one or more of the following: sensing data, measurement data, channel data, neural network data of the AI ​​model, and potential output data of the AI ​​model.

[0240] Event K1 indicates that the current or best AI / ML model may be outdated, and the UE should switch to non-AI mode.

[0241] The entry condition for event K1 is considered satisfied when the specified condition K1-1 is met.

[0242] Diff11–Hys11>Thresh11(K1-1) Diff11 is the difference between the first data point and the first anchor point. Hys11 is the hysteresis parameter for event K1, predefined or configured by the BS. Thresh11 is the threshold parameter for event K1, predefined or configured by the BS. The units of Thresh11 and Hys11 are the same as those of Diff11.

[0243] The departure condition of event K1 is considered to be met when the following conditions K1-2 are satisfied.

[0244] Diff11+Hys12 <Thresh12(K1-2) Hys12 is the hysteresis parameter for event K1, predefined or configured by the BS. Thresh12 is the threshold parameter for event K1, predefined or configured by the BS. The units of Hys12 and Thresh12 are the same as those of Diff11. Hys11 and Hys12 can be the same or different; for example, they can be similar. Thresh11 and Thresh12 can be the same or different; for example, they can be similar.

[0245] This is determined by whether the BS configuration parameter "reportOnLeave" is enabled or disabled. If disabled, no report will be sent when the leave condition is met.

[0246] In addition, BS also configures a parameter called "trigger time", which specifies the range of values ​​for the time used to trigger the parameter and is related to the time when specific criteria for the event need to be met to trigger a report.

[0247] Another possible entry or exit condition for K1 is the presence of an additional offset.

[0248] The entry condition for event K1 is considered satisfied when the conditions K1-3 specified below are met.

[0249] Diff11+offset11–Hys11>Thresh11(K1-3) The exit condition of event K1 is considered to be met when the conditions K1-4 specified below are satisfied.

[0250] Diff11+offset12+Hys12 <Thresh12(K1-4) offset11 and offset12 can be the same or different. The values ​​of offset11 and offset12 can be positive or negative.

[0251] In some possible implementations, the value of Hys11 or Hys12 can be set to 0, or it may not be predefined or configured.

[0252] The method provided in this application defines an event K1 and the corresponding entry and exit conditions for event K1. When any one of these conditions is met, the mode can be actively switched.

[0253] Event K2 indicates that the current AI / ML model (referred to as Model 1) may be outdated, and the UE should switch to another model (Model 2).

[0254] In event K2, the current model becomes worse than Thresh21, while another model becomes better than Thresh22. "Worse than Thresh21" means the second difference between the first data point and the second anchor point is greater than Thresh21. "Better than Thresh22" means the third difference between the first data point and the third anchor point is less than Thresh22. The second anchor point is the data anchor point associated with the first AI model (Model 1), and the third anchor point is the data anchor point associated with the second AI model (Model 2).

[0255] The entry condition for event K2 is considered satisfied when the specified condition K2-1 is met.

[0256] Diff21–Hys21>Thresh21 and Diff22+Hys22 <Thresh22(K2-1) Diff21 is the difference between the first data point and the second anchor point, where the second anchor point can be pre-configured, such as the data anchor point associated with model 1. Diff22 is the difference between the first data point and the third anchor point, where the third anchor point can be pre-configured, such as the data anchor point associated with model 2. Hys21 and Hys22 are predefined or configured parameters corresponding to event K2. Thresh21 is a predefined or configured threshold corresponding to the first AI model, and Thresh22 is a predefined or configured threshold corresponding to the second AI model. Thresh21 and Thresh22 can be configured to be the same or different.

[0257] The exit condition of event K2 is considered to be met when either condition K2-2 or K2-3 specified below is satisfied.

[0258] Diff21+Hys23 <Thresh23(K2-2) Diff22–Hys24>Thresh24(K2-3) Condition K2-2 indicates that the first AI model is better than threshold23. Condition K2-3 indicates that the second AI model is worse than threshold24. If either condition K2-2 or K2-3 is met, the UE leaves event K2.

