Terminal management method and apparatus, communication device, readable storage medium, and program product

Abstract

This application relates to a terminal management method, apparatus, communication device, storage medium, and computer program product. The method includes: receiving one or more inference configuration information sent by a network, the inference configuration information being carried by a Channel State Information (CSI) reported configuration message; when the inference configuration information is inapplicable, maintaining the inference configuration, and sending a first message to the network, the first message containing at least one of the following: a first applicable function, a first inapplicable function, a first reason, and a preferred inference configuration for the terminal. This method can improve the inference adaptability of the terminal model, thereby improving the inference performance of the terminal model.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a terminal management method, apparatus, communication device, computer-readable storage medium, and computer program product. Background Technology

[0002] With the development of 6G technology, 6G networks will seamlessly integrate with AI (Artificial Intelligence) / ML (Machine Learning) technologies, developing and progressing together. Thus, with the upgrading of terminal hardware capabilities and advancements in AI / ML model compression technology, terminals will possess the intelligent computing capabilities to perform AI / ML inference, enabling the deployment of AI / ML functions on terminals and making AI terminals a reality.

[0003] In traditional technologies, inference configuration information is sent in advance by the network to configure the AI ​​terminal for inference. The AI ​​terminal is then instructed to perform inference functions in the corresponding scenarios by using the inference functions specified in the inference configuration information.

[0004] However, since the network cannot obtain the additional conditions and model availability on the terminal side in advance, the inference configuration sent by the network may be inapplicable in the AI ​​terminal, resulting in poor model inference performance of the AI ​​terminal. Summary of the Invention

[0005] Therefore, it is necessary to provide a terminal management method, device, computer equipment, computer-readable storage medium, and computer program product to address the aforementioned technical problems.

[0006] Firstly, this application provides a terminal management method, which is applied to a terminal and includes:

[0007] Receive one or more inference configuration information sent by the network, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message;

[0008] When the inference configuration information is not applicable, the inference configuration is maintained, and a first message is sent to the network. The first message contains at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

[0009] In one embodiment, prior to receiving one or more inference configurations from the receiving network, the method further includes:

[0010] Receive terminal capability query information sent by the network, the terminal capability query information being used to initiate the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network;

[0011] The terminal capability information is sent to the network, and the terminal capability information includes at least one of the terminal's supported functions and model information.

[0012] In one embodiment, the supported functions of the terminal are functions that the terminal can use, characterized by terminal capability information.

[0013] In one embodiment, the model information includes at least one of the following: a trained model, a downloaded model, and an updated model possessed by the terminal.

[0014] In one embodiment, the inference configuration information is set by the network based on the support functions and / or model information sent by the terminal.

[0015] In one embodiment, the inference configuration information includes at least one of the following:

[0016] Association identifier, model input information, model output information, beam type of model input and output, mapping relationship between model input and output, and CSI report type.

[0017] In one embodiment, the association identifier is used to ensure consistency of the terminal-side model during training and inference.

[0018] In one embodiment, the model input information includes at least one of the following: a first CSI resource configuration identifier, a set of beam indexes output by the model, a number of beams, a beam identifier, a beam order, a beam selection mode, a number of historical time instances, and a time interval between two consecutive historical time instances.

[0019] In one embodiment, the beam selection mode includes at least one of the following:

[0020] Fixed configuration mode, random selection mode, pre-configured continuous beam mode and discontinuous beam mode.

[0021] In one embodiment, the model output information includes at least one of the following:

[0022] The second CSI resource configuration identifier, the beam index set of model input, the number of beams, the beam identifier, the beam order, the number of future time instances, and the time interval between two consecutive future time instances.

[0023] In one embodiment, the beam types of the model input and output include at least one of narrowband and wideband.

[0024] In one embodiment, the mapping relationship between the model input and output includes at least one of the following:

[0025] The input beam is different from the output beam; the input beam is a subset of the output beam; and the input beam is the same as the output beam.

[0026] In one embodiment, when the mapping relationship between the model input and output is such that the input beam is a subset of the output beam, or the input beam is the same as the output beam, the inference configuration information contains only one association identifier.

[0027] In one embodiment, when the mapping relationship between the model input and output is such that the input beam and the output beam are different, the inference configuration information includes two association identifiers, one of which is configured for the model input and the other of which is configured for the model output.

[0028] In one embodiment, the CSI report type includes one of periodic, semi-continuous, and non-periodic.

[0029] In one embodiment, the first applicable function is characterized as a function that the terminal can use for reasoning, and the first applicable function is a subset of the supported functions;

[0030] The first unapplicable function is the complement of the first applicable function in the supported functions.

[0031] In one embodiment, the first applicable function is determined by the association identifier, terminal assistance information, and model availability.

[0032] In one embodiment, the first applicable function is indicated by a bitmap or index or a CSI-ReportConfigID identifier.

[0033] In one embodiment, the terminal auxiliary information includes at least one of terminal battery level, memory usage, CPU utilization, GPU utilization, and network transmission latency.

[0034] In one embodiment, model availability includes having at least one of a trained model, an updated model, and a downloaded model.

[0035] In one embodiment, the first cause includes at least one of the following:

[0036] It does not meet the requirements for terminal auxiliary information, network auxiliary information, and model availability.

[0037] In one embodiment, the first message is carried by RRC reconfiguration completion and / or terminal assistance information signaling.

[0038] In one embodiment, after sending the first message to the network, the method further includes:

[0039] Do not activate periodic CSI reports.

[0040] In one embodiment, after sending the first message to the network, the method further includes:

[0041] Start the timer.

[0042] In one embodiment, after receiving a second message retransmitted by the network, the second message containing at least one of releasing inapplicable inference configuration information and updated inference configuration information, the method further includes:

[0043] Send a third message to the network, the third message including at least one of a second applicable function, a second inapplicable function, and a second reason.

[0044] In one embodiment, the method further includes:

[0045] If the timer expires, perform at least one of the following communication processes: actively release inapplicable inference configurations, or send a message to the network.

[0046] In one embodiment, the timer stops when at least one of the following conditions is received: receiving the second message indicating that the network releases an inapplicable inference configuration and the inapplicable inference configuration becomes applicable.

[0047] In one embodiment, after sending the first message to the network, the method further includes:

[0048] If the terminal enters at least one of the following states: Radio Resource Control (RRC) inactive state (RRC__INATIVE), Radio Resource Control (RRC_IDLE) idle state (RRC_IDLE), and handover state, release the inference configuration of applicable and / or inapplicable functions.

[0049] Secondly, this application provides a terminal management method, which is applied to a network, and the method includes:

[0050] Send one or more inference configuration information to the terminal, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message;

[0051] The terminal receives a first message, the first message containing at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

[0052] In one embodiment, the method further includes:

[0053] Send terminal capability query information to the terminal, the terminal capability query information being used to initiate the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network;

[0054] The terminal receives terminal capability information sent by the terminal, the terminal capability information including at least one of the terminal's supported functions and model information.

[0055] In one embodiment, after receiving the first message sent by the terminal, the method further includes:

[0056] A second message is sent to the terminal, the second message including at least one of releasing an inapplicable inference configuration and an updated inference configuration.

