Communication method, communication apparatus, storage medium and program product

By receiving and transmitting information related to the model processing unit in the 5G-Advanced communication system, the challenges of terminal-side resource management and scheduling are solved, achieving efficient CSI report transmission and scheduling, and improving the system's processing efficiency and reliability.

WO2026237685A1PCT designated stage Publication Date: 2026-11-19ZTE CORP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2026/092497
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-16
Filing Date
2026-04-23
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

In 5G-Advanced communication systems, the introduction of artificial intelligence/machine learning technologies increases the complexity of CSI processing on the terminal side, leading to challenges in resource management and scheduling. In particular, when coordinating the management of traditional CPU resources and new APU resources across different AI use cases, how to efficiently coordinate resource usage to ensure the feasibility and efficiency of CSI processing has become a problem.

Method used

A communication method is provided that, by receiving and sending information related to the instruction model processing unit, allows network nodes to adjust configurations or reschedule, so as to efficiently manage and coordinate resource usage and ensure that CSI reports on the terminal side can be reliably sent.

Benefits of technology

It enables efficient and reasonable assistance to network nodes in managing resources in 5G-Advanced communication systems, ensures reliable transmission and scheduling of CSI reports, improves terminal-side processing efficiency, and reduces latency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2026092497_19112026_PF_FP_ABST
    Figure CN2026092497_19112026_PF_FP_ABST
Patent Text Reader

Abstract

A communication method, a communication apparatus, a storage medium and a program product. The method is applied to a first node. The method comprises: receiving configuration information from a second node, wherein the configuration information is used by a first node to send a first report; and sending first information to the second node, wherein the first information is used for indicating information related to a model processing unit of the first node.
Need to check novelty before this filing date? Find Prior Art

Description

Communication methods, communication devices, storage media and software products

[0001] This disclosure claims priority to Chinese patent application No. 202510640429.5, filed on May 16, 2025, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This disclosure relates to the field of communication technology, and in particular to a communication method, communication device, storage medium, and program product. Background Technology

[0003] In the 5G-A communication standard, artificial intelligence / machine learning (AI / ML) technologies are gradually being introduced into the physical layer of the communication system. This introduction of AI / ML technologies increases the complexity of terminal-side processing. Summary of the Invention

[0004] On the one hand, a communication method is provided, including:

[0005] The configuration information is received from the second node; the configuration information is used by the first node to send the first report.

[0006] Send first information to the second node; the first information is used to indicate information related to the model processing unit of the first node.

[0007] On the other hand, a communication device is provided, which includes a receiving module and a transmitting module.

[0008] A receiving module is used to receive configuration information from a second node; the configuration information is used by the first node to send the first report.

[0009] The sending module is used to send first information to the second node; the first information is used to indicate information related to the Model Processing Unit (MPU) of the first node.

[0010] On the other hand, a communication method is provided for application to a second node, the method including:

[0011] Send configuration information to the first node; the configuration information is used by the first node to send the first report;

[0012] Receive first information from the first node; the first information is used to indicate information related to the model processing unit.

[0013] In another aspect, a communication device is provided, comprising: a transmitting module and a receiving module;

[0014] A sending module is used to send configuration information to the first node; the configuration information is used by the first node to send the first report.

[0015] The receiving module is used to receive first information from the first node; the first information is used to indicate information related to the model processing unit.

[0016] In another aspect, a communication device is provided, comprising: a memory and a processor; the memory and the processor are coupled; the memory is used to store a computer program; and the processor implements the above-described communication method when executing the computer program.

[0017] In another aspect, a computer-readable storage medium is provided, on which computer program instructions are stored, which, when executed by a processor, implement the above-described communication method.

[0018] In another aspect, a computer program product is provided, which includes computer program instructions that, when executed by a processor, implement the aforementioned communication method. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this disclosure, the accompanying drawings used in some embodiments of this disclosure will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings.

[0020] Figure 1 is a system architecture diagram of a communication system according to some embodiments.

[0021] Figure 2 is a flowchart of a communication method provided according to some embodiments.

[0022] Figure 3 is a schematic diagram of a bilateral model performance monitoring process provided according to some embodiments.

[0023] Figure 4 is a flowchart of another communication method provided according to some embodiments.

[0024] Figure 5 is a block diagram of a communication device according to some embodiments.

[0025] Figure 6 is a block diagram of another communication device provided according to some embodiments.

[0026] Figure 7 is a block diagram of another communication device provided according to some embodiments. Detailed Implementation

[0027] The technical solutions of this disclosure will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0028] It should be noted that, in this disclosure, the words "exemplarily" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplarily" or "for example" in this disclosure should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the words "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner.

[0029] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0030] In the description of this disclosure, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. "And / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "more than one" means two or more.