[0259] Hys23 and Hys24 are predefined or configured parameters corresponding to event K2. Thresh23 is a predefined or configured threshold corresponding to the first AI model, and Thresh24 is a predefined or configured threshold corresponding to the second AI model. Hys21 and Hys23 can be the same or different; for example, they can be similar. Hys22 and Hys24 can be the same or different; for example, they can be similar. Thresh21 and Thresh23 can be the same or different; for example, they can be similar. Thresh22 and Thresh24 can be the same or different; for example, they can be similar.

[0260] This is determined by whether the BS configuration parameter "reportOnLeave" is enabled or disabled. If disabled, no report will be sent when the leave condition is met.

[0261] In addition, BS also configures a parameter called "trigger time", which specifies the range of values ​​for the time used to trigger the parameter and is related to the time when specific criteria for the event need to be met to trigger a report.

[0262] Another possible entry or exit condition for K2 is the presence of an additional offset.

[0263] The entry condition for event K2 is considered satisfied when the conditions K2-4 specified below are met.

[0264] Diff21+offset21–Hys21>Thresh21 and Diff22+offset22+Hys21 <Thresh22(K2-4) The exit condition of event K2 is considered to be met when either condition K2-5 or K2-6 specified below is satisfied.

[0265] Diff21+offset23+Hys23 <Thresh23(K2-5) Diff22+offset24–Hys24>Thresh24(K2-6) offset21 and offset23 can be the same or different; for example, they can be similar. offset22 and offset24 can also be the same or different; for example, they can be similar. The values ​​of offset21, offset22, offset23, and offset24 can be positive or negative.

[0266] In some possible implementations, the values ​​of Hys21, Hys22, Hys23, or Hys24 can be set to 0, or they may not be predefined or configured.

[0267] The method provided in this application defines event K2 and the corresponding entry and exit conditions for event K2. When any one of these conditions is met, the active switching of the model can be achieved.

[0268] Event K3 indicates that the AI / ML model has become better than the threshold, and the UE should switch from non-AI mode to AI mode.

[0269] In event K3, the fourth difference between the first data and the fourth anchor point is less than a threshold. The fourth anchor point can be indicated by BS, or the fourth anchor point is the data anchor point closest to the first data; that is, the fourth anchor point is the data anchor point with the smallest difference from the first data among multiple anchor points.

[0270] The entry condition for event K3 is considered satisfied when the specified condition K3-1 is met.

[0271] Diff31+Hys31 <Thresh31(K3-1) Diff31 is the difference between the first data point and the fourth anchor point. Hys31 is a predefined or configured hysteresis parameter for event K3. Thresh31 is a predefined or configured threshold for event K3. The units of Thresh31 and Hys31 are the same as those of Diff31.

[0272] When the entry condition of event K3 is met, the UE can report the index of the nearest anchor point.

[0273] The departure condition of event K3 is considered to be met when the specified condition K3-2 is satisfied.

[0274] Diff31–Hys32>Thresh32(K3-2) Hys32 is a predefined or configured hysteresis parameter for event K3. Thresh32 is a predefined or configured threshold corresponding to event K3. Hys31 and Hys32 can be the same or different; for example, they can be similar. Thresh31 and thresh32 can be the same or different; for example, they can be similar.

[0275] This is determined by whether the BS configuration parameter "reportOnLeave" is enabled or disabled. If disabled, no report will be sent when the leave condition is met.

[0276] In addition, BS also configures a parameter called "trigger time", which specifies the range of values ​​for the time used to trigger the parameter and is related to the time when specific criteria for the event need to be met to trigger a report.

[0277] Another possible entry or exit condition for K3 is the presence of an additional offset.

[0278] The entry condition for event K3 is considered satisfied when the specified condition K3-3 is met.

[0279] Diff31+offset31+Hys31 <Thresh31(K3-3) The departure condition of event K3 is considered to be met when the conditions K3-4 specified below are satisfied.