[0057] Thirdly, this application also provides a terminal management device, which is applied to a terminal and includes:

[0058] The first receiving module is used to receive one or more inference configuration information sent by the network, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message;

[0059] The first sending module is configured to send a first message to the network when the inference configuration information is not applicable, wherein the first message includes at least one of the terminal's first applicable function, first inapplicable function, and first reason.

[0060] Fourthly, this application also provides a terminal management device, which is applied to a network, and the device includes:

[0061] The first sending module is used to send one or more inference configuration information to the terminal, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message.

[0062] A first receiving module is configured to receive a first message sent by the terminal, the first message including at least one of the terminal's first applicable function, first inapplicable function, and first reason.

[0063] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0064] Receive one or more inference configuration information sent by the network, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message;

[0065] When the inference configuration information is not applicable, the inference configuration is maintained, and a first message is sent to the network. The first message contains at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

[0066] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0067] Receive one or more inference configuration information sent by the network, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message;

[0068] When the inference configuration information is not applicable, the inference configuration is maintained, and a first message is sent to the network. The first message contains at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

[0069] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0070] Receive one or more inference configuration information sent by the network, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message;

[0071] When the inference configuration information is not applicable, the inference configuration is maintained, and a first message is sent to the network. The first message contains at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

[0072] The aforementioned terminal management method, apparatus, computer equipment, computer-readable storage medium, and computer program product receive one or more inference configuration information sent by the network. This inference configuration information is carried by a Channel State Information (CSI) reported configuration message. When the inference configuration information is inapplicable, the inference configuration is maintained, and a first message is sent to the network. The first message includes at least one of the following: a first applicable function, a first inapplicable function, a first reason, and a preferred inference configuration for the terminal. Using this method, the terminal reports a first message to the network, clarifying the applicable or inapplicable functions of the terminal. This allows the network to adjust the inference configuration information of the terminal model, improving the inference adaptability of the terminal model and, consequently, enhancing the inference performance of the terminal model. Attached Figure Description

[0073] Figure 1 This is an application environment diagram of the terminal management method in one embodiment;

[0074] Figure 2 This is a flowchart illustrating a terminal management method on the terminal side in one embodiment;

[0075] Figure 3 This is a flowchart illustrating the steps of terminal-network interaction to obtain terminal capability information in one embodiment.

[0076] Figure 4 This is a flowchart illustrating the step of pausing periodic CSI reporting after sending the first message in one embodiment.

[0077] Figure 5 This is a flowchart illustrating the step of starting a timer in one embodiment;

[0078] Figure 6 This is a flowchart illustrating the step of sending a third message to the network in one embodiment;

[0079] Figure 7 This is a flowchart illustrating the communication processing steps of a terminal for inapplicable inference configuration information in one embodiment.

[0080] Figure 8 This is a flowchart illustrating a terminal management method on the network side in one embodiment;

[0081] Figure 9 This is a flowchart illustrating the steps of network and terminal interaction to obtain terminal capability information in one embodiment.

[0082] Figure 10 This is a flowchart illustrating the step of sending a second message to a terminal in one embodiment;

[0083] Figure 11 This is a flowchart illustrating an example of terminal management in one embodiment;

[0084] Figure 12 This is a flowchart illustrating another example of terminal management in one embodiment;

[0085] Figure 13 This is a structural block diagram of a terminal management device on the terminal side in one embodiment;

[0086] Figure 14 This is a structural block diagram of a terminal management device on the network side in one embodiment;

[0087] Figure 15 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0088] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0089] Figure 1 This is a schematic diagram illustrating an application scenario of a terminal management method provided in an embodiment of this application. For example... Figure 1As shown, this scenario includes a terminal 100 and a network 200. Data transmission can occur between the terminal 100 and the network 200.

[0090] Terminal 100 can be a wireless terminal, which can be a device providing voice and / or other service data connectivity to a user, a handheld device with wireless connectivity, or other processing devices connected to a wireless modem. The wireless terminal can communicate with one or more core networks via a Radio Access Network (RAN). The wireless terminal can be a mobile terminal, such as a mobile phone (or "cellular" phone) or a computer with a mobile terminal, for example, a portable, pocket-sized, handheld, computer-embedded, or vehicle-mounted mobile device, exchanging voice and / or data with the RAN. The wireless terminal can also be referred to as a system, subscriber unit, subscriber station, mobile station, mobile, remote station, remote terminal, access terminal, user terminal, user agent, or user device or user equipment, without limitation herein.

[0091] In traditional technologies, 6G networks seamlessly integrate with AI (Artificial Intelligence) / ML (Machine Learning) technologies, developing and progressing together. Thus, with upgrades in terminal hardware capabilities and advancements in AI / ML model compression technology, terminals possess the intelligent computing capabilities to perform AI / ML inference, enabling the deployment of AI / ML functions on terminals and making AI terminals a reality. The network pre-sends inference configuration information to configure the AI ​​terminal for inference, and through the inference functions specified in the configuration information, instructs the AI ​​terminal to perform inference functions in the corresponding scenarios.

[0092] However, in traditional technologies, since the network cannot obtain the additional conditions and model availability on the terminal side in advance, the inference configuration sent by the network may be inapplicable to the AI ​​terminal, resulting in poor model inference performance of the AI ​​terminal.

[0093] Based on the aforementioned traditional technologies, this application provides a terminal management method. By having the terminal report a first message to the network, the applicable or inapplicable functions of the terminal are clarified. This allows the network to adjust the inference configuration information of the terminal model, thereby improving the inference adaptability of the terminal model and ultimately enhancing the inference performance of the terminal model.

[0094] It should be noted that the beneficial effects or technical problems solved by the embodiments of this application are not limited to this one, but may also be other implicit or related problems. For details, please refer to the description of the embodiments below.

[0095] Before introducing specific embodiments of the present invention, the technical terms involved in the present invention will be explained:

[0096] AI (Artificial Intelligence): AI refers to computer systems that simulate human brain activity, enabling them to perform activities that require human intelligence, such as perception, understanding, reasoning, and decision-making.

[0097] ML (Machine Learning): ML is a branch of AI that focuses on automatically learning data features through data and algorithms, thereby identifying patterns and rules, and ultimately achieving the goals of prediction and classification;

[0098] RRC (Radio Resource Control): The RRC layer is primarily responsible for signaling interaction between the radio access network and terminals in the 5G NR protocol layer. RRC signaling plays a crucial role in the connection establishment, maintenance, and release process of the entire radio access network. Specifically, 5G NR supports three states for RRC: RRC_IDLE, RRC_INACTIVE, and RRC_CONNECTED. RRC_INACTIVE is a state between RRC_IDLE and RRC_CONNECTED, allowing the terminal to release the radio bearer while retaining the signaling connection between the terminal and the core network and the user data channel.

[0099] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0100] In one embodiment, such as Figure 2 As shown, a terminal management method is provided, which is applied to Figure 1 Taking terminal 100 as an example, the explanation includes the following steps:

[0101] Step 202: Receive one or more inference configuration information sent by the network.