[0031] In some embodiments, the term "determine" may encompass a wide variety of actions. For example, "determine" may include calculation, processing, deduction, investigation, instruction, lookup (e.g., searching in a table, database, or other data structure), etc. Furthermore, "determine" may include sending (e.g., sending information), receiving (e.g., receiving information), accessing (e.g., accessing data in memory), etc. Additionally, "determine" may include resolving, selecting, establishing, etc.

[0032] In wireless communication systems, terminals require specific processing resources to measure, calculate, and report Channel State Information (CSI). Fourth-generation (4G) and 5G communication standards introduce the concept of a CSI Processing Unit (CPU), whose number reflects the terminal's CSI processing capability. The protocols also define different CPU usage numbers for different CSI processing use cases to reflect the complexity of the CSI processing. Furthermore, the network side can schedule CSI reports based on the number of CPUs on the terminal to ensure that the terminal can process scheduled CSI reports as much as possible. In the 5G-Advanced (5G-A) communication standard, artificial intelligence / machine learning technologies are gradually being introduced into the physical layer technology of communication systems. Typical examples include CSI-related enhanced use cases (e.g., CSI compression feedback, CSI prediction) and beam management-related enhanced use cases (e.g., spatial beam prediction, temporal beam prediction). These functions face challenges in terms of computing power and latency required for new CSI processing procedures. For these AI / ML-enhanced use cases, AI models are often a common processing method. Typically, AI / ML models require at least one of the following resources: computing resources, computing power resources, storage resources, time resources, software resources, and hardware resources. This necessitates that the terminal be equipped with a dedicated AI / ML Processing Unit (APU) to efficiently execute these computationally complex services in parallel, improving computational efficiency and reducing latency. With the upcoming 6G era, more physical layer use cases will adopt AI / ML technology for enhancement (including both network-managed AI functions and UE-implemented AI functions). This may present new challenges for CSI processing resource management on the network side: how to coordinate the management of traditional CPU resources and new APU resources, and how to coordinate the APU resources used by different AI use cases to ensure that the terminal can efficiently and completely process CSI reports. Furthermore, when models and CSI reporting are associated, the scheduling of CSI reporting needs to consider the impact of different functions, resource usage, and model activation / switching latency on the CSI processing timeline. This issue urgently needs to be addressed.

[0033] To address the aforementioned technical problems, this disclosure provides a communication method. After receiving configuration information for a first node to send a first report, the first node can send first information to a second node to notify the second node of information related to the first node's model processing unit. This allows the second node to adjust its configuration or reschedule based on the information related to the first node's model processing unit, ensuring the feasibility of the second node's configuration or scheduling of the first node. The model processing unit can abstract the resources occupied by the model in the first node. Therefore, the first information, by indicating information related to the first node's model processing unit, can efficiently and conveniently report relevant information about the model's resource usage and efficiently and reasonably assist / support the second node in performing configuration or scheduling.

[0034] The communication method provided in this disclosure can be applied to systems with various communication standards. For example, the systems to which the communication method provided in this disclosure is applicable include, but are not limited to, long-term evolution (LTE) systems, various versions based on LTE evolution, 5th generation mobile communication technology (5G) systems, future mobile communication networks (such as 6G mobile communication networks), or multiple converged communication systems. Furthermore, the communication method provided in this disclosure can also be applied to future-oriented communication systems.

[0035] For example, the above communication method can be applied to the communication system shown in FIG1. ​​As shown in FIG1, the communication system includes: a first node 101 and a second node 102.

[0036] Here, the first node 101 is used to receive configuration information from the second node 102; the configuration information is used by the first node 101 to send a first report; or, it is used to send first information to the second node 102; the first information is used to indicate information related to the model processing unit of the first node 101.

[0037] The second node 102 is used to send configuration information to the first node 101; the configuration information is used by the first node 101 to send a first report; or, it is used to receive first information from the first node 101; the first information is used to indicate information related to the model processing unit.

[0038] In some embodiments, the first node 101 can be a terminal and the second node 102 can be a base station; or the first node 101 can be a base station and the second node 102 can be a terminal. Figure 1 illustrates an example where the first node 101 can be a terminal and the second node 102 can be a base station.

[0039] In some embodiments, the terminal can be a device with wireless transceiver capabilities, which can be deployed on land, including indoors or outdoors, handheld, wearable, or vehicle-mounted; it can also be deployed on water (such as on ships); and it can also be deployed in the air (e.g., on airplanes, balloons, and satellites). The terminal can be a mobile phone, tablet computer, computer with wireless transceiver capabilities, virtual reality (VR) terminal, augmented reality (AR) terminal, wireless terminal in industrial control, wireless terminal in self-driving, wireless terminal in remote medical care, wireless terminal in smart grid, wireless terminal in transportation safety, wireless terminal in smart city, wireless terminal in smart home, etc. The embodiments of this application do not limit the application scenarios. The term "terminal" can sometimes also refer to a user, user equipment (UE), access terminal, UE unit, UE station, mobile station, mobile station, remote station, remote terminal, mobile device, UE terminal, wireless communication device, UE agent, or UE device, etc., but the embodiments of this application do not limit this to these terms.