[0280] Diff31+offset32–Hys32>Thresh32(K3-4) offset31 and offset32 can be the same or different. The values ​​of offset31 and offset32 can be positive or negative.

[0281] In some possible implementations, the value of Hys31 or Hys32 can be set to 0, or it may not be predefined or configured.

[0282] The method provided in this application defines event K3 and the corresponding entry and exit conditions for event K3. When any one of these conditions is met, active mode switching can be achieved.

[0283] In event K4, another model (Model 1) is superior to the current model (Model 2), and the UE should switch to Model 1.

[0284] The entry condition for event K4 is considered satisfied when the specified condition K4-1 is met.

[0285] Diff42+Hys41 <Diff41(K4-1) Diff41 is the difference between the first data point and the fifth anchor point, where the fifth anchor point can be pre-configured, such as the data anchor point associated with model 2. Diff42 is the difference between the first data point and the sixth anchor point, where the sixth anchor point can be pre-configured, such as the data anchor point associated with model 1. Hys41 is a predefined or configured parameter corresponding to event K4.

[0286] The exit condition of event K4 is considered to be met when the following specified condition K4-2 is satisfied.

[0287] Diff42–Hys42>Diff41(K4-2) Hys42 is a predefined or configured parameter corresponding to event K4. Hys41 and Hys42 can be the same or different; for example, they can be similar.

[0288] This is determined by whether the BS configuration parameter "reportOnLeave" is enabled or disabled. If disabled, no report will be sent when the leave condition is met.

[0289] In addition, BS also configures a parameter called "trigger time", which specifies the range of values ​​for the time used to trigger the parameter and is related to the time when specific criteria for the event need to be met to trigger a report.

[0290] Another possible entry or exit condition for K4 is the presence of an additional offset.

[0291] The entry condition for event K4 is considered satisfied when the specified condition K4-3 is met.

[0292] Diff42+offset42+Hys41 <Diff41+offset41(K4-3) The exit condition of event K4 is considered to be met when the following specified condition K4-4 is satisfied.

[0293] Diff42+offset42–Hys42>Diff41+offset41(K4-4) offset41 and offset42 can be the same or different. The values ​​of offset41 and offset42 can be positive or negative.

[0294] In some possible implementations, the value of Hys41 or Hys42 can be set to 0, or it may not be predefined or configured.

[0295] The method provided in this application defines event K4 and the corresponding entry and exit conditions for event K4. When any one of these conditions is met, the active switching of the model can be achieved.

[0296] The above embodiments are examples of UE actively performing model or mode switching. The method of BS actively performing model or mode switching is similar to the above embodiments. For specific implementation methods, please refer to the description of events K1 to K4 above, which will not be repeated in this application.

[0297] Figure 18 This is a schematic block diagram of a communication device 1700 according to an embodiment of this application. The communication device 1700 includes a sending module 1710, configured to send a report indicating AI model switching or mode switching when an event entry condition or exit condition is met.

[0298] In a possible implementation, the event is an operation of switching from the AI mode to the non-AI mode. The entry condition is determined by Diff11 and Thresh11, or the exit condition is determined by Diff11 and Thresh12. Diff11 is the first difference, Thresh11 is the first threshold, Thresh12 is the second threshold. The first difference is the difference between the first data and the first anchor point, and the first anchor point includes one or more reference data. The first threshold and the second threshold are predefined or configured thresholds corresponding to the event.

[0299] In a possible implementation, the entry condition is Diff11 – Hys11 > Thresh11, or the exit condition is Diff11 + Hys12 < Thresh12.

[0300] In a possible implementation, the entry condition is Diff11 + offset11 – Hys11 > Thresh11, or the exit condition is Diff11 + offset12 + Hys12 < Thresh12, where offset11 and offset12 are predefined or configured offsets.

[0301] In a possible implementation, Hys11 is the first hysteresis parameter, Hys12 is the second hysteresis parameter, and the first hysteresis parameter and the second hysteresis parameter are predefined or configured parameters corresponding to the event.