[0102] The inference configuration information is carried in the channel state information (CSI) reporting configuration message.

[0103] Specifically, in scenarios where the terminal and network work together, the terminal will listen to and receive inference configuration information sent from the network side in real time through a preset network communication protocol. This inference configuration information can be a single instruction or a set of configurations containing multiple parameters. Optionally, the inference configuration information includes at least one of the following: association identifiers, various key parameters required for model operation, such as model input information, model output information, beam types of model input and output, and mapping relationship between model input and output.

[0104] In this way, through this inference configuration information, the network side performs inference configuration on the terminal side model to instruct the terminal side model to perform corresponding functions.

[0105] Step 204: When the inference configuration information is not applicable, maintain the inference configuration and send the first message to the network.

[0106] The first message includes at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

[0107] In implementation, after receiving inference configuration information, if the inference configuration information is not applicable, the terminal will maintain the current inference configuration and simultaneously send a first message to the network. This first message is used to provide feedback to the network on the terminal's understanding of the applicability of the current inference configuration information. In this way, by the terminal proactively reporting the applicability of the inference configuration to the network, the network can adaptively adjust the terminal's inference configuration information based on this information.

[0108] Specifically, when the inference configuration information received by the terminal does not match its hardware performance, software environment, or actual business scenario, resulting in the inability to function properly, the terminal will maintain its current valid inference configuration to ensure the continuity of model inference services. Simultaneously, the terminal will immediately send a first message containing key information to the network side. Specifically, the first message includes the terminal's currently supported first applicable function (e.g., lightweight model inference tasks suitable for local computing power); a list of the first unapplicable functions that cannot be implemented due to configuration conflicts or insufficient capabilities (e.g., real-time processing of highly complex models); a clear indication of the first reason for the inapplicable inference configuration, which may involve insufficient resources or protocol version incompatibility; and suggestions for preferred inference configuration information, such as recommended model parameters and data processing formats, so that the network side can quickly locate the problem, optimize configuration strategies, and achieve more efficient inference service adaptation.

[0109] In the aforementioned terminal management method, one or more inference configuration information messages are received from the network. These inference configuration information is carried by a Channel State Information (CSI) reported configuration message. When the inference configuration information is inapplicable, the inference configuration is maintained, and a first message is sent to the network. The first message includes at least one of the following: a first applicable function, a first inapplicable function, a first reason, and a preferred inference configuration for the terminal. Using this method, the terminal reports a first message to the network, clarifying the applicable or inapplicable functions of the terminal. This allows the network to adjust the inference configuration information of the terminal model, improving the inference adaptability of the terminal model and consequently enhancing its inference performance.

[0110] In one embodiment, such as Figure 3 As shown, prior to step 202, the method further includes:

[0111] Step 302: Receive terminal capability query information sent by the network.

[0112] Among them, the terminal capability query information is used to initiate the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network.

[0113] In practice, before the network sends one or more inference configuration information to the terminal, the network first sends a UE Capability Enquiry to the terminal. This UE Capability Enquiry is used to activate the terminal device and enable the terminal to report supported AI / ML functions to the network.

[0114] Step 304: Send terminal capability information to the network.

[0115] In practice, in response to the terminal capability query information, the terminal sends UECapability Information to the network.

[0116] Specifically, upon receiving the terminal capability query information from the network, the terminal immediately initiates a data integration process to comprehensively review its AI / ML functions and model resources. Then, the terminal encapsulates and packages the compiled terminal capability information, generating terminal capability information that is transmitted to the network via a stable and reliable communication link. This terminal capability information includes at least one of the terminal's supported functions and model information.

[0117] Optionally, the supported functions of the terminal are those functions that the terminal can use, represented by the terminal capability information. For example, supported functions include specific AI / ML functions such as speech recognition and image classification supported by the terminal.

[0118] Optionally, the model information includes at least one of the following: a trained model, a downloaded model, and an updated model. For example, the model information may include key model information such as the model type (e.g., convolutional neural network, recurrent neural network), parameter size, and training dataset, or both, to provide the network with an accurate profile of the terminal's capabilities, facilitating subsequent adaptive task allocation and resource scheduling.

[0119] In this embodiment, the terminal receives terminal capability query information sent by the network and accordingly initiates the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network. This establishes a capability interaction mechanism between the network and the terminal. Based on the terminal capability information, the network can reasonably allocate suitable tasks and resources to the terminal, avoiding inefficiency or resource waste caused by resource mismatch. At the same time, the terminal actively reports its own capabilities, which helps the network to have a more comprehensive understanding of the terminal status, plan and optimize business deployment in advance, and improve the adaptability and stability of AI / ML functions executed on the terminal side.

[0120] In one embodiment, the inference configuration information is set by the network based on the supported functions and / or model information sent by the terminal.

[0121] In one embodiment, the inference configuration information includes at least one of the following:

[0122] Association identifier, model input information, model output information, beam type of model input and output, mapping relationship between model input and output, and CSI report type.

[0123] Specifically, the setting of inference configuration information is a dynamic process based on the interaction between the terminal and the network. The network conducts a comprehensive and targeted analysis based on the supported functions and / or model information sent by the terminal. The supported functions fed back by the terminal clearly demonstrate its capabilities and strengths in areas such as artificial intelligence and data processing; while the model information covers key details such as the type, version, and parameters of the terminal's built-in models. Thus, based on the inference configuration information, the network comprehensively considers and optimizes the design in conjunction with factors such as business needs and resource availability, ultimately formulating inference configuration information adapted to the terminal. This inference configuration information not only fully leverages the performance advantages of the terminal but also ensures that the model on the terminal side efficiently and accurately achieves the expected functions when performing inference tasks, realizing reasonable resource matching and collaborative work between the network and the terminal.

[0124] Optionally, the association identifier in the inference configuration information is used to ensure the consistency of the terminal-side model during training and inference.

[0125] Specifically, the association identifier in the inference configuration information is a key element in ensuring the data and logic consistency of the terminal-side model during the training and inference phases. By generating a unique and specific identifier (i.e., the association identifier), it is embedded in the model's key data and parameters during training, recording core information, algorithm structure, data sources, and other characteristics of the model during training. When the model enters the inference phase, the association identifier plays a role again, verifying whether there have been any parameter tampering, structural changes, or data deviations during training and inference by comparing the model's state during inference with the feature information recorded during training. Only when the association identifiers in the training and inference phases match consistently can the model's behavior and output results meet expectations, ensuring consistency throughout the entire process from training to inference. This improves the stability and reliability of model operation, providing a solid guarantee for the terminal to perform accurate inference tasks based on the model.

[0126] Optionally, the model input information includes at least one of the following: a first CSI resource configuration identifier, a set of beam indexes output by the model, the number of beams, beam identifiers, beam order, beam selection mode, the number of historical time instances, and the time interval between two consecutive historical time instances.

