WO2025245664A1 - Model management methods and apparatuses, terminal device and network device - Google Patents

Abstract

Provided in the embodiments of the present application are model management methods and apparatuses, a terminal device and a network device. A method comprises: when a first event has occurred, a terminal device sends first information to a network device, the first event being related to a performance monitoring result of a model associated with the terminal device, the first information being used for one or more of the following: requesting an uplink transmission resource, indicating the uplink transmission resource, and requesting a downlink reference signal, the uplink transmission resource being used for sending model information, and the downlink reference signal being used for model monitoring or online training.

Description

Model management methods and devices, terminal equipment, network equipment Technical Field

This application relates to the field of mobile communication technology, specifically to a model management method and apparatus, terminal equipment, and network equipment.

Background Technology

For wireless communication based on Artificial Intelligence (AI), terminal devices can use AI models deployed on the terminal device side for various processes such as Channel State Information (CSI) feedback, channel coding, channel decoding, modulation, demodulation, channel estimation, and detection.

As the communication environment of terminal devices changes, these devices need to report the performance status of AI models in a timely manner to determine whether the AI models used are still suitable for the current scenario. If the performance of the AI model deteriorates, it is necessary to update the model and replace the original AI model with a new one to ensure transmission performance in the current scenario.

In practical applications, terminal devices need to monitor the performance of AI models based on periodic reference signals. In addition, terminal devices also need to periodically report the performance monitoring results of AI models. This requires network devices to pre-configure reporting resources and reference signals, resulting in resource waste.

Summary of the Invention

This application provides a model management transmission method and apparatus, a terminal device, and a network device.

In a first aspect, the model management method provided in the embodiments of this application includes:

In the event of a first event, the terminal device sends first information to the network device; the first event is related to the performance monitoring results of the model associated with the terminal device.

The first information is used for one or more of the following:

Requesting uplink transmission resources;

Indicates uplink transmission resources; the uplink transmission resources are used to send model information;

Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training.

Secondly, the model management method provided in the embodiments of this application includes:

The network device receives first information sent by the terminal device; the first information is used for one or more of the following:

Requesting uplink transmission resources;

Indicates uplink transmission resources; the uplink transmission resources are used to send model information;

Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training.

Thirdly, the model management device provided in this application embodiment is applied to a terminal device, and the device includes:

The first sending unit is configured to send first information to the network device in the event of a first event; the first event is related to the performance monitoring results of the model associated with the terminal device.

The first information is used for one or more of the following:

Requesting uplink transmission resources;

Indicates uplink transmission resources; the uplink transmission resources are used to send model information;

Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training.

Fourthly, the model management device provided in this application embodiment is applied to a network device, and the device includes:

The second receiving unit is configured to receive first information sent by the terminal device; the first information is used for one or more of the following:

Requesting uplink transmission resources;

Indicates uplink transmission resources; the uplink transmission resources are used to send model information;

Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training.

Fifthly, the terminal device provided in the embodiments of this application includes a processor and a memory. The memory is used to store computer programs, and the processor is used to call and run the computer programs stored in the memory to execute the model management method described above.

Sixthly, the network device provided in the embodiments of this application includes a processor and a memory. The memory is used to store computer programs, and the processor is used to call and run the computer programs stored in the memory to execute the model management method described above.

The chip provided in this application embodiment is used to implement the above-described model management method.

Specifically, the chip includes a processor for calling and running a computer program from memory, causing a device equipped with the chip to perform the aforementioned model management method.

The computer-readable storage medium provided in this application embodiment is used to store a computer program that causes a computer to execute the model management method described above.

The computer program product provided in this application includes computer program instructions that cause a computer to execute the model management method described above.

The computer program provided in this application embodiment, when run on a computer, causes the computer to execute the above-described model management method.

In the model management method provided in this application embodiment, a terminal device can send first information to a network device when a first event occurs; the first event is related to the performance monitoring results of the model associated with the terminal device; the first information is used for one or more of the following: requesting uplink transmission resources, indicating uplink transmission resources, and requesting downlink reference signals; wherein, the uplink transmission resources are used to send model information, and the downlink reference signals are used for model monitoring or online training. It is understood that the terminal device can report and monitor model information based on a first event-driven approach. This approach does not require the network device to pre-allocate reporting resources and reference signals, effectively reducing resource overhead caused by model monitoring and training, and avoiding unnecessary reporting of model information, thus improving resource utilization efficiency.

Attached Figure Description

The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

Figure 1 is a schematic diagram of a communication architecture provided in an embodiment of this application;

Figure 2 is a schematic diagram of a neuron structure provided in an embodiment of this application;

Figure 3 is a schematic diagram of a neural network structure provided in an embodiment of this application;

Figure 4 is a schematic flowchart of a model management method provided in an embodiment of this application;

Figure 5 is a schematic diagram of a model management method provided in an embodiment of this application;

Figure 6 is a schematic diagram of a model management method provided in an embodiment of this application;

Figure 7 is a schematic diagram of a model management method provided in an embodiment of this application;

Figure 8 is a schematic flowchart of a model management method provided in an embodiment of this application;

Figure 9 is a schematic flowchart of a model management method provided in an embodiment of this application;

Figure 10 is a schematic diagram of a model management method provided in an embodiment of this application;

Figure 11 is a schematic diagram of the structure of a model management device 1100 provided in an embodiment of this application;

Figure 12 is a schematic diagram of the structure of a model management device 1200 provided in an embodiment of this application;

Figure 13 is a schematic structural diagram of a communication device provided in an embodiment of this application;

Figure 14 is a schematic structural diagram of a chip according to an embodiment of this application;

Figure 15 is a schematic block diagram of a communication system provided in an embodiment of this application.

Detailed Implementation

The technical solutions of the embodiments of 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. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

Figure 1 is a schematic diagram of an application scenario of an embodiment of this application.

As shown in Figure 1, the communication system 100 may include a terminal device 110 and a network device 120. The network device 120 can communicate with the terminal device 110 via an air interface. Multi-service transmission is supported between the terminal device 110 and the network device 120.

It should be understood that the embodiments of this application are only illustrated by way of example with communication system 100, but the embodiments of this application are not limited thereto. That is to say, the technical solutions of the embodiments of this application can be applied to various communication systems, such as: Long Term Evolution (LTE) system, LTE Time Division Duplex (TDD), Universal Mobile Telecommunication System (UMTS), Internet of Things (IoT) system, Narrow Band Internet of Things (NBIoT) system. Internet of Things (NB-IoT) systems, enhanced machine-type communications (eMTC) systems, 5G communication systems (also known as New Radio (NR) communication systems), or future communication systems, etc.

In the communication system 100 shown in Figure 1, network device 120 may be an access network device that communicates with terminal device 110. The access network device can provide communication coverage for a specific geographical area and can communicate with terminal device 110 (e.g., UE) located within that coverage area.

Network device 120 may be an evolved Node B (eNB or eNodeB) in a Long Term Evolution (LTE) system, a Next Generation Radio Access Network (NG RAN) device, a base station (gNB) in an NR system, a radio controller in a Cloud Radio Access Network (CRAN), or a relay station, access point, vehicle-mounted device, wearable device, hub, switch, bridge, router, or network device in a future evolved Public Land Mobile Network (PLMN), etc.

Terminal device 110 can be any terminal device, including but not limited to terminal devices that are connected to network device 120 or other terminal devices via wired or wireless connections.

For example, the terminal device 110 may refer to an access terminal, user equipment (UE), user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent, or user device. The access terminal may be a cellular phone, cordless phone, Session Initiation Protocol (SIP) phone, IoT device, satellite handheld terminal, Wireless Local Loop (WLL) station, Personal Digital Assistant (PDA), handheld device with wireless communication capabilities, computing device or other processing device connected to a wireless modem, vehicle-mounted device, wearable device, terminal device in a 5G network, or terminal device in a future evolved network, etc.

Terminal device 110 can be used for device-to-device (D2D) communication.

The wireless communication system 100 may further include a core network device 130 that communicates with the network device 120. This core network device 130 may be a 5G core network (5G Core, 5GC) device, such as an Access and Mobility Management Function (AMF), an Authentication Server Function (AUSF), a User Plane Function (UPF), or a Session Management Function (SMF). Optionally, the core network device 130 may also be an Evolved Packet Core (EPC) device for an LTE network, such as a Session Management Function + Core Packet Gateway (SMF+PGW-C) device. It should be understood that SMF+PGW-C can simultaneously implement the functions of both SMF and PGW-C. During network evolution, the aforementioned core network equipment may also be called by other names, or new network entities may be formed by dividing the functions of the core network. This application does not impose any restrictions on this.

The various functional units in the communication system 100 can also communicate with each other through a next-generation (NG) interface.

For example, terminal devices establish air interface connections with access network devices through the NR interface for transmitting user plane data and control plane signaling; terminal devices can establish control plane signaling connections with the AMF through NG interface 1 (N1); access network devices, such as next-generation radio access base stations (gNB), can establish user plane data connections with the UPF through NG interface 3 (N3); access network devices can establish control plane signaling connections with the AMF through NG interface 2 (N2); the UPF can establish control plane signaling connections with the SMF through NG interface 4 (N4); the UPF can interact with the data network for user plane data through NG interface 6 (N6); the AMF can establish control plane signaling connections with the SMF through NG interface 11 (N11); and the SMF can establish control plane signaling connections with the PCF through NG interface 7 (N7).

Figure 1 exemplarily illustrates a network device, a core network device, and two terminal devices. Optionally, the wireless communication system 100 may include multiple network devices, and each network device may include other numbers of terminal devices within its coverage area. This application embodiment does not limit this.

It should be noted that Figure 1 is merely an example illustrating the system to which this application applies. Of course, the method shown in the embodiments of this application can also be applied to other systems. Furthermore, the terms "system" and "network" are often used interchangeably in this document. The term "and/or" in this document merely describes the relationship between related objects, indicating that three relationships can exist. For example, A and/or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character "/" in this document generally indicates that the preceding and following related objects have an "or" relationship. It should also be understood that "instruction" mentioned in the embodiments of this application can be a direct instruction, an indirect instruction, or an indication of a related relationship. For example, A instructing B can mean that A directly instructs B, for example, B can be obtained through A; it can also mean that A indirectly instructs B, for example, A instructs C, B can be obtained through C; or it can mean that there is a related relationship between A and B. It should also be understood that "correspondence" mentioned in the embodiments of this application can indicate a direct or indirect correspondence between two things, or an related relationship between two things, or a relationship of instruction and being instructed, configuration and being configured, etc. It should also be understood that the terms "predefined" or "predefined rules" mentioned in the embodiments of this application refer to... This can be achieved by pre-storing the corresponding codes, tables, or other means of indicating relevant information in the device (e.g., including terminal devices and network devices), and this application does not limit the specific implementation method. For example, pre-definition can refer to what is defined in the protocol. It should also be understood that in the embodiments of this application, the "protocol" can refer to standard protocols in the field of communication, such as the LTE protocol, the NR protocol, and related protocols applied to future communication systems, and this application does not limit this.