[0302] In a possible implementation, the event is an operation of switching from the first AI model to the second AI model. The entry condition is determined by Diff21, Diff22, Thresh21, and Thresh22, or the exit condition is determined by Diff21, Diff22, Thresh23, and Thresh24. Diff21 is the second difference, Diff22 is the third difference, Thresh21 is the third threshold, Thresh22 is the fourth threshold, Thresh23 is the fifth threshold, Thresh24 is the sixth threshold. The second difference is the difference between the first data and the second anchor point, and the second anchor point is the data anchor point associated with the first AI model. The third difference is the difference between the first data and the third anchor point, and the third anchor point is the data anchor point associated with the second AI model. The third threshold and the fifth threshold are predefined or configured thresholds corresponding to the first AI model, and the fourth threshold and the sixth threshold are predefined or configured thresholds corresponding to the second AI model.

[0303] In a possible implementation, the entry condition is Diff21 – Hys21 > Thresh21 and Diff22 + Hys22 < Thresh22, or the exit condition is Diff21 + Hys23 < Thresh23 or Diff22 – Hys24 > Thresh24.

[0304] In a possible implementation, the entry condition is Diff21 + offset21 – Hys21 > Thresh21 and Diff22 + offset22 + Hys22 < Thresh22, and the exit condition is Diff21 + offset23 + Hys23 < Thresh23 or Diff22 + offset24 – Hys24 > Thresh24, where offset21, offset22, offset23, and offset24 are predefined or configured offsets.

[0305] In a possible implementation, Hys21 is the third hysteresis parameter, Hys22 is the fourth hysteresis parameter, Hys23 is the fifth hysteresis parameter, and Hys24 is the sixth hysteresis parameter. The third hysteresis parameter, the fourth hysteresis parameter, the fifth hysteresis parameter, and the sixth hysteresis parameter are predefined or configured parameters corresponding to the event.

[0306] In a possible implementation, the event is an operation of switching from the non-AI mode to the AI mode. The entry condition is determined by Diff31 and Thresh31, or the exit condition is determined by Diff31 and Thresh32. Diff31 is the fourth difference, Thresh31 is the seventh threshold, Thresh32 is the eighth threshold. The fourth difference is the difference between the first data and the fourth anchor point, and the fourth anchor point includes one or more reference data. The seventh threshold and the eighth threshold are predefined or configured thresholds corresponding to the event.

[0307] In a possible implementation, the entry condition is Diff31 + Hys31 < Thresh31, or the exit condition is Diff31 – Hys32 > Thresh32.

[0308] In a possible implementation, the entry condition is Diff31 + offset31 + Hys31 < Thresh31, and the exit condition is Diff31 + offset32 – Hys32 > Thresh32, where offset31 and offset32 are predefined or configured offsets.

[0309] In a possible implementation, Hys31 is the seventh hysteresis parameter, Hys32 is the eighth hysteresis parameter, and the seventh and eighth hysteresis parameters are predefined or configured parameters corresponding to the event.

[0310] In a possible implementation, the event is an operation of switching from the third AI model to the fourth AI model. The entry condition is determined by Diff41 and Diff42, or the exit condition is determined by Diff41 and Diff42. Diff41 is the fifth difference, Diff42 is the sixth difference. The fifth difference is the difference between the first data and the fifth anchor point, and the fifth anchor point is the data anchor point associated with the third AI model. The sixth difference is the difference between the first data and the sixth anchor point, and the sixth anchor point is the data anchor point associated with the fourth AI model.

[0311] In a possible implementation, the entry condition is Diff42 + Hys41 < Diff41, and the exit condition is Diff42 – Hys42 > Diff41.

[0312] In a possible implementation, the entry condition is Diff42 + offset42 + Hys41 < Diff41 + offset41, and the exit condition is Diff42 + offset42 – Hys42 > Diff41 + offset41, where offset41 and offset42 are predefined or configured offsets.