[0127] Specifically, the model input information is the key basis for the terminal-side model to perform inference operations, encompassing a wide variety of parameters. Among them, the first CSI resource configuration identifier represents a specific channel state information resource configuration, providing the model with basic communication environment parameters; the beam index set, number of beams, beam identifier, beam order, and beam selection mode output by the model are directly related to the core elements of beam management in wireless communication, helping the model accurately grasp the distribution, characteristics, and selection strategies of beams to optimize signal transmission; and the number of historical time instances and the time interval between two consecutive historical time instances endow the model with the ability to perceive data change trends from a time dimension, enabling it to perform inference tasks more accurately based on historical data patterns and current model input information, providing strong support for the efficient operation of wireless communication.

[0128] In one optional embodiment, the beam selection mode includes at least one of the following:

[0129] Fixed configuration mode, random selection mode, pre-configured continuous beam mode and discontinuous beam mode.

[0130] Specifically, among all the information included in the model input, beam selection mode is a key parameter with several different settings. Fixed configuration mode refers to beam selection rules executed according to pre-set, fixed parameters that are not easily changed, selecting beams with the same logic each time. This fixed configuration mode is suitable for scenarios with stable environments and fixed requirements. Random selection mode introduces randomness in beam selection, which can, to some extent, avoid the limitations caused by fixed modes. Pre-configured continuous beam mode selects adjacent continuous beams sequentially according to a pre-set order, ensuring the continuity and orderliness of beam coverage. Discontinuous beam mode breaks the continuity constraint, selecting beams discontinuously according to specific rules or requirements, offering greater advantages in flexibly adapting to complex and changing environmental needs.

[0131] In one embodiment, the model output information includes at least one of the following:

[0132] The second CSI resource configuration identifier, the beam index set of model input, the number of beams, the beam identifier, the beam order, the number of future time instances, and the time interval between two consecutive future time instances.

[0133] Specifically, in wireless communication systems, the second CSI (Channel State Information) resource configuration identifier is used to specify specific channel state information acquisition and feedback resources, providing a basic resource framework for subsequent model inputs. The beam index set for model input contains a series of numbered beam identifiers, each corresponding to a different signal transmission direction and characteristic, allowing for precise location of the required beam resources. The number of beams clarifies the total number of beams available or requiring processing in a specific scenario, providing crucial information for system resource allocation and scheduling. Beam identifiers assign unique labels to each beam, facilitating system differentiation and management of different beams. The beam order specifies the sequence of beams in transmission and processing, directly impacting signal processing flow and communication performance. The number of future time instances refers to the number of time segments divided when predicting or processing communication conditions within a specific future time period. The time interval between two consecutive future time instances determines the fineness of the time granularity. Together, these elements help the system more accurately plan and optimize future communication resources and strategies, ensuring efficient and stable operation of the communication system.

[0134] In one embodiment, the beam types of the model input and output include at least one of narrowband and wideband.

[0135] Specifically, narrowband beams possess strong directivity and concentrated signal characteristics, enabling efficient signal transmission and reception in specific directions. During model input, narrowband beams can precisely focus on specific areas or targets, enhancing signal anti-interference capabilities and providing AI models with more accurate and stable channel state information. For example, in indoor communication scenarios, inputting the model with a narrowband beam can better capture signal characteristics at specific locations, optimizing the model's adaptability to the local environment. During model output, narrowband beams accurately transmit the processed signal to the target terminal, effectively reducing signal scattering and interference, and improving communication reliability and quality. Wideband beams, on the other hand, have a wider spectral range, capable of carrying more data. During model input, wideband beams can comprehensively reflect channel changes across the entire frequency band, providing the model with rich, multi-dimensional information, helping it to more accurately understand complex communication environments. For instance, in high-speed mobile scenarios, wideband beam input allows the model to promptly perceive rapidly changing channel characteristics, thereby quickly adjusting communication strategies. During the model output phase, broadband beams can transmit data at higher rates, meeting the needs of high-capacity, high-speed communication, such as video streaming and large data file transfer.

[0136] In one embodiment, the mapping relationship between model inputs and outputs includes at least one of the following:

[0137] The input beam is different from the output beam; the input beam is a subset of the output beam; and the input beam is the same as the output beam.

[0138] Specifically, when a model processes beam-related data, the mapping relationship between its input and output can take several forms. First, the input and output beams may exhibit completely different states, meaning the model has thoroughly transformed the input beam, generating an output beam with entirely new characteristics. Second, the input beam can be a subset of the output beam, indicating that the AI ​​model, during processing, further expands and derives more beam information based on the input beam, making the output beam encompass the input beam and contain additional content. Finally, the input and output beams may be identical, meaning the model has not changed the essential characteristics and composition of the beam after processing, preserving the original state of the input beam and providing a stable mapping guarantee for beam transmission or processing in specific scenarios.

[0139] In one embodiment, when the mapping relationship between model input and output is such that the input beam is a subset of the output beam, or the input beam is the same as the output beam, the inference configuration information contains only one association identifier.

[0140] Specifically, in the beamforming process of AI models, the mapping relationship between model inputs and outputs plays a decisive role in constructing inference configuration information. The design of inference configuration information exhibits concise and efficient characteristics under two specific mapping scenarios: first, when the input beam is a subset of the output beam; and second, when the input and output beams are identical. In these cases, since the model's input and output beams are either identical or partially identical, the inference configuration information does not require complex structures and multi-dimensional identifiers; it only needs to incorporate a single association identifier. This association identifier accurately matches the correspondence between input and output beams, avoiding the storage and transmission burden caused by redundant information and helping the model quickly locate and retrieve beam data during inference, significantly improving inference efficiency.

[0141] In one embodiment, when the mapping relationship between the model input and output is such that the input beam and the output beam are different, the inference configuration information includes two association identifiers: one association identifier is configured for the model input, and the other association identifier is configured for the model output.

[0142] In one embodiment, the CSI report type includes one of periodic, semi-continuous, and non-periodic.

[0143] In one embodiment, the first applicable function is characterized as a function that the terminal can use for reasoning, and the first applicable function is a subset of the supported functions; the first inapplicable function is the complement of the first applicable function to the supported functions.

[0144] In one embodiment, the first applicable function is determined by the association identifier, terminal assistance information, and model availability.

[0145] In one embodiment, terminal auxiliary information includes at least one of terminal battery level, memory usage, CPU utilization, GPU utilization, and network transmission latency.

[0146] Among these metrics, battery power is a crucial indicator of a device's battery life, directly impacting its usage time and stability. For example, low battery power may trigger the device's power-saving mode, affecting the performance of certain functions. Memory, as a key resource for storing and processing data, reflects the amount of space available for task processing. Excessive memory usage can slow down or even cause the device to lag. CPU utilization reflects the workload of the device's central processing unit. A higher CPU utilization indicates that the device is handling complex tasks, potentially consuming more power and system resources. GPU utilization primarily reflects the usage of the graphics processing unit. For devices requiring high-load graphics tasks such as graphics rendering and video processing, GPU utilization is an important indicator of their graphics processing capabilities and performance. Network transmission latency represents the time it takes for data to be transmitted between the device and the network. Lower latency ensures fast and stable data transmission, while higher latency can lead to untimely data transmission, affecting user experience. For instance, in real-time communication scenarios, higher latency can cause audio or video stuttering.

[0147] In one embodiment, model availability includes having at least one of a trained model, an updated model, and a downloaded model.