To facilitate understanding of the technical solutions of the embodiments of this application, the relevant technologies of the embodiments of this application are described below. The following relevant technologies are optional solutions and can be combined with the technical solutions of the embodiments of this application in any way, and they all fall within the protection scope of the embodiments of this application.

AI models are models capable of handling a variety of tasks. They possess the ability to learn and adapt, dynamically adjusting and making decisions based on changes in the environment. AI models can also be called machine learning (ML) models; the two are equivalent or interchangeable.

In practical applications, AI models can be constructed from neural networks. A neural network is a computational model consisting of multiple interconnected neurons. The connections between nodes represent weighted values from the input signal to the output signal, called weights. Each node performs a weighted summation of different input signals and outputs the result through a specific activation function. Referring to the schematic diagram of the neuron structure shown in Figure 2, a1, a2, ..., an and 1 are the inputs of the neuron, w1, w2, ..., wn and b represent the weights, Sum represents the summation function, f represents the activation function, and t represents the output result.

A simple neural network, as shown in Figure 3, consists of an input layer, hidden layers, and an output layer. Through different connections, weights, and activation functions of multiple neurons, different outputs can be generated, thus fitting a mapping relationship from input to output. Each node in the previous level is connected to all its nodes in the next level. This fully connected model can also be called a DNN, or deep neural network.

An AI model can be trained and obtained through processes such as dataset construction, training, validation, and testing. Training can be divided into offline training and online training. Offline training can yield a static training result. During the use of the AI model by network devices or terminal devices, as the terminal devices further measure and/or report, the network devices can continue to collect more data for real-time online training to optimize the AI model's parameters and achieve better inference and prediction results. After obtaining the AI model, by inputting the currently obtained information into the AI model, the corresponding model output can be obtained through inference.

When AI models are used in wireless communication, they can be divided into single-ended models and dual-ended models. Single-ended models can be deployed on only one side of the terminal device or network device, and the training of the AI model can also be carried out on only one side. Dual-ended models need to be deployed in pairs on the terminal device and network device sides, and the models on both sides need to be trained together. That is to say, the models deployed on both sides are corresponding and cannot be used or updated independently.

To achieve different communication functions, different AI models are introduced into the communication system, defining the corresponding inputs and outputs for each communication function.

For example, when using an AI model for CSI feedback, the obtained channel information (such as feature vectors, beam information, and delay information) can be used as input to the AI model to infer the corresponding CSI quantization bits. On the network side, there will be a corresponding AI model that uses the CSI quantization bits as input to infer the corresponding channel information.

For example, when using the AI model for beam management, the terminal device can use the reference signal received power (RSRP) corresponding to multiple beams in the measured second beam set (represented by the CSI-RS resource index) as input to the AI model, thereby inferring the best beam (represented by the CSI-RS resource index) and its corresponding RSRP in the first beam set, and then report the inference result to the network device. The second beam set can be a subset of the first beam set.

In addition, AI models can also be used for other communication processes such as localization, channel coding, channel decoding, modulation and demodulation, and channel estimation.

It should be understood that the communication environment surrounding terminal devices changes in real time. As the communication environment of terminal devices changes, the terminal devices need to report the performance status of the model in a timely manner to determine whether the AI model used is still suitable for the current scenario. If the performance of the AI model deteriorates, the AI model needs to be updated, and a new AI model needs to be used to replace the original AI model to ensure transmission performance in the current scenario.

In practical applications, terminal devices need to monitor the performance of AI models based on periodic reference signals. In addition, terminal devices also need to periodically report the performance monitoring results of AI models. This requires network devices to pre-configure reporting resources and reference signals, which increases resource overhead and causes resource waste.

Based on this, embodiments of this application provide a model management method, wherein a terminal device can send first information to a network device upon the occurrence of a first event; the first event is related to the performance monitoring results of a model associated with the terminal device; the first information is used for one or more of the following: requesting uplink transmission resources, indicating uplink transmission resources, and requesting downlink reference signals; wherein the uplink transmission resources are used to send model information, and the downlink reference signals are used for model monitoring or online training. It is understood that the terminal device can report and monitor model information based on a first event-driven approach. This approach does not require the network device to pre-allocate reporting resources and reference signals, effectively reducing resource overhead caused by model monitoring and training, and avoiding unnecessary reporting of model information, thus improving resource utilization efficiency.

To facilitate understanding of the technical solutions of the embodiments of this application, the technical solutions of this application are described in detail below through specific embodiments. The above-mentioned related technologies are optional solutions and can be arbitrarily combined with the technical solutions of the embodiments of this application, all of which fall within the protection scope of the embodiments of this application. The embodiments of this application include at least some of the following contents.

Figure 4 illustrates a model management method provided by an embodiment of this application, which may include the following steps.

S410. Upon the occurrence of a first event, the terminal device sends first information to the network device; the first event is related to the performance monitoring results of the model associated with the terminal device. Accordingly, the network device receives the first information sent by the terminal device.

The first piece of information is used for one or more of the following:

Requesting uplink transmission resources;

Indicates uplink transmission resources; the uplink transmission resources are used to send model information;

Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training.

In this embodiment of the application, the terminal device can use the model to communicate with the network device.

It should be noted that the model mentioned in the embodiments of this application can also be called an AI model, and the two are equivalent or interchangeable.

It should be noted that the models mentioned in the embodiments of this application can be used in communication processes such as CSI feedback, beam management, positioning, channel coding, channel decoding, modulation and demodulation, and channel estimation. The embodiments of this application do not limit this.

In some embodiments, the above model may be a single-end model, specifically a model deployed separately on the terminal device side.

In other embodiments, the above model may also be a model deployed on the terminal device side in a dual-end model, and correspondingly, another model corresponding to this model may be deployed on the network device side.

In this embodiment, the first event is related to the performance monitoring results of the model associated with the terminal device. It is understood that the terminal device can perform performance monitoring on its associated model, obtain the performance monitoring results, and then determine whether the first event has occurred based on the performance monitoring results.

It should be noted that the model associated with the terminal device can also be called the model supported by the terminal device. The model associated with the terminal device can be the model currently used by the terminal device, as well as the model not used by the terminal device.

It should also be noted that the performance monitoring results here can be the performance metrics corresponding to the AI models supported by the terminal device (which can be currently in use or not).

In this embodiment, the terminal device sends first information to the network device only when the first event occurs, requesting uplink transmission resources, and/or instructing on uplink transmission resources, and/or requesting downlink reference signals. Here, the uplink transmission resources are used by the terminal device to report model information to the network device, and the downlink reference signals are used for further model monitoring or online training. That is, the terminal device can perform subsequent model information reporting and more detailed model monitoring and training only after the first event occurs. This way, the network device does not need to pre-allocate reporting resources and reference signals, effectively reducing resource overhead and avoiding unnecessary model reporting and monitoring.

It should be noted that the timing of when the terminal device performs performance monitoring on the associated model can be implemented in different ways.

In some embodiments, the terminal device may periodically monitor the performance of its associated models.

Understandably, terminal devices can periodically monitor the performance of associated models, thus allowing them to periodically determine whether the first event has occurred.

It should be noted that the period for the terminal device to perform performance monitoring on the associated model can be predefined by the protocol, configured by the network device, or agreed upon by the terminal device and the network device. This application embodiment does not impose any restrictions on this.

In other embodiments, the terminal device can perform performance monitoring on the associated model if a first condition is met:

The changes in large-scale channel parameters measured by the terminal device exceed the first range;

The metric obtained by the terminal device exceeds the second range;

The serving cell of the terminal device has changed.

In some embodiments, large-scale channel parameters may include channel delay power spectrum, presence of line of sight (LOS) path, Doppler shift, etc., and the embodiments of this application do not limit these parameters.

Among them, the large-scale reference change of the channel exceeding the first range can refer to the change of the channel delay power spectrum of the terminal device exceeding a certain range, the existence of a change between LOS and non-line-of-sight (non-LOS, NLOS), the change of Doppler frequency shift exceeding a certain range, etc. The embodiments of this application do not limit this.

In some embodiments, the measurement quantities of the terminal device may include Reference Signal Receiving Power (RSRP), Signal to Interference plus Noise Ratio (SINR), Generalized Cosine Similarity (GCS), Square of Generalized Cosine Similarity (SGCS), Normalized Mean Squared Error (NMSE), etc., and the embodiments of this application do not limit these.

Among them, the measured quantity exceeding the second range can be RSRP, SINR, GCS, SGCS, or NMSE being higher or lower than a certain threshold value.

It should be noted that the aforementioned first and/or second scopes may be specified by the protocol or configured by the network device, and the embodiments of this application do not impose any restrictions on them.

In addition, a change in the serving cell of a terminal device could mean that the terminal device has undergone cell handover.

Understandably, changes in large-scale channel parameters, measurements, and the terminal device's cell can reflect changes in the terminal device's communication environment to some extent. In this embodiment, the terminal device only initiates performance monitoring of the associated model when its communication environment undergoes significant changes. This reduces the frequency of performance monitoring by the terminal device, thereby saving power consumption.

In this embodiment of the application, the terminal device can obtain the performance indicators of the associated model by monitoring the performance of the associated model.

It should be noted that the model associated with the terminal device can be a model supported by the terminal device. In other words, the terminal device can use these models to perform the process of communicating with network devices.

In some embodiments, the model associated with the terminal device can be referred to as the candidate model of the terminal device. The model used by the terminal device when performing communication can be any one of the candidate models.

It should be noted that the model associated with the terminal device can be a model configured by the network device for the terminal device, or a model reported by the terminal device to the network device, or a model pre-agreed upon by the terminal device and the network device.

In some embodiments, the model associated with the terminal device may include a first model and one or more second models. The first model may be a model currently used by the terminal device, and the one or more second models may be models supported by the terminal device, and the one or more second models are different from the first model.

Understandably, the terminal device can monitor the performance of the first model currently in use, as well as the second model that is not currently in use, to obtain the performance indicators of each model.

It should be noted that performance metrics are used to evaluate the performance of a model. They can include physical quantities such as SINR, Block Error Rate (BLER), Throughput, SGCS, GCS, NMSE, etc., or the difference between the above physical quantities and the reference values.

In one possible implementation, the model's performance metric could be a comparison between the model's output and a second reference label (ground truth label).