[0313] In a possible implementation, Hys41 is the ninth hysteresis parameter, Hys42 is the tenth hysteresis parameter, and the ninth and tenth hysteresis parameters are predefined or configured parameters corresponding to the event.

[0314] In a possible implementation, the first data includes monitoring data or measurement data of a user device or a network device.

[0315] In a possible implementation, the first data includes monitoring data or measurement data related to an AI model of a user device or a network device.

[0316] In a possible implementation, the first data includes any one or more of the following: perception data, measurement data, channel data, neuron data of an AI model, and potential output data of an AI model.

[0317] In a possible implementation, the device further includes a processing module 1720 for communicating according to the report.

[0318] In a possible implementation, the device is located on a user device or a network device.

[0319] Such as Figure 19As shown, the communication device 2200 may include a processor 2210 and a transceiver 2220. Optionally, the communication device 2200 may also include a memory 2230. The memory 2230 may be used to store instruction information, or to store code and instructions to be executed by the processor 2210.

[0320] The memory 2230 may include random access memory, flash memory, read-only memory, programmable read-only memory, non-volatile memory, registers, etc. The processor 2210 may be a central processing unit (CPU).

[0321] For other functions and operations of the communication device 2200, please refer to... Figures 5 to 16 The process of the method embodiment shown will not be repeated here to avoid repetition.

[0322] This application also provides a computer storage medium that can store program instructions to execute the steps in the above method.

[0323] Alternatively, the storage medium may specifically be memory 2230.

[0324] This application also provides a computer program product. The computer program product includes computer program code. When the computer program code is run on a computer, the computer is able to perform the steps in the above-described method.

[0325] Optionally, all or part of the computer program code may be stored in the first storage medium. The first storage medium may be packaged together with the processor or packaged separately from the processor.

[0326] This application also provides a chip system, the system chip including an input / output interface, at least one processor, at least one memory, and a bus. The at least one memory is used to store instructions, and the at least one processor is used to invoke the instructions from the at least one memory to perform the operations in the methods described above.

[0327] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium. When the program runs, it executes the processes of the methods in the above embodiments. The storage medium may include: a magnetic disk, an optical disk, a read-only memory (ROM), or a random-access memory (RAM).

[0328] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the described apparatus embodiments are merely exemplary. For example, the unit division is only a logical functional division, and other division methods may be used in actual implementation. For example, multiple units or components may be merged or integrated into another system, or some features may be ignored or not performed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be implemented through some interfaces. Indirect coupling or communication connection between devices or units can be implemented electronically, mechanically, or otherwise.

[0329] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiment solution according to actual needs.

[0330] Furthermore, the functional units in the embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0331] The above are merely preferred exemplary embodiments of the present invention. Various modifications and variations can be made to the present invention by those skilled in the art without departing from the scope and purpose of the invention.

Claims

1. A communication method characterized by comprising: Comprising: Upon meeting an entry condition or an exit condition of an event, sending a report indicating an artificial intelligence (AI) model switching or a mode switching.

2. The method of claim 1, wherein, The event is an operation of switching from an AI mode to a non-AI mode, the entry condition is determined by Diff11 and Thresh11, or the exit condition is determined by Diff11 and Thresh12, Diff11 is a first difference value, Thresh11 is a first threshold value, Thresh12 is a second threshold value, the first difference value is a difference value between a first data and a first anchor point, the first anchor point comprises one or more reference data, the first threshold value and the second threshold value are predefined or configured threshold values corresponding to the event.

3. The method of claim 2, wherein, The entry condition is Diff11 – Hys11 > Thresh11, or the exit condition is Diff11 + Hys12 < Thresh12.

4. The method of claim 2, wherein, The entry condition is Diff11 + offset11 – Hys11 > Thresh11, or the exit condition is Diff11 + offset12 + Hys12 < Thresh12, offset11 and offset12 are predefined or configured offsets.

5. The method according to claim 3 or 4, characterized in that, Hys11 is a first hysteresis parameter, Hys12 is a second hysteresis parameter, the first hysteresis parameter and the second hysteresis parameter are predefined or configured parameters corresponding to the event.