[0148] In one embodiment, the first cause includes at least one of the following:

[0149] It does not meet the requirements for terminal auxiliary information, network auxiliary information, and model availability.

[0150] Specifically, when one or more inference configuration information sent by the network is inapplicable, the terminal sends a first message to the network to clarify that the terminal is not compatible with the current inference configuration information. This first message includes a first reason for the incompatibility. This first reason could be twofold: First, the inference configuration information may not meet the terminal's auxiliary information. For example, the terminal's battery level may be low, while the computational intensity set in the inference configuration information is too high, leading to rapid battery depletion or even terminal malfunction; or the terminal's memory, CPU, or GPU utilization may be saturated, making it difficult to support the complex computational tasks specified in the inference configuration information. Second, the inference configuration information may not meet the network's auxiliary information. For example, the network transmission latency may be high, but the inference configuration information requires high real-time data interaction, thus failing to guarantee the stability and timeliness of data transmission. Third, the inference configuration information may not meet the model's availability requirements. For example, some key components of the model may be missing, or the versions may be incompatible, causing the functions in the inference configuration information to malfunction. These reasons can appear individually or in combination. The terminal accurately feeds back this key information to the network through the first message, providing a comprehensive and accurate basis for the network to optimize and configure the inference information in the future, so as to achieve a more efficient and adaptable AI model inference process.

[0151] In one embodiment, the first message is carried by RRC reconfiguration completion and / or terminal assistance information signaling.

[0152] In one embodiment, such as Figure 4 As shown, after step 204, the method further includes:

[0153] Step 401: Do not activate periodic CSI reports.

[0154] In implementation, when a terminal detects that the inference configuration information sent by the network is inapplicable, it sends a first message to the network containing the reason for the incompatibility. Furthermore, to optimize resource usage and avoid unnecessary data transmission, the terminal disables periodic CSI (Channel State Information) reports. This reduces its own data processing pressure and power consumption, while preventing the network from continuously sending incompatible inference configuration information to the terminal based on CSI reports. This improves the overall system's operating efficiency and the rationality of resource utilization.

[0155] In one embodiment, such as Figure 5 As shown, after step 204, the method further includes:

[0156] Step 501: Start the timer.

[0157] In practice, after sending the first message to the network, the terminal starts a timer to monitor the time the terminal waits for network feedback, thereby releasing inapplicable inference configuration information within the effective time.

[0158] Optionally, if a second message is received from the network before the timer expires, the terminal releases the inapplicable inference configuration. The second message includes at least one of releasing inapplicable inference configuration information and updated inference configuration information. Specifically, if the terminal successfully receives a second message from the network before the timer expires, and the content of the second message covers at least one of releasing inapplicable inference configuration information and updated inference configuration information, the terminal will respond quickly, stop using the currently incompatible inference configuration, and further update or wait for updated new inference configuration information.

[0159] In one embodiment, after step 204, the method further includes:

[0160] If the terminal enters at least one of the RRC__INATIVE state, RRC_IDLE state, and handover state, release the inference configuration of applicable and / or inapplicable functions.

[0161] In implementation, when a terminal enters at least one of the following states: RRC_INACTIVE, RRC_IDLE, and handover state, the terminal will perform fine-grained management of relevant functions based on preset policies and business requirements, releasing some inference configurations for used functions and / or inapplicable functions.

[0162] Specifically, in the RRC_INACTIVE state, to reduce power consumption and save network resources, the terminal retains only partial connection information with the network. At this time, some non-critical real-time service functions, such as high-frequency data synchronization functions, will be released, while basic paging and monitoring functions remain available. When the terminal is in the RRC_IDLE state, due to its low connection with the network, most user-specific service functions will be released, except for necessary terminal message reception and cell selection / reselection functions, to minimize resource consumption. During handover, to ensure the stability and reliability of the handover process, functions unrelated to handover, such as ongoing background data transmission functions, will be suspended or released. At the same time, functions optimized for the handover process, such as measurement and control functions, will be activated to assist the terminal in quickly completing the network handover. After the handover is completed, applicable functions will be reactivated and inapplicable functions will be deactivated according to actual service needs.

[0163] In one embodiment, such as Figure 6As shown, after receiving a second message retransmitted by the network, the second message containing at least one of releasing inapplicable inference configuration information and updated inference configuration information, the method further includes:

[0164] Step 601: Send a third message to the network.

[0165] The third message includes at least one of the following: a second applicable function, a second inapplicable function, and a second reason.

[0166] In implementation, before the timer expires, the terminal receives a second message from the network. This second message may contain at least one of releasing inapplicable inference configuration information and updating inference configuration information. This allows the terminal to reconfigure its local AI model based on the new inference configuration information from the network. However, if the new configuration inference information from the network is still inapplicable, the terminal can still send a third message to the network, reporting that the inference configuration is still unreasonable.

[0167] In one embodiment, such as Figure 7 As shown, the method also includes:

[0168] Step 701: If the timer expires, implement at least one of the following communication processes:

[0169] Actively release inapplicable inference configurations and send messages to the network.

[0170] In implementation, during the terminal-network inference configuration interaction process, once the terminal starts its timer, it enters a phase of waiting for network feedback. If, as time progresses, the timer reaches its preset duration without any network response (i.e., the timer expires), the terminal can proactively release inapplicable inference configurations, promptly ceasing the use of inference configuration schemes that are unsuitable for the current operating environment or business requirements. This avoids resource waste and system performance degradation caused by the continued operation of invalid configurations, freeing up resource space for subsequent reconfiguration or receiving new inference configurations. On the other hand, the terminal can also send messages containing relevant information (e.g., first message, third message) to the network, detailing the inapplicability of the current inference configuration and the timer's expiration. This allows the network to understand the terminal's operating status in a timely manner, enabling it to re-plan and adjust its inference configuration strategy based on terminal feedback, thus promoting the configuration optimization process between the terminal and the network.

[0171] In one embodiment, the timer stops when at least one of the following conditions is received: the network releases a second message indicating that the inapplicable inference configuration has been changed to an applicable configuration.

[0172] In one embodiment, such as Figure 8As shown, a terminal management method is provided, which is applied to a network. The method includes:

[0173] Step 801: Send one or more inference configuration information to the terminal.

[0174] The inference configuration information is carried in the channel state information (CSI) reporting configuration message.

[0175] In implementation, during the management of terminals, the network can send one or more inference configuration messages to the terminals to configure the terminal's AI model for inference, instructing the model on the terminal side to perform corresponding functions. These inference configuration messages can be a single instruction or a set of configurations containing multiple parameters. Optionally, the inference configuration messages may include at least one of the following: association identifiers, various key parameters required for model operation, such as model input information, model output information, beam types for model input and output, and the mapping relationship between model input and output.

[0176] Step 802: Receive the first message sent by the terminal.

[0177] The first message includes at least one of the following: a first applicable function, a first inapplicable function, a first reason, and a preferred inference configuration for the terminal. The first message is used to update and adjust the terminal's inference configuration information.