For example, for a model used to estimate channel information, the reference label of the model can be the channel information measured by the terminal device. The terminal device can use the values of GCS, SGCS, or NMSE between the measured channel information and the channel information output by the model as the performance index of the model.

For example, for a model used to calculate the precoding matrix, its reference label can be the expected channel capacity or expected throughput (e.g., a capacity ceiling calculated theoretically). The terminal device can calculate the corresponding channel capacity or throughput based on the precoding matrix output by the model. The terminal device can use the difference between the calculated channel capacity and the expected channel capacity, or the difference between the calculated throughput and the expected throughput, as a performance metric for the model.

In another possible implementation, the model's performance metric can be a comparison between the model's output and the output obtained by a non-AI method.

For example, the terminal device can calculate the corresponding SINR based on the model's output, and use the difference between it and the SINR calculated based on traditional methods as the performance index of the model.

In another possible implementation, the model's performance metrics can be calculated based on the model's output.

For example, the terminal device can estimate the corresponding SINR, NMSE, BLER, or throughput based on the model's output to obtain the model's performance metrics.

It should be noted that the above performance indicators may be predefined by the protocol or indicated by the network device, and this application embodiment does not impose any restrictions on them.

In one embodiment of this application, after the terminal device obtains the performance indicators of the first model, or the first model and the second model, it can determine whether the first event has occurred based on the respective performance indicators of the models.

In some embodiments, the first event may include one or more of the following:

The performance metrics of the first model are worse than the first threshold value;

The performance metrics of the first model and one or more second models are all worse than the second threshold value;

In one or more second models, at least one second model has a better performance metric than the first model.

In one or more second models, at least one second model has a better performance metric than the first model, and the difference between the performance metric of at least one second model and the performance metric of the first model is greater than a third threshold value.

At least one of the second models outperforms the fourth threshold in terms of performance metrics;

The performance metric of the first model is worse than the fifth threshold, and at least one of the first two models has a performance metric better than the sixth threshold.

The absolute value of the difference between the performance metric of the first model and the performance metric of at least one of the one or more second models is less than the seventh threshold value;

At least one of the second models outperforms the performance of the reference model.

In one or more second models, the performance metric of at least one second model is better than that of the reference model, and the difference between the performance metric of at least one second model and the performance metric of the reference model is greater than the eighth threshold.

In one or more second models, the performance metrics of two or more second models are better than those of the first model;

Two or more of the second models outperform the performance metrics of the first model, and the difference between the performance metrics of the two or more second models and the performance metrics of the first model is greater than the ninth threshold.

At least one of the two second models has a performance metric that is no worse than that of the first model, and the complexity of at least one second model is lower than that of the first model.

In some embodiments, the first to ninth threshold values can be numbers greater than 0.

In some embodiments, one or more of the first to ninth thresholds described above are determined based on any of the following methods:

Protocol predefined;

The first configuration information sent by the network device.

Understandably, the above threshold values can be predefined by the protocol or configured by the network device.

It should be noted that the aforementioned reference model can be a model agreed upon by the terminal device and the network device, and is based on this reference model. The terminal device and the network device can perform basic communication; that is, the reference model is one that can guarantee normal communication between the terminal device and the network device.

In some embodiments, the reference model may also be called the baseline model, and the performance index of the reference model can be used as a benchmark value for judging the performance index of other models.

In some embodiments, model complexity can be reflected by the number of model parameters and/or the computational load of model inference. A larger number of model parameters and a higher computational load during inference indicate higher model complexity. Conversely, a smaller number of model parameters and a lower computational load during inference indicate lower model complexity.

In some embodiments, the first event may be determined based on any of the following methods:

Protocol predefined;

The second configuration information sent by the network device.

In other words, the first event can be either predefined by the protocol or configured by the network device.

It should be noted that the first event can be a combination of the above events. In this case, the terminal device will only be driven to send the first information when all the events in the combination are satisfied.

In some embodiments, the terminal device may determine whether the first event has occurred based on the result of a single performance monitoring session. For example, the terminal device may determine that the first event has occurred as long as the condition for the first event is met even once.

In some embodiments, the terminal device determines whether a first event has occurred based on multiple performance monitoring results. For example, the terminal device determines that the first event has occurred only if the conditions for the first event are met multiple times (e.g., M times) within a certain time window. The length of the time window and the number of times the conditions need to be met, M, can be configured by the network device or agreed upon by the terminal device and the network device. Alternatively, the terminal device determines that the first event has occurred only if the conditions for the first event are met consecutively multiple times. This method does not rely on a single performance monitoring result, thereby effectively avoiding unnecessary information reporting caused by occasional performance degradation.

Therefore, it can be seen that the terminal device can trigger the sending of the first information under different circumstances, such as the performance index of the first model being poor while the performance index of the second model is better, or the complexity of the first model being higher than that of the second model with the same performance index.

In some embodiments, the first information is carried by Media Access Control (MAC) layer signaling, or by Physical Uplink Control Channel (PUCCH).

In the embodiments of this application, the first information has multiple functions. Specifically, the first information can be used to request uplink transmission resources, to indicate uplink transmission resources, to request downlink reference signals, to indicate and request uplink transmission resources, to request both uplink transmission resources and downlink reference signals, to indicate both uplink transmission resources and downlink reference signals, and to indicate and request both uplink transmission resources and downlink reference signals.

It should be noted that uplink transmission resources can also be called uplink channel resources, uplink transmission resources, uplink time-frequency resources, etc., and the above concepts are equivalent or interchangeable.

In some embodiments, the uplink transmission resource is a Physical Uplink Shared Channel (PUSCH) resource, or the uplink transmission resource is a PUCCH resource.

The different functions of the first piece of information will be explained below.

In one possible implementation, the first information is used to request uplink transmission resources.

Referring to the flowchart shown in Figure 5, when the first information is used to request uplink transmission resources, the model management method provided in this application embodiment may further include the following steps:

S420a, the network device sends the second information, and the corresponding terminal device receives the second information, which is used to indicate uplink transmission resources.

Understandably, when the first message is used to request uplink resources, the network device needs to configure uplink transmission resources for the terminal device after receiving the first message. The network device can indicate (or configure) the uplink transmission resources through the second message.

In some embodiments, the second information may indicate uplink transmission resources by indicating time-domain resource offset and frequency-domain resource location (PRB).

In some embodiments, the second information can indicate uplink transmission resources from the resource pool. The second information can indicate uplink transmission resources by resource index value or offset value relative to a reference resource. This application embodiment does not limit the indication method of transmission resources.

It should also be noted that the resource pool can be predefined or configured by the network device, and this application embodiment does not impose any restrictions on this. The resource pool may include one or more transmission resources. For example, the resource pool may include at least one PUSCH resource and/or at least one PUCCH resource.

In some embodiments, the second information may be downlink control information (DCI). It is understood that network devices can dynamically indicate uplink transmission resources used for model information reporting to terminal devices via DCI signaling. For example, network devices can trigger aperiodic CSI reporting via DCI, causing terminal devices to feed back model information on the PUSCH resource indicated by the DCI. Alternatively, network devices can dynamically indicate a PUCCH resource index from pre-configured PUCCH resources as the uplink transmission resource via DCI.

In some embodiments, the second information may also employ other signaling to indicate uplink transmission resources, such as MAC layer signaling.

In some embodiments, based on S420a and referring to the flowchart shown in FIG5, the model management method provided in this application embodiment further includes the following steps:

S430a, The terminal device sends model information to the network device on the uplink transmission resources indicated by the second information.

Correspondingly, the network device can receive the model information sent by the terminal device on the uplink transmission resources indicated by the second information.

Understandably, after receiving the uplink transmission resources indicated by the network device, the terminal device can report model information on the indicated uplink transmission resources.

Based on this method, the terminal device can request uplink transmission resources from the network device when the first event occurs, thereby reporting model information. This avoids a large number of unnecessary reports in periodic reporting and also avoids the waste of resources caused by pre-allocating uplink resources, thus improving reporting efficiency.

In another possible implementation, the first information is used to indicate uplink transmission resources.

In some embodiments, the first information can indicate uplink transmission resources by indicating time-domain resource offset and frequency-domain resource location (PRB).

In some embodiments, the first information can indicate uplink transmission resources from a resource pool. The first information can indicate uplink transmission resources by a resource index value or an offset value relative to a reference resource. This application embodiment does not limit the indication method of transmission resources.

It should be noted that the resource pool can be predefined or configured by the network device, and this application embodiment does not impose any restrictions on this. The resource pool may include one or more transmission resources. For example, the resource pool may include at least one PUSCH resource and/or at least one PUCCH resource.

In some embodiments, referring to the flowchart shown in FIG6, when the first information is used to indicate uplink transmission resources, the model management method provided in this application embodiment may further include the following steps:

S420b: The terminal device sends model information to the network device on the uplink transmission resource indicated by the first information. Correspondingly, the network device can receive the model information sent by the terminal device on the uplink transmission resource indicated by the first information.

Understandably, after indicating uplink transmission resources to the network device, the terminal device can report model information on the indicated uplink transmission resources.

In some embodiments, the first information is used to indicate whether to send the model information on a pre-configured uplink transmission resource. When the first information indicates to send the model information on a pre-configured uplink transmission resource, the model information is sent to the network device based on the pre-configured uplink transmission resource.

Understandably, before the S410 terminal device sends the first information to the network device, the network device can pre-configure uplink transmission resources for the terminal device through static configuration. These uplink transmission resources include configured grant-based resources and/or periodic PUCCH resources.

For example, network devices can pre-configure configured grant-based (i.e., periodic) PUSCH resources via higher-layer signaling, and the first information is used to indicate whether to send the model information on these PUSCH resources. When the first information indicates on these PUSCH resources... When the model information is sent on the PUSCH resource, the terminal device sends the model information to the network device on the most recent configured grant based PUSCH resource.

For example, network devices can pre-configure periodic PUCCH resources via higher-layer signaling. When the first information indicates that the model information should be sent on these PUCCH resources, the device sends the model information to the network device on the most recent PUCCH resource thereafter.

In some embodiments, the first information can use one bit to indicate whether the model information should be sent on pre-configured uplink transmission resources. For example, when the bit is set to a first value, it indicates that the model information should be sent on the pre-configured uplink transmission resources; when the bit is set to a second value, it indicates that the model information should not be sent on the pre-configured uplink transmission resources. It is understood that by pre-configuring uplink transmission resources and indicating whether to send model information on the pre-configured uplink transmission resources based on the first information, the signaling overhead of the first information can be reduced.