6. The method of claim 1, wherein, The event is an operation of switching from a first AI model to a second AI model, the entry condition is determined by Diff21, Diff22, Thresh21 and Thresh22, or the exit condition is determined by Diff21, Diff22, Thresh23 and Thresh24, Diff21 is a second difference value, Diff22 is a third difference value, Thresh21 is a third threshold value, Thresh22 is a fourth threshold value, Thresh23 is a fifth threshold value, Thresh24 is a sixth threshold value, the second difference value is a difference value between a first data and a second anchor point, the second anchor point is a data anchor point associated with the first AI model, the third difference value is a difference value between the first data and a third anchor point, the third anchor point is a data anchor point associated with the second AI model, the third threshold value and the fifth threshold value are predefined or configured threshold values corresponding to the first AI model, the fourth threshold value and the sixth threshold value are predefined or configured threshold values corresponding to the second AI model.

7. The method of claim 6, wherein, The entry condition is Diff21 – Hys21 > Thresh21 and Diff22 + Hys22 < Thresh22, or the exit condition is Diff21 + Hys23 < Thresh23 or Diff22 – Hys24 > Thresh24.

8. The method of claim 6, wherein, The entering condition is Diff21+offset21-Hys21>Thresh21 and Diff22+offset22+Hys22<Thresh22, and the leaving condition is Diff21+offset23+Hys23<Thresh23 or Diff22+offset24-Hys24>Thresh24, offset21, offset22, offset23 and offset24 are predefined or configured offsets.

9. The method according to claim 7 or 8, characterized in that, Hys21 is a third hysteresis parameter, Hys22 is a fourth hysteresis parameter, Hys23 is a fifth hysteresis parameter, and Hys24 is a sixth hysteresis parameter, the third hysteresis parameter, the fourth hysteresis parameter, the fifth hysteresis parameter and the sixth hysteresis parameter are predefined or configured parameters corresponding to the event.

10. The method of claim 1, wherein, The event is an operation of switching from a non-AI mode to an AI mode, the entering condition is determined by Diff31 and Thresh31, or the leaving condition is determined by Diff31 and Thresh32, Diff31 is a fourth difference value, Thresh31 is a seventh threshold value, and Thresh32 is an eighth threshold value, the fourth difference value is a difference value between the first data and a fourth anchor point, the fourth anchor point includes one or more reference data, and the seventh threshold value and the eighth threshold value are predefined or configured threshold values corresponding to the event.

11. The method of claim 10, wherein, The entering condition is Diff31+Hys31<Thresh31, or the leaving condition is Diff31-Hys32>Thresh32.

12. The method of claim 10, wherein, The entering condition is Diff31+offset31+Hys31<Thresh31, and the leaving condition is Diff31+offset32-Hys32>Thresh32, offset31 and offset32 are predefined or configured offsets.

13. The method according to claim 11 or 12, characterized in that, Hys31 is a seventh hysteresis parameter, and Hys32 is an eighth hysteresis parameter, the seventh hysteresis parameter and the eighth hysteresis parameter are predefined or configured parameters corresponding to the event.

14. The method of claim 1, wherein, The event is an operation of switching from a third AI model to a fourth AI model, the entering condition is determined by Diff41 and Diff42, or the leaving condition is determined by Diff41 and Diff42, Diff41 is a fifth difference value, Diff42 is a sixth difference value, the fifth difference value is a difference value between the first data and a fifth anchor point, the fifth anchor point is a data anchor point associated with the third AI model, and the sixth difference value is a difference value between the first data and a sixth anchor point, the sixth anchor point is a data anchor point associated with the fourth AI model.

15. The method of claim 14, wherein, The entering condition is Diff42+Hys41<Diff41, and the leaving condition is Diff42-Hys42>Diff41.