[0178] In implementation, after the network sends inference configuration information to the terminal, the terminal provides feedback on the applicability of this configuration information. This results in the network receiving the first message from the terminal. This first message provides feedback to the network regarding the applicability of the current inference configuration information to the terminal. If the inference configuration information is not applicable to the terminal, the network will quickly initiate an optimization process based on the detailed feedback from the terminal, updating and adjusting the original inference configuration accordingly. This ensures that the new inference configuration accurately matches the actual needs of the terminal, thereby improving the efficiency and accuracy of AI model inference and achieving efficient collaboration and deep integration between the network and the terminal in AI inference tasks.

[0179] In one embodiment, such as Figure 9 As shown, the method also includes:

[0180] Step 901: Send terminal capability query information to the terminal.

[0181] Among them, the terminal capability query information is used to initiate the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network.

[0182] In implementation, to achieve efficient collaboration and accurate resource allocation between the network and terminals, the network proactively sends a UE Capability Enquiry to the terminal. This UE Capability Enquiry triggers a comprehensive review and feedback process of the terminal's own AI / ML capabilities. Specifically, upon receiving the UE Capability Enquiry, the terminal quickly initiates its internal capability detection mechanism to systematically scan its supported AI / ML functions, including but not limited to key information such as model training capabilities, inference algorithm types, and data processing scale. Subsequently, the terminal integrates and encapsulates these supported AI / ML functions and provides feedback to the network in a standardized report format according to established communication protocols and formats. Through this process, the network clearly understands the boundaries of the terminal's AI / ML capabilities, providing a data foundation for subsequently developing appropriate inference configuration strategies and optimizing resource allocation schemes, ensuring that network resources match the terminal's needs.

[0183] Step 902: Receive terminal capability information sent by the terminal.

[0184] The terminal capability information includes at least one of the terminal's supported functions and model information.

[0185] In implementation, after sending terminal capability query information, the network enters a phase of waiting for the terminal's response. Once the terminal has completed the sorting and integration of its own AI / ML capabilities, it will transmit terminal capability information, including core capabilities, back to the network according to the established communication protocol. The terminal capability information received by the network is rich and crucial, covering at least one of the terminal's supported functions and model information. If it includes supported functions, the network can clearly understand the specific capabilities the terminal possesses in the AI / ML field, such as what artificial intelligence tasks the terminal supports, such as image recognition and speech processing, or which machine learning algorithms it has mastered for operation and optimization. If it involves model information, the network can obtain details such as the type, version, and parameter scale of the AI / ML model adapted to the terminal, and know the complexity of model training and inference that the terminal can handle. This helps the network comprehensively build a terminal capability profile, providing accurate basis for subsequent customized inference configurations, reasonable allocation of network resources, and promotion of efficient collaboration between the network and the terminal in AI / ML tasks.

[0186] In this embodiment, the network proactively sends terminal capability query information, triggering the terminal to respond with its supported AI / ML functions. This effectively avoids the passive lag of terminal capability information and ensures that the network can promptly grasp the terminal's dynamics. This allows the network to gain a comprehensive and detailed understanding of the terminal's capability boundaries and technical characteristics in the AI / ML field, providing a solid data foundation for the network to subsequently customize and adapt inference configuration strategies and optimize resource allocation.

[0187] In one embodiment, such as Figure 10 As shown, after step 802, the method further includes:

[0188] Step 1001: Send the second message to the terminal.

[0189] The second message includes at least one of releasing an inapplicable inference configuration and updating the inference configuration.

[0190] In implementation, after receiving the first message from the terminal, the network side determines that the current terminal's model is not applicable to the initially sent inference configuration information. Therefore, the network sends a second message to the terminal containing at least one of the following: releasing the inapplicable inference configuration and updating the inference configuration. This effectively improves the collaborative efficiency between the network and the terminal, as well as the inference performance of the AI ​​model. On the one hand, the network side instructs the terminal to release the inapplicable inference configuration in a timely manner, avoiding problems such as wasted computing resources, increased energy consumption, and reduced processing efficiency caused by the terminal continuously using invalid configurations. On the other hand, the updated inference configuration issued by the network side is customized according to the actual needs and operating status of the terminal, accurately adapting to the terminal's hardware capabilities, network conditions, and model availability, ensuring that the AI ​​model inference runs under more optimized parameters and rules. The combination of these two aspects makes the network's management of the terminal's inference configuration more flexible and intelligent.

[0191] In one exemplary embodiment, such as Figure 11 As shown, a terminal management method is provided, and an example flow of the terminal management method includes:

[0192] Step 1101: The network sends UE Capability Query Information to the terminal. This UE Capability Query Information is used to instruct the terminal to report the supported AI / ML functions to the network.

[0193] Step 1102: The terminal sends UE Capability Information to the network, which includes at least one of supported functions and model information.

[0194] Step 1103: The network sends one or more inference configuration information to the terminal. The inference configuration information is carried by a CSI report (CSI-ReportConfig). The CSI report includes at least one of the following: association identifier, model input information, model output information, beam type of model input and output, and mapping relationship between model input and output.

[0195] Step 1104: The inference configuration received by the terminal is not applicable; retain the inference configuration.

[0196] Step 1105: The terminal sends a first message, which includes at least one of a first applicable function, a first inapplicable function, and a first reason.

[0197] Step 1106: Start the timer on the terminal.

[0198] Step 1107: The network sends a second message to the terminal, the second message including at least one of releasing an inapplicable inference configuration and updating the inference configuration.

[0199] Step 1108: The terminal receives a second message before the timer expires, releasing the inapplicable inference configuration.

[0200] Step 1109: The terminal sends a third message to the network, the third message including at least one of a second applicable function, a second inapplicable function, and a second reason.

[0201] In one exemplary embodiment, such as Figure 12 As shown, a terminal management method is provided, and an example flow of the terminal management method includes:

[0202] Step 1201: The network sends UE Capability Query Information to the terminal. This UE Capability Query Information is used to instruct the terminal to report the supported AI / ML functions to the network.

[0203] Step 1202: The terminal sends UE Capability Information to the network, which includes at least one of supported functions and model information.

[0204] Step 1203: The network sends one or more inference configuration information to the terminal. The inference configuration information is carried by a CSI report (CSI-ReportConfig). The CSI report includes at least one of the following: association identifier, model input information, model output information, beam type of model input and output, and mapping relationship between model input and output.

[0205] Step 1204: The inference configuration received by the terminal is not applicable; retain the inference configuration.

[0206] Step 1205: The terminal sends a first message, which includes at least one of a first applicable function, a first inapplicable function, and a first reason.

[0207] Step 1206: Start the timer on the terminal.

[0208] Step 1207: The network sends a second message to the terminal, the second message including at least one of releasing an inapplicable inference configuration and updating the inference configuration.

[0209] Step 1208: The timer expires, and the terminal actively releases inapplicable inference configuration information.

[0210] It should be understood that, although Figures 2-12 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figures 2-12 At least some of the steps in the process may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but may be executed at different times. The execution order of these steps or stages is not necessarily sequential, but may be executed in turn or alternately with other steps or at least some of the steps or stages in other steps.