Based on this method, the terminal device can indicate uplink transmission resources to the network device when the first event occurs, and report model information on the indicated uplink transmission resources. This avoids a large number of unnecessary reports in periodic reporting, as well as the waste of resources caused by pre-allocating uplink resources, and improves reporting efficiency.

In another possible implementation, the first information is used to indicate uplink transmission resources and request uplink transmission resources.

Understandably, the first piece of information can be used simultaneously to indicate uplink transmission resources and to request uplink transmission resources. In other words, the terminal device can request the network device to configure the indicated uplink transmission resources for it.

In some embodiments, referring to the flowchart shown in FIG7, when the first information is used to indicate uplink transmission resources and request uplink transmission resources, the model management method provided in this application embodiment may include the following steps:

S420c, the network device sends third information, and the terminal device receives the third information accordingly; wherein, the third information is used to indicate whether to use the uplink transmission resource indicated by the first information, and/or to indicate another uplink transmission resource.

Understandably, after receiving the first information, the network device can determine whether the uplink transmission resources indicated by the terminal device in the first information are occupied.

If the terminal device indicates in the first information that the uplink transmission resources are not occupied, the network device can respond to the terminal device's request and configure the uplink transmission resources indicated by the terminal device. Specifically, the network device can instruct the terminal device to use the uplink transmission resources indicated in the first information for model information reporting through the third information.

If the uplink transmission resource indicated by the terminal device in the first information is already occupied, the network device can re-indicate another uplink transmission resource for the terminal device to report its model information. Specifically, the network device can indicate to the terminal device through the third information that it cannot use the uplink transmission resource indicated by the first information. Furthermore, the network device can indicate another uplink transmission resource for the terminal device through the third information.

It should be noted that the third information can indicate another uplink transmission resource by indicating the time-domain resource offset and the frequency-domain resource location (PRB). Alternatively, the third information can indicate another uplink transmission resource from the resource pool. For example, the third information can indicate another uplink transmission resource by a resource index value or an offset value relative to a reference resource. The embodiments of this application do not limit the indication method of transmission resources.

In some embodiments, the third information can use one bit to indicate whether the uplink transmission resources indicated by the first information are used. When the value of this bit is a first value, it indicates that the uplink transmission resources indicated by the first information are used for model information reporting; when the value of this bit is a second value, it indicates that the uplink transmission resources indicated by the first information cannot be used for model information reporting.

Furthermore, after receiving the third information, the terminal device can, according to the instructions of the third information, use the uplink transmission resources indicated by the first information to report model information, or use another uplink transmission resource indicated by the third information to report model information.

Based on this method, the terminal device can negotiate uplink transmission resources with the network device when the first event occurs, and report model information on the negotiated uplink transmission resources. This avoids a large number of unnecessary reports in periodic reporting, as well as the waste of resources caused by pre-allocating uplink resources, and improves reporting efficiency.

In this embodiment of the application, the above-mentioned model information may include a first reference label and/or model performance monitoring results.

The first reference label includes a reference value for the output of the third model, which is the model currently used by the network device and corresponds to the first model currently used by the terminal device; the first reference label is used for performance monitoring of the third model.

In some embodiments, the terminal device and the network device can communicate using corresponding dual-end models. A first model is deployed on the terminal device, and a third model corresponding to the first model is deployed on the network device.

It should be noted that the correspondence between the first model and the third model can mean that the operations implemented by the first model and the third model are in one-to-one correspondence. For example, if the first model is used for downlink CSI feedback, then the third model is used for decoding downlink CSI feedback information; or, if the first model is used for generating downlink beam information, then the third model can be used for decoding downlink beam feedback information; if the first model is used for channel coding of uplink data, then the third model can be used for channel decoding of uplink data, and so on.

Understandably, the first and third models are deployed in pairs and cannot be used or updated independently. When performing model monitoring on the terminal device side, the terminal device can include reference values of the third model's output results in the reported model information, so that the network device can simultaneously monitor the performance of the deployed third model.

For example, if the third model is used for CSI recovery, the first reference label is the downlink channel information, and the terminal device can feed back the measured downlink channel information to the network side.

In some embodiments, model performance monitoring results include one or more of the following:

Performance metrics of the first model;

The performance metric of the second model that has the best performance metric among one or more second models;

The performance metric of at least one of the one or more second models is better than the performance metric of the first model.

The difference between the performance metrics of the first model and the performance metrics of the reference model;

The difference between the performance metric of at least one of the two second models and the performance metric of the reference model, wherein the performance metric of at least one second model is better than the performance metric of the reference model;

The difference between the performance index of at least one of the first two models and the performance index of the reference model, wherein the performance index of at least one second model is better than that of the first reference model.

The model parameters and/or model structure of the second model with the best performance index among one or more second models;

The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than that of the first model;

Model identification information and/or dataset identification information of the second model with the best performance index among one or more second models;

Model identification information and/or dataset identification information of at least one of the one or more second models, and the performance metric of at least one second model is better than that of the first model;

The model parameters and/or model structure of at least one of the one or more second models, and the performance index of at least one second model is better than the tenth threshold value;

The model identification information and/or dataset identification information of at least one of the one or more second models, and the performance metric of at least one second model is better than the tenth threshold value.

It should be noted that model parameters and/or model structure can be used to determine a new model. Additionally, model identification information can be used by network devices to determine a new model, and dataset identification information can be used by network devices to determine a new model or a dataset for a new model.

Understandably, terminal devices can report parameters of one or more superior second models to the network device, including model parameters, model structure, model identification information, and dataset identification information. This allows the network device to update the model currently used by the terminal device based on this information, thereby further matching the deployed model to the current scenario.

It should be noted that if the terminal device and the network device communicate using a dual-end model, then when the network device updates the model based on the model information reported by the terminal device, it can update the models deployed on both the terminal device side and the network device side simultaneously.

In another possible implementation, the first information is used to request a downlink reference signal.

The downlink reference signal can be used for further model performance monitoring and/or online training. It is understood that the downlink reference signal can be a dedicated reference signal for further model monitoring and online model training. Terminal devices can perform more accurate and complete model performance monitoring and online training based on the downlink reference signal, thereby obtaining more accurate model information.

In some embodiments, referring to the flowchart shown in FIG8, when the first information is used to request a downlink reference signal, the model management method provided in this application embodiment may include the following steps:

S420d: The network device sends downlink reference signals, and the corresponding terminal device receives downlink reference signals.

S430d and terminal devices perform performance monitoring and/or online training on the models associated with the terminal devices based on downlink reference signals.

Understandably, after receiving the first information sent by the terminal device, the network device can configure the downlink reference signal based on the first information.

It should be noted that the downlink reference signal can be a non-periodic reference signal. For example, the downlink reference signal can be CSI-RS, demodulation reference signal (DMRS), and phase tracking reference signal (PTRS), etc., and the embodiments of this application do not limit it.

In this embodiment of the application, performance monitoring of the model associated with the terminal device may be performed on the first model, or on the first model and one or more second models.

In addition, the terminal device can perform performance monitoring on the model associated with the terminal device based on the downlink reference signal. This can be achieved by the terminal device performing channel measurement based on the downlink reference signal, using the measurement result as the input of the aforementioned model, obtaining the output of the aforementioned model through inference, and then obtaining the performance monitoring result of the model based on the output of the model.

It should be noted that the model's performance monitoring results can be characterized by the model's performance metrics, or in other words, the model's performance monitoring results can be the performance metrics corresponding to the model. The model's performance metrics can be referred to in the description of the above embodiments; for brevity... This will not be elaborated upon here.

In this embodiment, the terminal device performs channel measurements based on the downlink reference signal and uses the measured data as a training dataset for online model training. Through online model training, the terminal can update the model parameters, thereby further adapting the model to the current scenario.

Understandably, the downlink reference signal can be a dedicated reference signal used for model monitoring and online model training. Terminal devices can perform more accurate and comprehensive model performance monitoring and online training based on the downlink reference signal.

In some embodiments, the first information may, in addition to requesting a downlink reference signal, also indicate and/or request uplink transmission resources.

Understandably, the terminal device requests a downlink reference signal for model performance monitoring. Furthermore, the terminal device needs to report the model information obtained from the downlink reference signal-based model performance monitoring to the network device. Based on this, the terminal device can simultaneously request the downlink reference signal and/or indicate and/or request uplink transmission resources. In this way, the terminal device can perform model performance monitoring based on the requested downlink reference signal, obtain model information, and then report the model information based on the indicated and/or requested uplink transmission resources.

It should be noted that the method of indicating and/or requesting uplink transmission resources in the first information can be referred to the description in the above embodiments, and will not be repeated here for the sake of brevity.

In some embodiments, where the first information is used to request a downlink reference signal and also to indicate and/or request uplink transmission resources, the model management method provided in this application embodiment may further include the following steps:

The terminal device sends performance monitoring results on uplink transmission resources. The performance monitoring results are obtained by monitoring the performance of the model based on the downlink reference signal.

Understandably, the terminal device may send performance monitoring results obtained by monitoring the model based on the downlink reference signal on the first information indication and/or request for uplink transmission resources.

In some embodiments, the time-domain resources for transmitting downlink reference signals are spaced apart from the time-domain resources for uplink transmission resources by a first duration; the first duration is determined based on the processing capability of the terminal device.

It should be noted that the process of obtaining performance monitoring results for the model associated with the terminal device based on the downlink reference signal takes a certain amount of time. This time is related to the processing power of the terminal device. The stronger the processing power of the terminal device, the shorter the time required.

Therefore, the time-domain resources for receiving downlink reference signals by the terminal device need to be separated from the time-domain resources for uplink transmission of transmission model performance monitoring results by a certain time interval (referred to as the first time interval in this embodiment) so that the terminal device can obtain the performance monitoring results of the model based on the downlink reference signals within this time interval.

It should be noted that the first duration can be greater than or equal to the duration required for the terminal device to perform performance monitoring on the model associated with the terminal device based on the downlink reference signal and obtain the performance monitoring results of the model.

In some embodiments, the terminal device may report its processing capabilities to the network device. For example, the terminal device may directly report to the network device the time required for the terminal device to perform performance monitoring on the model associated with the terminal device based on the downlink reference signal to obtain the performance monitoring results of the model, so that the network device can configure the downlink reference signal and uplink transmission resources based on the processing capabilities of the terminal device.

For example, the terminal device may indicate its processing capabilities in the first information.

This method enables terminal devices to report and monitor model information based on a first event-driven approach. This approach does not require network devices to pre-allocate reporting resources and reference signals, which can effectively reduce resource overhead caused by model monitoring and training, avoid unnecessary reporting of model information, and improve resource utilization efficiency.