16. The method of claim 14, wherein, The entering condition is Diff42+offset42+Hys41<Diff41+offset41, and the leaving condition is Diff42+offset42-Hys42>Diff41+offset41, offset41 and offset42 are predefined or configured offsets.

17. The method according to claim 15 or 16, characterized in that Hys41 is a ninth hysteresis parameter, and Hys42 is a tenth hysteresis parameter, the ninth hysteresis parameter and the tenth hysteresis parameter are predefined or configured parameters corresponding to the event.

18. The method according to any one of claims 2 to 17, characterized in that, The first data includes monitoring data or measurement data of a user equipment or a network equipment.

19. The method of claim 18, wherein, The first data includes monitoring data or measurement data related to an AI model of the user equipment or the network equipment.

20. The method of any one of claims 2-19, wherein, The first data includes any one or more of the following: perception data, measurement data, channel data, neuron data of an AI model, and potential output data of the AI model.

21. The method of any one of claims 1 to 20, wherein, Further comprising: According to the report, communication is performed.

22. The method of any one of claims 1 to 21, wherein, The method is performed by a user equipment or a network equipment.

23. A communications device, characterized by Comprising: A sending module configured to send a report indicating an AI model switching or mode switching when an entering condition or a leaving condition of an event is met.

24. The communication apparatus according to claim 23, wherein, The event is an operation of switching from an AI mode to a non-AI mode, the entering condition is determined by Diff11 and Thresh11, or the leaving condition is determined by Diff11 and Thresh12, Diff11 is a first difference value, Thresh11 is a first threshold value, Thresh12 is a second threshold value, the first difference value is a difference value between first data and a first anchor point, the first anchor point includes one or more reference data, and the first threshold value and the second threshold value are predefined or configured threshold values corresponding to the event.

25. The communication apparatus according to claim 24, wherein, The entering condition is Diff11-Hys11>Thresh11, or the leaving condition is Diff11+Hys12<Thresh12.

26. The communication apparatus according to claim 24, wherein The entering condition is Diff11+offset11-Hys11>Thresh11, or the leaving condition is Diff11+offset12+Hys12<Thresh12, offset11 and offset12 are predefined or configured offsets.

27. The communication apparatus according to claim 25 or 26, wherein, Hys11 is a first hysteresis parameter, and Hys12 is a second hysteresis parameter, the first hysteresis parameter and the second hysteresis parameter are predefined or configured parameters corresponding to the event.

28. The communication apparatus of claim 23, wherein The event is an operation of switching from a first AI model to a second AI model, the entry condition is determined by Diff21, Diff22, Thresh21 and Thresh22, or the exit condition is determined by Diff21, Diff22, Thresh23 and Thresh24, Diff21 is a second difference value, Diff22 is a third difference value, Thresh21 is a third threshold value, Thresh22 is a fourth threshold value, Thresh23 is a fifth threshold value, and Thresh24 is a sixth threshold value, the second difference value is a difference value between the first data and a second anchor point, the second anchor point is a data anchor point associated with the first AI model, the third difference value is a difference value between the first data and a third anchor point, the third anchor point is a data anchor point associated with the second AI model, the third threshold value and the fifth threshold value are predefined or configured threshold values corresponding to the first AI model, and the fourth threshold value and the sixth threshold value are predefined or configured threshold values corresponding to the second AI model.

29. The communication apparatus according to claim 28, wherein, The entry condition is Diff21-Hys21>Thresh21 and Diff22+Hys22<Thresh22, or the exit condition is Diff21+Hys23<Thresh23 or Diff22-Hys24>Thresh24.

30. The communication apparatus of claim 28, wherein, The entry condition is Diff21+offset21-Hys21>Thresh21 and Diff22+offset22+Hys22<Thresh22, and the exit condition is Diff21+offset23+Hys23<Thresh23 or Diff22+offset24-Hys24>Thresh24, offset21, offset22, offset23 and offset24 are predefined or configured offsets.