[0211] In one embodiment, such as Figure 13 As shown, a terminal management device 1300 is provided, including: a first receiving module 1301 and a first transmitting module 1302, wherein:

[0212] The first receiving module 1301 is used to receive one or more inference configuration information sent by the network. The inference configuration information is carried by the channel state information (CSI) reported configuration message.

[0213] The first sending module 1302 is used to send a first message to the network when there is inapplicable inference configuration information. The first message includes at least one of the terminal's first applicable function, first inapplicable function, and first reason.

[0214] In one embodiment, the device 1300 further includes:

[0215] The second receiving module is used to receive terminal capability query information sent by the network. The terminal capability query information is used to initiate the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network.

[0216] The second sending module is used to send terminal capability information to the network. The terminal capability information includes at least one of the terminal's supported functions and model information.

[0217] In one embodiment, the supported functions of the terminal are functions that the terminal can use, characterized by terminal capability information.

[0218] In one embodiment, the model information includes at least one of the following: a trained model, a downloaded model, and an updated model possessed by the terminal.

[0219] In one embodiment, the inference configuration information is set by the network based on the supported functions and / or model information sent by the terminal.

[0220] In one embodiment, the inference configuration information includes at least one of the following:

[0221] Association identifier, model input information, model output information, beam type of model input and output, mapping relationship between model input and output, and CSI report type.

[0222] In one embodiment, the association identifier is used to ensure consistency of the terminal-side model during training and inference.

[0223] In one embodiment, the model input information includes at least one of the following: a first CSI resource configuration identifier, a set of beam indexes output by the model, the number of beams, beam identifiers, beam order, beam selection mode, the number of historical time instances, and the time interval between two consecutive historical time instances.

[0224] In one embodiment, the beam selection mode includes at least one of the following:

[0225] Fixed configuration mode, random selection mode, pre-configured continuous beam mode and discontinuous beam mode.

[0226] In one embodiment, the model output information includes at least one of the following:

[0227] The second CSI resource configuration identifier, the beam index set of model input, the number of beams, the beam identifier, the beam order, the number of future time instances, and the time interval between two consecutive future time instances.

[0228] In one embodiment, the beam types of the model input and output include at least one of narrowband and wideband.

[0229] In one embodiment, the mapping relationship between the model input and output includes at least one of the following:

[0230] The input beam is different from the output beam; the input beam is a subset of the output beam; and the input beam is the same as the output beam.

[0231] In one embodiment, when the mapping relationship between model input and output is such that the input beam is a subset of the output beam, or the input beam is the same as the output beam, the inference configuration information contains only one association identifier.

[0232] In one embodiment, when the mapping relationship between the model input and output is such that the input beam and the output beam are different, the inference configuration information includes two association identifiers, one of which is configured for the model input and the other of which is configured for the model output.

[0233] In one embodiment, the CSI report type includes one of periodic, semi-continuous, and non-periodic.

[0234] In one embodiment, the first applicable function is characterized as a function that the terminal can use for reasoning, and the first applicable function is a subset of the supported functions;

[0235] The first inapplicable function is the complement of the first applicable function in the supported functions.

[0236] In one embodiment, the first applicable function is determined by the association identifier, terminal assistance information, and model availability.

[0237] In one embodiment, the first applicable function is indicated by a bitmap or index or a CSI-ReportConfigID identifier.

[0238] In one embodiment, the terminal auxiliary information includes at least one of terminal battery level, memory usage, CPU utilization, GPU utilization, and network transmission latency.

[0239] In one embodiment, model availability includes having at least one of a trained model, an updated model, and a downloaded model.

[0240] In one embodiment, the first cause includes at least one of the following:

[0241] It does not meet the requirements for terminal auxiliary information, network auxiliary information, and model availability.

[0242] In one embodiment, the first message is carried by RRC reconfiguration completion and / or terminal assistance information signaling.

[0243] In one embodiment, the device 1300 further includes:

[0244] Do not activate periodic CSI reports.

[0245] In one embodiment, the device 1300 further includes:

[0246] Start the timer.

[0247] In one embodiment, the device 1300 further includes:

[0248] Send a third message to the network, the third message including at least one of a second applicable function, a second inapplicable function, and a second reason.

[0249] In one embodiment, the device 1300 further includes:

[0250] If the timer expires, implement at least one of the following communication processes: actively release inapplicable inference configurations, or send a message to the network.

[0251] In one embodiment, the timer stops when at least one of the following conditions is received: receiving a second message indicating that the network releases an inapplicable inference configuration and the inapplicable inference configuration becomes applicable.

[0252] In one embodiment, the device 1300 further includes:

[0253] The configuration release module is used to release the inference configuration of applicable and / or inapplicable functions when the terminal enters at least one of the Radio Resource Control (RRC) inactive state (RRC__INATIVE), Radio Resource Control (RRC_IDLE) idle state (RRC_IDLE), and handover state.

[0254] In one embodiment, such as Figure 14 As shown, a terminal management device 1400 is provided, including: a first transmitting module 1401 and a first receiving module 1402, wherein:

[0255] The first sending module 1401 is used to send one or more inference configuration information to the terminal, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message.

[0256] The first receiving module 1402 is configured to receive a first message sent by the terminal, the first message including at least one of the terminal's first applicable function, first inapplicable function, and first reason.

[0257] In one embodiment, the device 1400 further includes:

[0258] The second sending module is used to send terminal capability query information to the terminal, and the terminal capability query information is used to initiate the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network;

[0259] The second receiving module is used to receive terminal capability information sent by the terminal, wherein the terminal capability information includes at least one of the terminal's supported functions and model information.

[0260] In one embodiment, the device 1400 further includes:

[0261] The third sending module is used to send a second message to the terminal, the second message including at least one of releasing an inapplicable inference configuration and an updated inference configuration.

[0262] For specific limitations regarding the terminal management device, please refer to the limitations on the terminal management method above, which will not be repeated here. Each module in the aforementioned terminal management device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0263] In one embodiment, a communication device is provided, see [link to previous document]. Figure 15 . Figure 15 This is a schematic diagram of the structure of the terminal device provided in an embodiment of the present invention. Figure 15 The terminal device 1500 shown includes at least one processor 1501, a memory 1502, at least one network interface 1504, and a user interface 1503. The various components in the terminal device 1500 are coupled together via a bus system 1505. It is understood that the bus system 1505 is used to implement communication between these components. In addition to a data bus, the bus system 1505 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 1 Various buses are designated as bus system 1505. Additionally, this embodiment of the invention includes a transceiver 1506, which may consist of multiple components, including a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium.

[0264] The user interface 1503 may include a display, keyboard, or clicking device (e.g., mouse, trackball, touchpad, or touchscreen).

[0265] It is understood that the memory 1502 in this embodiment of the invention 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 (DRRAM). The memory 1502 of the systems and methods described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0266] In some implementations, memory 1502 stores elements, executable modules or data structures, or subsets thereof, or extended sets thereof: operating system 15021 and application program 15022.

[0267] The operating system 15021 includes various system programs, such as the framework layer, core library layer, and driver layer, used to implement various basic business functions and handle hardware-based tasks. The application program 15022 includes various applications, such as a media player and a browser, used to implement various application functions. The program implementing the method of this embodiment can be included in the application program 15022.