The methods provided in the embodiments of this application will be described in detail below with reference to specific application scenarios.

The model management method provided in this application embodiment is an event-driven model management method. The terminal device determines whether a predefined event is met based on the model performance monitoring results, and when the event is met, sends first information to the network device to request uplink channel resources or downlink reference signals. The uplink channel resources are used for model information reporting, and the downlink reference signals are used for further model monitoring or online training. By using an event-driven approach for model management and reporting, the network device does not need to pre-allocate reporting resources and reference signals, which can effectively reduce the resource overhead caused by model management, especially model monitoring, and avoid unnecessary model information reporting.

The above method will be explained based on two different scenarios below.

Example 1

Referring to Figure 9, the model management method provided in this embodiment includes the following steps:

S1. The terminal device performs performance monitoring of the AI model and determines whether a predefined event has occurred based on the monitoring results.

It should be noted that the predefined event can be the first event mentioned above.

In this embodiment of the application, the terminal device can monitor the performance of the AI model in the following two ways:

Method 1: The terminal device periodically monitors the performance of the AI model, and the monitoring period can be configured by the network device.

Method 2: Only when the first condition is met will the performance of the AI model be monitored.

The first condition includes at least one of the following conditions:

Condition 1: The channel's large-scale parameters measured by the terminal equipment change beyond a certain range. These large-scale parameters may include the channel delay power spectrum, the presence of a LOS path, and Doppler shift. For example, channel delay changes exceeding a certain range, there may be a shift between LOS and NLOS, and Doppler shift changes exceeding a certain range.

Condition 2: The metric obtained by the terminal device exceeds a certain range. The metric can be RSRP, SINR, SGCS, etc., for example, their values are higher or lower than a certain threshold.

Condition 3: The serving cell of the terminal device has changed. In other words, the terminal device has undergone cell handover.

In this embodiment, the terminal device can obtain the performance indicators of the AI model through performance monitoring. The AI model here can be the currently used model, or one or more candidate AI models that are not currently in use.

It should be noted that performance metrics are used to evaluate the performance of an AI model. These can be values of physical quantities such as SINR, BLER, throughput, SGCS, GCS, NMSE, or the difference between their values and reference values. Performance metrics can be indicated by network devices.

In one implementation, the terminal device obtains the performance metrics of the AI model by comparing the output of the AI model with the AI reference label (ground truth label).

For example, the AI reference label can be channel information measured by the terminal device. The terminal device can use the squared cosine similarity (SGCS) between the measured channel information and the channel information obtained by the AI model as a performance metric for the AI model.

For example, the AI reference label can be the expected channel capacity or the expected throughput (e.g., a capacity ceiling calculated theoretically). The terminal device can calculate the corresponding channel capacity or throughput of the AI model based on the precoding matrix output by the AI model. The difference between the calculated channel capacity and the expected channel capacity, or between the calculated throughput and the expected throughput, is used as a performance indicator.

In another implementation, the terminal device obtains a performance metric by comparing the output of the AI model with the output obtained by a non-AI method. For example, the terminal device calculates the corresponding SINR based on the output of the AI model and compares it with the SINR calculated based on a traditional method, thereby obtaining the corresponding SINR difference as a performance metric.

In another implementation, the terminal device can directly calculate performance metrics based on the output of the AI model. For example, the terminal device can estimate the corresponding SINR, NMSE, BLER, or throughput based on the output of the AI model, and use these as performance metrics.

In this embodiment of the application, the predefined event is one of the following events:

The current model's performance metrics are worse than the first threshold.

The performance metrics of the current model and all candidate models are worse than the second threshold.

At least one new model has better performance metrics than the current model.

At least one new model has a better performance metric than the current model, and the difference in performance metrics is greater than the third threshold.

At least one new model has a performance metric that is better than the fourth threshold.

The current model's performance metrics are worse than the fifth threshold, and at least one new model's performance metrics are better than the sixth threshold;

The absolute value of the difference between the performance metrics of the current model and the new model is less than the seventh threshold.

At least one new model outperforms the reference model in terms of performance metrics;

At least one new model has a better performance metric than the reference model, and the difference in performance metrics is greater than the eighth threshold.

Several new models outperform the current model in terms of performance metrics;

Several new models outperform the current model in terms of performance metrics, and the difference in performance metrics is greater than the ninth threshold.

At least one new model has a performance metric that is no worse than the current model, and the complexity of the new model is lower than that of the current model; in one implementation, the complexity can be reflected by the number of model parameters or the amount of computation in inference.

It should be noted that the new model mentioned above may be a different model from the current model among the candidate models; the candidate model is a model that the network device pre-configures for the terminal device, or a model that the terminal device pre-reports to the network device, or a model that the terminal device and the network device pre-agree on (obtained through offline training).

It should also be noted that the predefined event can also be a combination of the above events. In this case, the terminal device will only be driven to send the first information when all the events in the combination are satisfied.

S2. When a predefined event occurs (is satisfied), the terminal device sends first information to the network device, the first information being used to request or indicate uplink transmission resources.

The uplink channel resources are used for subsequent model information reporting.

It should be noted that the first message is carried by MAC layer signaling or PUCCH.

It should also be noted that the uplink transmission resources are either PUSCH resources or PUCCH resources.

In some embodiments, if the first information is used to indicate uplink channel resources, then the first information is used to indicate the uplink channel resources used for current model information reporting from the physical resources pre-configured by the network device.

S3, The network device receives the first information sent by the terminal device.

If the first message is used to request uplink transmission resources, then after receiving the first message, the network device needs to configure the uplink transmission resources of the terminal device.

In one implementation, the uplink transmission resources are uplink transmission resources indicated by the network device from a predefined resource pool via DCI signaling. The network device may also use other signaling, such as MAC layer signaling, to indicate uplink transmission resources from the predefined resource pool.

S4. The terminal equipment reports model information on the uplink transmission resources.

Understandably, if the first information is used to request uplink channel resources, then after sending the first information, the terminal device needs to receive the uplink channel resources configured by the network device and report model information on the uplink transmission resources.

In one implementation, the uplink channel resources are uplink channel resources indicated by the network device from a predefined resource pool via DCI signaling.

If the first information is used by the terminal device to indicate uplink channel resources from a predefined resource pool, then after sending the first information, the terminal device reports model information on the uplink channel resources.

Specifically, the model information may be the model's reference label or the model's performance monitoring results.

The reference labels for the model are a set of reference values output by the network-side model, used for performance monitoring of the network-side model. For example, if the network-side model is used for CSI recovery, the reference labels are downlink channel information. The terminal device can feed back the measured downlink channel information to the network side for network device performance monitoring.

In this embodiment of the application, the performance monitoring results of the model include at least one of the following:

Current model performance metrics;

The performance metrics of one or more new models with the best performance among the candidate models;

At least one new model has a performance metric that is better than the current model;

The difference in performance metrics between the new model and the current model, wherein the performance metrics of the new model are better than those of the current model;

The difference in performance metrics between the current model and the reference model;

The difference in performance metrics between at least one new model and a reference model, wherein the new model performs better than the reference model;

The difference in performance metrics between at least one new model and a reference model, wherein the performance metrics of the new model are better than those of the current model;

At least one new model with better model parameters and/or model structure than the current model;

At least one new model has a corresponding model ID or dataset ID, and the performance metrics of the new model are better than those of the current model; wherein, the model ID can be used by the network device to determine the new model, and the dataset ID can be used by the network device to determine the new model or the dataset of the new model.

S5. The network device receives the model information reported by the terminal device on the uplink transmission resources.

Specifically, the uplink transmission resources are the uplink channel resources configured by the network device in response to the first information, or the uplink channel resources indicated by the terminal device from a predefined resource pool through the first information.

Based on this method, terminal devices can request/instruct uplink resources from network devices when a predefined event occurs, thereby reporting model information. This avoids a large number of unnecessary reports in periodic reporting and also avoids resource waste caused by pre-allocating uplink resources, thus improving reporting efficiency.

Example 2

Referring to Figure 10, the model management method provided in this embodiment includes the following steps:

S1. The terminal device performs performance monitoring of the AI model and determines whether a predefined event has occurred based on the monitoring results.

It should be noted that the performance monitoring method is described in Example 1, and will not be repeated here for the sake of brevity.

In this embodiment, predefined events are pre-configured by the network device to the terminal device. Specifically, the network device may pre-indicate one or more of the following events:

The current model's performance metrics are worse than the first threshold.

The performance metrics of the current model and all candidate models are worse than the second threshold.

At least one new model has better performance metrics than the current model.

At least one new model has a better performance metric than the current model, and the difference in performance metrics is greater than the third threshold.

At least one new model has a performance metric that is better than the fourth threshold.

The current model's performance metrics are worse than the fifth threshold, and at least one new model's performance metrics are better than the sixth threshold;

The absolute value of the difference between the performance metrics of the current model and the new model is less than the seventh threshold.

At least one new model outperforms the reference model in terms of performance metrics;

At least one new model has a better performance metric than the reference model, and the difference in performance metrics is greater than the eighth threshold.

Several new models outperform the current model in terms of performance metrics;

Several new models outperform the current model in terms of performance metrics, and the difference in performance metrics is greater than the ninth threshold.

At least one new model has a performance metric that is no worse than the current model, and the complexity of the new model is lower than that of the current model.

It should be noted that when a network device indicates multiple events, the terminal device will only send the first information if all indicated events are met. For example, the events could be that the performance metric of the current model is worse than a first threshold and that the performance metric of at least one new model is better than the current model.

In this embodiment, the threshold value can be configured by the network device to the terminal, or it can be a fixed value agreed upon in advance between the terminal device and the network device.

S2. When the event occurs (is satisfied), the terminal device sends first information to the network device, the first information being used to request a downlink reference signal.

The downlink reference signal is used for further model monitoring or online training.

It should be noted that the first message is carried by MAC layer signaling or PUCCH.

It should also be noted that the uplink transmission resources are either PUSCH resources or PUCCH resources.

S3. The network device receives the first information sent by the terminal device and configures the downlink reference signal according to the first information.

Among them, network devices can trigger aperiodic reference signals, for example, triggering aperiodic CSI-RS signals via DCI signaling for further model monitoring or online training of terminal devices.

S4. The terminal device receives the downlink reference signal and performs further model monitoring or online training based on the reference signal.

In one implementation, the terminal device performs channel measurements based on the downlink reference signal, uses the measurement results as input to a model, obtains the model's output through inference, and performs performance monitoring based on the model's output. Here, the model can be the current model or a candidate model.

In another implementation, the terminal device performs channel measurements based on the downlink reference signal and uses the measured data as a training dataset for online model training. Through online model training, the terminal can update the model parameters, thereby further adapting the model to the current scenario.