31. The communication apparatus according to claim 29 or 30, wherein, Hys21 is a third hysteresis parameter, Hys22 is a fourth hysteresis parameter, Hys23 is a fifth hysteresis parameter, and Hys24 is a sixth hysteresis parameter, the third hysteresis parameter, the fourth hysteresis parameter, the fifth hysteresis parameter and the sixth hysteresis parameter are predefined or configured parameters corresponding to the event.

32. The communication apparatus of claim 23, wherein The event is an operation of switching from a non-AI mode to an AI mode, the entry condition is determined by Diff31 and Thresh31, or the exit condition is determined by Diff31 and Thresh32, Diff31 is a fourth difference value, Thresh31 is a seventh threshold value, and Thresh32 is an eighth threshold value, the fourth difference value is a difference value between the first data and a fourth anchor point, the fourth anchor point includes one or more reference data, and the seventh threshold value and the eighth threshold value are predefined or configured threshold values corresponding to the event.

33. The communication apparatus according to claim 32, wherein The entering condition is Diff31+Hys31<Thresh31, or the leaving condition is Diff31-Hys32>Thresh32.

34. The communication apparatus of claim 32, wherein The entering condition is Diff31+offset31+Hys31<Thresh31, and the leaving condition is Diff31+offset32-Hys32>Thresh32, offset31 and offset32 are predefined or configured offsets.

35. The communication apparatus according to claim 33 or 34, wherein, Hys31 is a seventh hysteresis parameter, and Hys32 is an eighth hysteresis parameter, the seventh hysteresis parameter and the eighth hysteresis parameter are predefined or configured parameters corresponding to the event.

36. The communication apparatus of claim 23, wherein The event is an operation of switching from a third AI model to a fourth AI model, the entering condition is determined by Diff41 and Diff42, or the leaving condition is determined by Diff41 and Diff42, Diff41 is a fifth difference value, Diff42 is a sixth difference value, the fifth difference value is a difference value between the first data and a fifth anchor point, the fifth anchor point is a data anchor point associated with the third AI model, and the sixth difference value is a difference value between the first data and a sixth anchor point, the sixth anchor point is a data anchor point associated with the fourth AI model.

37. The communication apparatus of claim 36, wherein The entering condition is Diff42+Hys41<Diff41, and the leaving condition is Diff42-Hys42>Diff41.

38. The communication apparatus of claim 36, wherein The entering condition is Diff42+offset42+Hys41<Diff41+offset41, and the leaving condition is Diff42+offset42-Hys42>Diff41+offset41, offset41 and offset42 are predefined or configured offsets.

39. The communication apparatus according to claim 37 or 38, wherein, Hys41 is a ninth hysteresis parameter, and Hys42 is a tenth hysteresis parameter, the ninth hysteresis parameter and the tenth hysteresis parameter are predefined or configured parameters corresponding to the event.

40. The communication apparatus according to any one of claims 24-39, wherein, The first data includes monitoring data or measurement data of a user equipment or a network equipment.

41. The communication apparatus of claim 40, wherein The first data includes monitoring data or measurement data related to an AI model of the user equipment or the network equipment.

42. The communication apparatus according to any one of claims 24-41, wherein, The first data includes any one or more of the following: perception data, measurement data, channel data, neuron data of an AI model, and potential output data of the AI model.

43. The communication apparatus according to any one of claims 23-42, wherein, The apparatus further includes a processing module configured to communicate according to the report.

44. The communication apparatus according to any one of claims 23-43, wherein, The apparatus is located on a user equipment or a network equipment.

45. A communications device, characterized by A processor and a memory, the processor is connected with the memory; the memory is configured to store instructions, and the processor is configured to execute the instructions; when the processor executes the instructions stored in the memory, the processor can execute the method according to any one of claims 1-22.

46. A computer-readable storage medium, comprising: The computer readable storage medium stores instructions, when the instructions are run on the processor, the processor can execute the method according to any one of claims 1-22.

47. A computer program product, characterised in that, comprising computer program code which, when the computer program code is run on a computer, the computer is capable of performing the method according to any one of claims 1 to 22.

Images