[0268] In this embodiment of the invention, by calling the program or instructions stored in the memory 1502, specifically the program or instructions stored in the application program 15022, the receiver is used to receive one or more inference configuration information sent by the network, the inference configuration information being carried by the Channel State Information (CSI) reporting configuration message; the transmitter is used to maintain the inference configuration when the inference configuration information is not applicable, and send a first message to the network, the first message containing at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

[0269] It is understood that the embodiments described in this invention can be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in this application, or combinations thereof.

[0270] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0271] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0272] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0273] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0274] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0275] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A terminal management method, characterized in that, The method is applied to a terminal, and the method includes: Receive one or more inference configuration information sent by the network, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message; When the inference configuration information is not applicable, the inference configuration is maintained, and a first message is sent to the network. The first message contains at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

2. The method according to claim 1, characterized in that, Prior to receiving one or more inference configurations from the receiving network, the method further includes: Receive terminal capability query information sent by the network, the terminal capability query information being used to initiate the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network; The terminal capability information is sent to the network, and the terminal capability information includes at least one of the terminal's supported functions and model information.

3. The method according to claim 2, characterized in that, The supported functions of the terminal are those functions that the terminal can use, represented by the terminal capability information.

4. The method according to claim 2, characterized in that, The model information includes at least one of the following: a trained model, a downloaded model, and an updated model possessed by the terminal.

5. The method according to any one of claims 1 to 4, characterized in that, The inference configuration information is set by the network based on the supported functions and / or model information sent by the terminal.

6. The method according to claim 1, characterized in that, The inference configuration information includes at least one of the following: Association identifier, model input information, model output information, beam type of model input and output, mapping relationship between model input and output, and CSI report type.

7. The method according to claim 6, characterized in that, The association identifier is used to ensure the consistency of the terminal-side model during training and inference.

8. The method according to claim 6, characterized in that, The model input information includes at least one of the following: first CSI resource configuration identifier, beam index set of model output, number of beams, beam identifier, beam order, beam selection mode, number of historical time instances, and time interval between two consecutive historical time instances.

9. The method according to claim 8, characterized in that, The beam selection mode includes at least one of the following: Fixed configuration mode, random selection mode, pre-configured continuous beam mode and discontinuous beam mode.

10. The method according to claim 6, characterized in that, The model output information includes at least one of the following: The second CSI resource configuration identifier, the beam index set of model input, the number of beams, the beam identifier, the beam order, the number of future time instances, and the time interval between two consecutive future time instances.

11. The method according to claim 6, characterized in that, The beam types of the model input and output include at least one of narrowband and wideband.

12. The method according to claim 6, characterized in that, The mapping relationship between the model input and output includes at least one of the following: The input beam is different from the output beam; the input beam is a subset of the output beam; and the input beam is the same as the output beam.

13. The method according to claim 12, characterized in that, When the mapping relationship between the model input and output is such that the input beam is a subset of the output beam, or the input beam is the same as the output beam, the inference configuration information contains only one association identifier.

14. The method according to claim 12, characterized in that, When the mapping relationship between the model input and output is such that the input beam and the output beam are different, the inference configuration information includes two association identifiers: one association identifier is configured for the model input, and the other association identifier is configured for the model output.

15. The method according to claim 6, characterized in that, CSI reporting types include periodic, semi-continuous, and non-periodic.

16. The method according to claim 1, characterized in that, The first applicable function is characterized as the function that the terminal can use for reasoning, and the first applicable function is a subset of the supported functions; The first unapplicable function is the complement of the first applicable function in the supported functions.

17. The method according to claim 1, characterized in that, The first applicable function is determined by the association identifier, terminal auxiliary information, and model availability.

18. The method according to claim 1, characterized in that, The first applicable function is indicated by a bitmap or index or CSI-ReportConfigID identifier.

19. The method according to claim 17, characterized in that, The terminal auxiliary information includes at least one of the following: terminal battery level, memory usage, CPU utilization, GPU utilization, and network transmission latency.

20. The method according to claim 17, characterized in that, The model availability includes having at least one of a fully trained model, a fully updated model, and a fully downloaded model.

21. The method according to claim 1, characterized in that, The first reason includes at least one of the following: It does not meet the requirements for terminal auxiliary information, network auxiliary information, and model availability.

22. The method according to claim 1, characterized in that, The first message is carried by RRC reconfiguration completion and / or terminal auxiliary information signaling.

23. The method according to claim 1, characterized in that, After sending the first message to the network, the method further includes: Do not activate periodic CSI reports.

24. The method according to claim 1, characterized in that, After sending the first message to the network, the method further includes: Start the timer.

25. The method according to claim 24, characterized in that, The method further includes: Send a third message to the network, the third message including at least one of a second applicable function, a second inapplicable function, and a second reason.

26. The method according to claim 24, characterized in that, The method further includes: If the timer expires, perform at least one of the following communication processes: actively release inapplicable inference configurations, or send a message to the network.

27. The method according to claim 24, characterized in that, The timer stops when at least one of the following conditions is received: receiving a second message indicating that the network has released an inapplicable inference configuration or that the inapplicable inference configuration has become applicable.

28. The method according to claim 1, characterized in that, After sending the first message to the network, the method further includes: If the terminal enters at least one of the following states: Radio Resource Control (RRC) inactive state (RRC__INATIVE), Radio Resource Control (RRC_IDLE) idle state (RRC_IDLE), and handover state, release the inference configuration of applicable and / or inapplicable functions.

29. A terminal management method, characterized in that, The method is applied to a network, and the method includes: Send one or more inference configuration information to the terminal, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message; The terminal receives a first message, the first message containing at least one of the terminal's first applicable function, first inapplicable function, first reason, and preferred inference configuration.

30. The method according to claim 29, characterized in that, The method further includes: Send terminal capability query information to the terminal, the terminal capability query information being used to initiate the terminal to report the supported AI / ML artificial intelligence / machine learning functions to the network; The terminal receives terminal capability information sent by the terminal, the terminal capability information including at least one of the terminal's supported functions and model information.

31. The method according to claim 29, characterized in that, After receiving the first message sent by the terminal, the method further includes: A second message is sent to the terminal, the second message including at least one of releasing an inapplicable inference configuration and an updated inference configuration.

32. A terminal management device, characterized in that, The device is used in a terminal, and the device includes: The first receiving module is used to receive one or more inference configuration information sent by the network, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message; The first sending module is configured to send a first message to the network when the inference configuration information is not applicable, wherein the first message includes at least one of the terminal's first applicable function, first inapplicable function, and first reason.

33. A terminal management device, characterized in that, The device is used in a network, and the device includes: The sending module is used to send one or more inference configuration information to the terminal, wherein the inference configuration information is carried by the Channel State Information (CSI) reporting configuration message; A receiving module is configured to receive a first message sent by the terminal, the first message including at least one of the terminal's first applicable function, first inapplicable function, and first reason.

34. A communication device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 28, or 29 to 31.

35. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 28, or 29 to 31.

36. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 28, or 29 to 31.

Images