Based on this method, terminal devices can request downlink reference signals from network devices when predefined events occur, thereby enabling further model monitoring and online training. This avoids resource waste caused by pre-configuring periodic downlink reference signals and improves resource efficiency.

The preferred embodiments of this application have been described in detail above with reference to the accompanying drawings. However, this application is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this application, various simple modifications can be made to the technical solutions of this application, and these simple modifications all fall within the protection scope of this application. For example, the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this application will not describe the various possible combinations separately. Furthermore, various different embodiments of this application can also be arbitrarily combined, as long as they do not violate the spirit of this application, they should also be considered as the content disclosed in this application. Moreover, without conflict, the various embodiments and/or the technical features in the various embodiments described in this application can be arbitrarily combined with the prior art, and the resulting technical solutions should also fall within the protection scope of this application.

It should also be understood that in the various method embodiments of this application, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. Furthermore, in the embodiments of this application, the terms "downlink," "uplink," and "sidelink" are used to indicate the transmission direction of signals or data. "Downlink" indicates that the transmission direction of signals or data is a first direction from the site to the user equipment in the cell; "uplink" indicates that the transmission direction of signals or data is a second direction from the user equipment in the cell to the site; and "sidelink" indicates that the transmission direction of signals or data is a third direction from user equipment 1 to user equipment 2. For example, "downlink signal" indicates that the transmission direction of the signal is the first direction. Additionally, in the embodiments of this application, the term "and/or" is merely a description of the association relationship between related objects, indicating that three relationships can exist. Specifically, A and/or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, the character "/" in this document generally indicates that the preceding and following related objects have an "or" relationship.

Figure 11 is a schematic diagram of the structure of the model management device 1100 provided in an embodiment of this application, which is applied to a terminal device. As shown in Figure 11, the model management device 1100 includes:

The first sending unit 1101 is configured to send first information to the network device in the event of a first event; the first event is related to the performance monitoring results of the model associated with the terminal device.

The first information is used for one or more of the following:

Requesting uplink transmission resources;

Indicates uplink transmission resources; the uplink transmission resources are used to send model information;

Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training.

In some embodiments, the model management device 1100 further includes a model monitoring unit configured to perform performance monitoring on the model associated with the terminal device when a first condition is met.

The measured large-scale parameters of the channel exceed the first range;

The measured quantity exceeds the second range;

The service area has changed;

The performance monitoring of the model associated with the terminal device is used to determine whether the first event has occurred.

In some embodiments, the model associated with the terminal device includes a first model, or includes a first model and one or more second models, wherein the first model is the model currently used by the terminal device, and the second model is a model supported by the terminal device and is different from the first model;

The first event includes one or more of the following:

The performance metrics of the first model are worse than the first threshold value;

The performance metrics of the first model and the one or more second models are all worse than the second threshold value;

At least one of the one or more second models has a better performance metric than the first model.

At least one of the one or more second models has a better performance index than the first model, and the difference between the performance index of the at least one second model and the performance index of the first model is greater than a third threshold value.

At least one of the one or more second models has a performance metric that is better than the fourth threshold value;

The performance metric of the first model is worse than the fifth threshold, and the performance metric of at least one of the one or more second models is better than the sixth threshold;

The absolute value of the difference between the performance metric of the first model and the performance metric of at least one of the one or more second models is less than the seventh threshold value;

At least one of the one or more second models has a performance metric that is better than that of the reference model;

At least one of the one or more second models has a performance metric that is better than that of the reference model, and the difference between the performance metric of the at least one second model and the performance metric of the reference model is greater than the eighth threshold.

The performance metrics of two or more of the one or more second models are better than those of the first model.

In the one or more second models, the performance index of two or more second models is better than that of the first model, and the difference between the performance index of the two or more second models and the performance index of the first model is greater than the ninth threshold value.

The performance metric of at least one of the one or more second models is no worse than that of the first model, and the complexity of the at least one second model is lower than that of the first model.

In some embodiments, one or more of the first threshold value to the ninth threshold value are determined based on any of the following:

Protocol predefined;

The first configuration information sent by the network device.

In some embodiments, the first event is determined based on any of the following methods:

Protocol predefined;

The second configuration information sent by the network device.

In some embodiments, the first information is carried via MAC signaling or via PUCCH.

In some embodiments, the uplink transmission resource is a PUSCH resource, or a PUCCH resource.

In some embodiments, the model management device 1100 further includes a first receiving unit. The first receiving unit is configured to allow the terminal device to receive second information, the second information being used to indicate uplink transmission resources.

In some embodiments, the first sending unit 1101 is further configured to send the model information to the network device on the uplink transmission resources indicated by the second information.

In some embodiments, the second information is DCI, used to indicate the uplink transmission resources from a predefined resource pool.

In some embodiments, the first information is used to indicate uplink transmission resources, and the first sending unit 1101 is further configured to send the model information to the network device on the uplink transmission resources indicated by the first information;

or,

The first information is used to indicate whether to send the model information on the pre-configured uplink transmission resources. The first sending unit 1101 is also configured to send the model information to the network device based on the pre-configured uplink transmission resources when the first information indicates that the model information is to be sent on the pre-configured uplink transmission resources.

In some embodiments, the first information is used to indicate uplink transmission resources and request uplink transmission resources, and the first receiving unit is further configured to receive third information; the third information is used to indicate whether to use the uplink transmission resource indicated by the first information, and/or to indicate another uplink transmission resource.

In some embodiments, the model information includes a first reference label and/or model performance monitoring results.

In some embodiments, the first reference label includes a reference value of the output result of the third model, the third model being the model currently used by the network device, and the third model corresponding to the first model currently used by the terminal device; the first reference label is used for performance monitoring of the third model.

In some embodiments, the model associated with the terminal device includes a first model, or includes a first model and one or more second models, wherein the first model is the model currently used by the terminal device, and the second model is a model supported by the terminal device and is different from the first model;

The model performance monitoring results include one or more of the following:

Performance metrics of the first model;

The performance index of the second model with the best performance index among the one or more second models;

The performance metric of at least one of the one or more second models is better than the performance metric of the first model.

The difference between the performance metrics of the first model and the performance metrics of the reference model;

The difference between the performance index of at least one of the one or more second models and the performance index of the reference model, wherein the performance index of the at least one second model is better than the performance index of the reference model;

The difference between the performance index of at least one of the one or more second models and the performance index of the reference model, wherein the performance index of the at least one second model is better than the performance index of the first reference model;

The model parameters and/or model structure of the second model with the best performance index among the one or more second models;

The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than that of the first model;

The model identification information and/or dataset identification information of the second model with the best performance index among the one or more second models;

The model identification information and/or dataset identification information of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the performance index of the first model;

The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the tenth threshold value;

The model identification information and/or dataset identification information of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the tenth threshold value.

In some embodiments, the first information is used to request a downlink reference signal, and the first receiving unit is further configured to receive the downlink reference signal;

The model monitoring unit is also configured to perform performance monitoring and/or online training on the model associated with the terminal device based on the downlink reference signal.

In some embodiments, the first information is further used to request uplink transmission resources and/or indicate uplink transmission resources. The first sending unit 1101 is also configured to send performance monitoring results on the uplink transmission resources, the performance monitoring results being obtained by performing performance monitoring on the model based on the downlink reference signal.

In some embodiments, the time-domain resources for transmitting the downlink reference signal are spaced apart from the time-domain resources for the uplink transmission resources by a first duration; the first duration is determined based on the processing capability of the terminal device.

In some embodiments, the model associated with the terminal device is a model configured by the network device for the terminal device;

Alternatively, the model associated with the terminal device is the model reported by the terminal device to the network device;

Alternatively, the model associated with the terminal device is a model pre-agreed upon by the terminal device and the network device.

Those skilled in the art should understand that the description of the model management device in the embodiments of this application can be understood with reference to the description of the model management method in the embodiments of this application.

Figure 12 is a schematic diagram of the structure of the model management device 1200 provided in an embodiment of this application, which is applied to a network device. As shown in Figure 12, the model management device 1200 includes:

The second receiving unit 1201 is configured to receive first information sent by the terminal device; the first information is used for one or more of the following:

Requesting uplink transmission resources;

Indicates uplink transmission resources; the uplink transmission resources are used to send model information;

Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training.

In some embodiments, the first information is sent upon the occurrence of a first event; the first event relates to the performance monitoring results of a model associated with the terminal device.

In some embodiments, the model associated with the terminal device includes a first model, or includes a first model and one or more second models, wherein the first model is the model currently used by the terminal device, and the second model is the model supported by the terminal device. The model is different from the first model;

The first event includes one or more of the following:

The performance metrics of the first model are worse than the first threshold value;

The performance metrics of the first model and the one or more second models are all worse than the second threshold value;

At least one of the one or more second models has a better performance metric than the first model.

At least one of the one or more second models has a better performance index than the first model, and the difference between the performance index of the at least one second model and the performance index of the first model is greater than a third threshold value.

At least one of the one or more second models has a performance metric that is better than the fourth threshold value;

The performance metric of the first model is worse than the fifth threshold, and the performance metric of at least one of the one or more second models is better than the sixth threshold;

The absolute value of the difference between the performance metric of the first model and the performance metric of at least one of the one or more second models is less than the seventh threshold value;

At least one of the one or more second models has a performance metric that is better than that of the reference model;

At least one of the one or more second models has a performance metric that is better than that of the reference model, and the difference between the performance metric of the at least one second model and the performance metric of the reference model is greater than the eighth threshold.

The performance metrics of two or more of the one or more second models are better than those of the first model;

The performance metrics of two or more of the one or more second models are better than those of the first model, and the difference between the performance metrics of the two or more second models and the performance metrics of the first model is greater than the ninth threshold value.

The performance metric of at least one of the one or more second models is no worse than that of the first model, and the complexity of the at least one second model is lower than that of the first model.

In some embodiments, one or more of the first threshold value to the ninth threshold value are determined based on any of the following:

Protocol predefined;

The first configuration information sent by the network device.

In some embodiments, the first event is determined based on any of the following methods:

Protocol predefined;

The second configuration information sent by the network device.

In some embodiments, the first information is carried via MAC signaling or via PUCCH.

In some embodiments, the uplink transmission resource is a PUSCH resource, or the uplink transmission resource is a PUCCH resource.

In some embodiments, the first information is used to request uplink transmission resources, and the model management device 1200 further includes a second sending unit configured to send second information, the second information being used to indicate uplink transmission resources.

In some embodiments, the second information is a DCI, which is used to indicate the uplink transmission resources from a predefined resource pool.

In some embodiments, the first information is used to indicate uplink transmission resources, and the second receiving unit 1201 is further configured to receive the model information on the uplink transmission resources indicated by the first information;

or,

The first information is used to indicate whether the model information is sent on a pre-configured uplink transmission resource. The second receiving unit 1201 is further configured to receive the model information based on the pre-configured uplink transmission resource when the first information indicates that the model information is sent on the pre-configured uplink transmission resource.

In some embodiments, the first information is used to indicate uplink transmission resources and request uplink transmission resources, and the second sending unit is configured to send third information; the third information is used to indicate whether to use the uplink transmission resource indicated by the first information, and/or to indicate another uplink transmission resource.

In some embodiments, the model information includes a first reference label and/or model performance monitoring results.

In some embodiments, the first reference label includes a reference value of the output result of the third model, the third model being the model currently used by the network device, and the third model corresponding to the first model currently used by the terminal device; the first reference label is used for performance monitoring of the third model.

In some embodiments, the model associated with the terminal device includes a first model, or includes a first model and one or more second models, wherein the first model is the model currently used by the terminal device, and the second model is a model supported by the terminal device and is different from the first model;

The model performance monitoring results include one or more of the following:

Performance metrics of the first model;

The performance index of the second model with the best performance index among the one or more second models;

The performance metric of at least one of the one or more second models is better than the performance metric of the first model.

The difference between the performance metrics of the first model and the performance metrics of the reference model;

The difference between the performance index of at least one of the one or more second models and the performance index of the reference model, wherein the performance index of the at least one second model is better than the performance index of the reference model;

The difference between the performance index of at least one of the one or more second models and the performance index of the reference model, wherein the performance index of the at least one second model is better than the performance index of the first reference model;

The model parameters and/or model structure of the second model with the best performance index among the one or more second models;

The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than that of the first model;

The model identification information and/or dataset identification information of the second model with the best performance index among the one or more second models;

The model identification information and/or dataset identification information of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the performance index of the first model;

The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the tenth threshold value;

The model identification information and/or dataset identification information of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the tenth threshold value.

In some embodiments, the first information is used to request a downlink reference signal, and the second transmitting unit is further configured to transmit the downlink reference signal; the downlink reference signal is used by the terminal device to perform performance monitoring and/or online training of the model associated with the terminal device.

In some embodiments, the first information is further used to request uplink transmission resources and/or indicate uplink transmission resources, and the second receiving unit 1201 is further configured to receive performance monitoring results sent by the terminal device on the uplink transmission resources, wherein the performance monitoring results are obtained by performing performance monitoring on the model based on the downlink reference signal.

In some embodiments, the time-domain resources for transmitting the downlink reference signal are spaced apart from the time-domain resources for the uplink transmission resources by a first duration; the first duration is determined based on the processing capability of the terminal device.

In some embodiments, the model associated with the terminal device is the model configured by the network device for the terminal device;

Alternatively, the model associated with the terminal device is the model reported by the terminal device to the network device;

Alternatively, the model associated with the terminal device is a model pre-agreed upon by the terminal device and the network device.

Those skilled in the art should understand that the description of the model management device in the embodiments of this application can be understood with reference to the description of the model management method in the embodiments of this application.

Figure 13 is a schematic structural diagram of a communication device 1300 provided in an embodiment of this application. This communication device can be a terminal device or a network device. The communication device 1300 shown in Figure 13 includes a processor 1310, which can call and run computer programs from memory to implement the methods in the embodiments of this application.

Optionally, as shown in FIG13, the communication device 1300 may further include a memory 1320. The processor 1310 may retrieve and run computer programs from the memory 1320 to implement the methods described in the embodiments of this application.

The memory 1320 can be a separate device independent of the processor 1310, or it can be integrated into the processor 1310.

Optionally, as shown in FIG13, the communication device 1300 may further include a transceiver 1330, and the processor 1310 may control the transceiver 1330 to communicate with other devices. Specifically, it may send information or data to other devices or receive information or data sent by other devices.

The transceiver 1330 may include a transmitter and a receiver. The transceiver 1330 may further include an antenna, and the number of antennas may be one or more.

Optionally, the communication device 1300 may specifically be a network device in the embodiments of this application, and the communication device 1300 may implement the corresponding processes implemented by the network device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

Optionally, the communication device 1300 may specifically be a mobile terminal/terminal device in the embodiments of this application, and the communication device 1300 may implement the corresponding processes implemented by the mobile terminal/terminal device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

Figure 14 is a schematic structural diagram of a chip according to an embodiment of this application. The chip 1400 shown in Figure 14 includes a processor 1410, which can call and run computer programs from memory to implement the methods in the embodiments of this application.

Optionally, as shown in FIG14, chip 1400 may further include memory 1420. Processor 1410 may retrieve and run computer programs from memory 1420 to implement the methods in the embodiments of this application.

The memory 1420 can be a separate device independent of the processor 1410, or it can be integrated into the processor 1410.

Optionally, the chip 1400 may also include an input interface 1430. The processor 1410 can control the input interface 1430 to communicate with other devices or chips; specifically, it can acquire information or data sent by other devices or chips.

Optionally, the chip 1400 may also include an output interface 1440. The processor 1410 can control the output interface 1440 to communicate with other devices or chips, specifically, to output information or data to other devices or chips.

Optionally, the chip can be applied to the network device in the embodiments of this application, and the chip can implement the corresponding processes implemented by the network device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

Optionally, the chip can be applied to the mobile terminal/terminal device in the embodiments of this application, and the chip can implement the corresponding processes implemented by the mobile terminal/terminal device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

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

This application also provides a computer storage medium storing one or more programs, which can be executed by one or more processors to implement the methods in this application.

Figure 15 is a schematic block diagram of a communication system 1500 provided in an embodiment of this application. As shown in Figure 15, the communication system 1500 includes a terminal device 1510 and a network device 1520.

The terminal device 1510 can be used to implement the corresponding functions implemented by the terminal device in the above method, and the network device 1520 can be used to implement the corresponding functions implemented by the network device in the above method. For the sake of brevity, it will not be described in detail here.

It should be understood that the processor in the embodiments of this application may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method embodiments can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor described above may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

It should be understood that the above-described memory is exemplary but not restrictive. For example, the memory in the embodiments of this application may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM), etc. That is to say, the memory in the embodiments of this application is intended to include, but is not limited to, these and any other suitable types of memory.

This application also provides a computer-readable storage medium for storing computer programs.

Optionally, the computer-readable storage medium can be applied to the network device in the embodiments of this application, and the computer program causes the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

Optionally, the computer-readable storage medium can be applied to the mobile terminal/terminal device in the embodiments of this application, and the computer program causes the computer to execute the corresponding processes implemented by the mobile terminal/terminal device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

This application also provides a computer program product, including computer program instructions.

Optionally, the computer program product can be applied to the network device in the embodiments of this application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of this application. For the sake of brevity, they will not be described in detail here.

Optionally, the computer program product can be applied to the mobile terminal/terminal device in the embodiments of this application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the mobile terminal/terminal device in the various methods of the embodiments of this application. For the sake of brevity, they will not be described in detail here.

This application also provides a computer program.

Optionally, the computer program can be applied to the network device in the embodiments of this application. When the computer program is run on the computer, it causes the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

Optionally, the computer program can be applied to the mobile terminal/terminal device in the embodiments of this application. When the computer program is run on a computer, it causes the computer to execute the corresponding processes implemented by the mobile terminal/terminal device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

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

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

In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

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

The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims (45)

A model management method, the method comprising: In the event of a first event, the terminal device sends first information to the network device; the first event is related to the performance monitoring results of the model associated with the terminal device. The first information is used for one or more of the following: Requesting uplink transmission resources; Indicates uplink transmission resources; the uplink transmission resources are used to send model information; Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training. The method according to claim 1, further comprising: Under the condition that the terminal device meets the first condition, the terminal device performs performance monitoring on the model associated with the terminal device: The changes in the large-scale parameters of the channel measured by the terminal device exceed the first range; The measurement obtained by the terminal device exceeds the second range; The serving cell of the terminal device has changed; The performance monitoring of the model associated with the terminal device is used to determine whether the first event has occurred. According to the method of claim 1 or 2, the model associated with the terminal device includes a first model, or includes a first model and one or more second models, wherein the first model is the model currently used by the terminal device, and the second model is a model supported by the terminal device and is different from the first model; The first event includes one or more of the following: The performance metrics of the first model are worse than the first threshold value; The performance metrics of the first model and the one or more second models are all worse than the second threshold value; At least one of the one or more second models has a better performance metric than the first model. At least one of the one or more second models has a better performance index than the first model, and the difference between the performance index of the at least one second model and the performance index of the first model is greater than a third threshold value. At least one of the one or more second models has a performance metric that is better than the fourth threshold value; The performance metric of the first model is worse than the fifth threshold, and the performance metric of at least one of the one or more second models is better than the sixth threshold; The absolute value of the difference between the performance metric of the first model and the performance metric of at least one of the one or more second models is less than the seventh threshold value; At least one of the one or more second models has a performance metric that is better than that of the reference model; At least one of the one or more second models has a performance metric that is better than that of the reference model, and the difference between the performance metric of the at least one second model and the performance metric of the reference model is greater than the eighth threshold. The performance metrics of two or more of the one or more second models are better than those of the first model. The performance metrics of two or more of the one or more second models are better than the performance metrics of the first model, and the difference between the performance metrics of the two or more second models and the performance metrics of the first model is greater than the ninth threshold value. The performance metric of at least one of the one or more second models is no worse than that of the first model, and the complexity of the at least one second model is lower than that of the first model. According to the method of claim 3, one or more of the first threshold value to the ninth threshold value are determined based on any of the following: Protocol predefined; The first configuration information sent by the network device. The method according to any one of claims 1-4, wherein the first event is determined based on any one of the following: Protocol predefined; The second configuration information sent by the network device. The method according to any one of claims 1-5, wherein the first information is carried by Media Access Control (MAC) signaling, or the first information is carried by Physical Uplink Control Channel (PUCCH). The method according to any one of claims 1-6, wherein the uplink transmission resource is a Physical Uplink Data Channel (PUSCH) resource, or the uplink transmission resource is a Physical Uplink Control Channel (PUCCH) resource. The method according to any one of claims 1-7, wherein the first information is used to request uplink transmission resources, the method The law also includes: The terminal device receives second information, which is used to indicate uplink transmission resources. The method according to claim 8, wherein the method further comprises: The terminal device sends the model information to the network device on the uplink transmission resources indicated by the second information. According to the method of claim 8 or 9, the second information is downlink control information (DCI) used to indicate the uplink transmission resources from a predefined resource pool. The method according to any one of claims 1-7, further comprising: The first information is used to indicate uplink transmission resources, and the terminal device sends the model information to the network device on the uplink transmission resources indicated by the first information; or, The first information is used to indicate whether to send the model information on the pre-configured uplink transmission resources. When the first information indicates to send the model information on the pre-configured uplink transmission resources, the model information is sent to the network device based on the pre-configured uplink transmission resources. The method according to any one of claims 1-7, wherein the first information is used to indicate uplink transmission resources and request uplink transmission resources, the method further comprising: The terminal device receives third information; the third information is used to indicate whether to use the uplink transmission resource indicated by the first information, and/or to indicate another uplink transmission resource. The method according to any one of claims 1-12, wherein the model information includes a first reference label, and/or model performance monitoring results. The method according to claim 13, wherein, The first reference label includes a reference value for the output result of the third model, which is the model currently used by the network device and corresponds to the first model currently used by the terminal device; the first reference label is used for performance monitoring of the third model. According to the method of claim 13 or 14, the model associated with the terminal device includes a first model, or includes a first model and one or more second models, wherein the first model is the model currently used by the terminal device, and the second model is a model supported by the terminal device and is different from the first model; The model performance monitoring results include one or more of the following: Performance metrics of the first model; The performance index of the second model with the best performance index among the one or more second models; The performance metric of at least one of the one or more second models is better than the performance metric of the first model. The difference between the performance metrics of the first model and the performance metrics of the reference model; The difference between the performance index of at least one of the one or more second models and the performance index of the reference model, wherein the performance index of the at least one second model is better than the performance index of the reference model; The difference between the performance index of at least one of the one or more second models and the performance index of the reference model, wherein the performance index of the at least one second model is better than the performance index of the first reference model; The model parameters and/or model structure of the second model with the best performance index among the one or more second models; The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than that of the first model; The model identification information and/or dataset identification information of the second model with the best performance index among the one or more second models; The model identification information and/or dataset identification information of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the performance index of the first model; The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the tenth threshold value; The model identification information and/or dataset identification information of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the tenth threshold value. The method according to any one of claims 1-15, wherein the first information is used to request a downlink reference signal, the method further comprising: The terminal device receives the downlink reference signal; The terminal device performs performance monitoring and/or online training on the model associated with the terminal device based on the downlink reference signal. According to the method of claim 16, wherein the first information is further used to request uplink transmission resources, and/or to indicate uplink transmission resources, the method further includes: The terminal device sends performance monitoring results on the uplink transmission resources. The performance monitoring results are obtained by monitoring the performance of the model based on the downlink reference signal. The method according to claim 17, wherein, The time-domain resources for transmitting the downlink reference signal are spaced apart from the time-domain resources for transmitting the uplink transmission resources by a first duration; the first duration is determined based on the processing capability of the terminal device. The method according to any one of claims 1-18, wherein the model associated with the terminal device is a model configured by the network device for the terminal device; Alternatively, the model associated with the terminal device is the model reported by the terminal device to the network device; Alternatively, the model associated with the terminal device is a model pre-agreed upon by the terminal device and the network device. A model management method, the method comprising: The network device receives first information sent by the terminal device; the first information is used for one or more of the following: Requesting uplink transmission resources; Indicates uplink transmission resources; the uplink transmission resources are used to send model information; Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training. According to the method of claim 20, the first information is sent upon the occurrence of a first event; the first event relates to the performance monitoring results of a model associated with the terminal device. According to the method of claim 20 or 21, the model associated with the terminal device includes a first model, or includes a first model and one or more second models, wherein the first model is the model currently used by the terminal device, and the second model is a model supported by the terminal device and is different from the first model; The first event includes one or more of the following: The performance metrics of the first model are worse than the first threshold value; The performance metrics of the first model and the one or more second models are all worse than the second threshold value; At least one of the one or more second models has a better performance metric than the first model. At least one of the one or more second models has a better performance index than the first model, and the difference between the performance index of the at least one second model and the performance index of the first model is greater than a third threshold value. At least one of the one or more second models has a performance metric that is better than the fourth threshold value; The performance metric of the first model is worse than the fifth threshold, and the performance metric of at least one of the one or more second models is better than the sixth threshold; The absolute value of the difference between the performance metric of the first model and the performance metric of at least one of the one or more second models is less than the seventh threshold value; At least one of the one or more second models has a performance metric that is better than that of the reference model; At least one of the one or more second models has a performance metric that is better than that of the reference model, and the difference between the performance metric of the at least one second model and the performance metric of the reference model is greater than the eighth threshold. The performance metrics of two or more of the one or more second models are better than those of the first model; The performance metrics of two or more of the one or more second models are better than those of the first model, and the difference between the performance metrics of the two or more second models and the performance metrics of the first model is greater than the ninth threshold value. The performance metric of at least one of the one or more second models is no worse than that of the first model, and the complexity of the at least one second model is lower than that of the first model. According to the method of claim 22, one or more of the first threshold value to the ninth threshold value are determined based on any of the following: Protocol predefined; The first configuration information sent by the network device. The method according to any one of claims 21-23, wherein the first event is determined based on any one of the following: Protocol predefined; The second configuration information sent by the network device. The method according to any one of claims 20-24, wherein the first information is carried by Media Access Control (MAC) signaling, or the first information is carried by Physical Uplink Control Channel (PUCCH). The method according to any one of claims 20-25, wherein the uplink transmission resource is a Physical Uplink Data Channel (PUSCH) resource, or the uplink transmission resource is a Physical Uplink Control Channel (PUCCH) resource. The method according to any one of claims 20-26, wherein the first information is used to request uplink transmission resources, the method further comprising: The network device sends a second message, which is used to indicate uplink transmission resources. The method according to claim 27, wherein the second information is downlink control information (DCI), and the second information is used to indicate the uplink transmission resources from a predefined resource pool. The method according to any one of claims 20-26, further comprising: The first information is used to indicate uplink transmission resources, and the network device receives the model information on the uplink transmission resources indicated by the first information; Alternatively, the first information may be used to indicate whether the model information is sent on a pre-configured uplink transmission resource. When the first information indicates that the model information is sent on a pre-configured uplink transmission resource, the model information is received based on the pre-configured uplink transmission resource. The method according to any one of claims 20-26, wherein the first information is used to indicate uplink transmission resources and request uplink transmission resources, the method further comprising: The network device sends third information; the third information is used to indicate whether to use the uplink transmission resource indicated by the first information, and/or to indicate another uplink transmission resource. The method according to any one of claims 20-30, wherein the model information includes a first reference label and/or model performance monitoring results. The method according to claim 31, wherein, The first reference label includes a reference value for the output result of the third model, which is the model currently used by the network device and corresponds to the first model currently used by the terminal device; the first reference label is used for performance monitoring of the third model. According to the method of claim 31 or 32, the model associated with the terminal device includes a first model, or includes a first model and one or more second models, wherein the first model is the model currently used by the terminal device, and the second model is a model supported by the terminal device and is different from the first model; The model performance monitoring results include one or more of the following: Performance metrics of the first model; The performance index of the second model with the best performance index among the one or more second models; The performance metric of at least one of the one or more second models is better than the performance metric of the first model. The difference between the performance metrics of the first model and the performance metrics of the reference model; The difference between the performance index of at least one of the one or more second models and the performance index of the reference model, wherein the performance index of the at least one second model is better than the performance index of the reference model; The difference between the performance index of at least one of the one or more second models and the performance index of the reference model, wherein the performance index of the at least one second model is better than the performance index of the first reference model; The model parameters and/or model structure of the second model with the best performance index among the one or more second models; The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than that of the first model; The model identification information and/or dataset identification information of the second model with the best performance index among the one or more second models; The model identification information and/or dataset identification information of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the performance index of the first model; The model parameters and/or model structure of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the tenth threshold value; The model identification information and/or dataset identification information of at least one of the one or more second models, wherein the performance index of the at least one second model is better than the tenth threshold value. The method according to any one of claims 20-33, wherein the first information is used to request a downlink reference signal, the method further comprising: The network device sends the downlink reference signal; the downlink reference signal is used by the terminal device to perform performance monitoring and/or online training of the model associated with the terminal device. According to the method of claim 34, wherein the first information is further used to request uplink transmission resources, and/or to indicate uplink transmission resources, the method further includes: The network device receives performance monitoring results sent by the terminal device on the uplink transmission resources. The performance monitoring results are obtained by monitoring the performance of the model based on the downlink reference signal. According to the method of claim 35, the time-domain resources for transmitting the downlink reference signal are related to the uplink transmission. The resource time interval is a first duration; the first duration is determined based on the processing capability of the terminal device. The method according to any one of claims 21-36, wherein the model associated with the terminal device is a model configured by the network device for the terminal device; Alternatively, the model associated with the terminal device is the model reported by the terminal device to the network device; Alternatively, the model associated with the terminal device is a model pre-agreed upon by the terminal device and the network device. A model management device, applied to a terminal device, the device comprising: The first sending unit is configured to send first information to the network device in the event of a first event; the first event is related to the performance monitoring results of the model associated with the terminal device. The first information is used for one or more of the following: Requesting uplink transmission resources; Indicates uplink transmission resources; the uplink transmission resources are used to send model information; Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training. A model management device, applied to a network device, the device comprising: The second receiving unit is configured to receive first information sent by the terminal device; the first information is used for one or more of the following: Requesting uplink transmission resources; Indicates uplink transmission resources; the uplink transmission resources are used to send model information; Request a downlink reference signal; the downlink reference signal is used for model monitoring or online training. A terminal device, comprising: Memory is used to store executable instructions for a computer; A processor, connected to the memory, is configured to implement the method of any one of claims 1 to 19 by executing the computer-executable instructions. A network device, comprising: Memory is used to store executable instructions for a computer; A processor, connected to the memory, is configured to implement the method of any one of claims 20 to 37 by executing the computer-executable instructions. A chip, the chip comprising: A processor for retrieving and running a computer program from memory, causing a device having the chip mounted to perform the method as described in any one of claims 1 to 19, or to perform the method as described in any one of claims 20 to 37. A computer-readable storage medium storing a computer program that, when executed by at least one processor, implements the method as claimed in any one of claims 1 to 19, or implements the method as claimed in any one of claims 20 to 37. A computer program product comprising a computer storage medium storing a computer program, the computer program comprising instructions executable by at least one processor, wherein when the instructions are executed by the at least one processor, the method of any one of claims 1 to 19 is implemented, or the method of any one of claims 20 to 37 is implemented. A computer program that causes a computer to perform the method as described in any one of claims 1 to 19, or to implement the method as described in any one of claims 20 to 37.