[0040] In some embodiments, the base station may be a base station in Long Term Evolution (LTE), Long Term Evolution Advanced (LTEA), or an evolved Node B (eNB or eNodeB), a base station device in a 5G network, or a base station in a future communication system. The base station may include various macro base stations, micro base stations, femtocell base stations, wireless remote extensions, reconfigurable intelligent surfaces (RISs), routers, wireless fidelity (WIFI) devices, or various network-side devices such as primary cells and secondary cells.

[0041] In some embodiments, higher-layer signaling includes, but is not limited to, radio resource control (RRC), media access control-control element (MAC CE), or other higher-layer signaling above the physical layer. Physical layer signaling includes, but is not limited to, downlink control information (DCI) and uplink control information (UCI). As an example, physical layer signaling can be transmitted between the base station and the terminal on the physical downlink control channel (PDCCH) or on the physical uplink control channel (PUCCH).

[0042] In some embodiments, the uplink channel may be a PUCCH or a Physical Uplink Shared Channel (PUSCH). In some embodiments, the downlink channel may be a PDCCH or a Physical Downlink Shared Channel (PDSCH). In some embodiments, the uplink reference signal may be a Sounding Reference Signal (SRS), a Physical Random Access Channel (PRACH), or a Demodulation Reference Signal (DMRS). In some embodiments, the downlink reference signal may be a Synchronization Signal Block (SSB), a Channel State Information Reference Signal (CSI-RS), a Demodulation Reference Signal (DMRS), a Tracking Reference Signal (TRS), or a Phase-Tracking Reference Signal (PTRS). In some embodiments, the uplink signal may be an uplink channel or an uplink reference signal. In some embodiments, the downlink signal may be a downlink channel or a downlink reference signal. In some embodiments, DCI is equivalent to PDCCH.

[0043] In some embodiments, the parameter indicator can also be called an index or an identifier (ID), and these terms are equivalent. For example, a resource identifier in a wireless system can also be called a resource indicator or a resource index. Here, the resource identifier in a wireless system includes, but is not limited to, one of the following: identifiers corresponding to reference signal resources, reference signal resource groups, reference signal resource configurations, channel state information (CSI) reports, CSI report sets, terminals, base stations, panels, neural networks, sub-neural networks, and neural network layers. The base station can indicate the identifier of one or a group of resources to the terminal through various higher-layer signaling or physical-layer signaling. The terminal can feed back the identifier of one or a group of resources to the base station through various higher-layer signaling and / or physical-layer signaling.

[0044] In some embodiments, time-domain units can be slots, sub-slots, first slots, last slots, frames, sub-frames, transmission time intervals (TTI), transmission occasions, opportunities, symbols, sub-symbols, first symbols, and last symbols. A slot or sub-slot includes at least one symbol. A symbol refers to a time unit within a subframe, frame, or slot, such as an orthogonal frequency division multiplexing (OFDM) symbol, a single-carrier frequency division multiple access (SC-FDMA) symbol, or an orthogonal frequency division multiple access (OFDMA) symbol. Specific time units include seconds, milliseconds, microseconds, or other typical time units.

[0045] In some embodiments, frequency domain units may be a bandwidth part (BWP), a carrier, a subcarrier, a resource element (RE), a group of resource elements (REG), a resource block (RB), a physical resource block (PRB), a group of resource blocks (RBG), a subband (SB), or a subband group (SBG).

[0046] In some embodiments, a spatial element may be a Transmission and Reception Point (TRP), a transmitting antenna, a transmitting antenna group, a transmitting antenna port, a transmitting antenna port group, a receiving antenna, a receiving antenna group, a receiving antenna port, a receiving antenna port group, an antenna layout, an antenna polarization direction, a reference signal port, or a reference signal port group.

[0047] In some embodiments, to calculate channel state information or perform channel estimation, mobility management, positioning, etc., base stations or users need to transmit reference signals (RS). Reference signals include, but are not limited to, channel-state information reference signals (CSI-RS), which include zero-power CSI-RS (ZP CSI-RS) and non-zero-power CSI-RS (NZP CSI-RS), channel-state information-interference measurement (CSI-IM), sounding reference signals (SRS), synchronization signals blocks (SSB), physical broadcast channels (PBCH), and synchronization signal block / physical broadcast channel (SSB / PBCH). NZP CSI-RS can be used to measure channel or interference. CSI-RS can also be used for tracking, called tracking reference signals (TRS), while CSI-IM is generally used to measure interference, and SRS is used to measure the uplink channel. Additionally, the time-frequency resources used for transmitting reference signals comprise a set of resource elements (REs) called reference signal resources, such as CSI-RS resource, SRS resource, CSI-IM resource, and SSB resource. In this paper, SSB includes synchronization signal blocks and / or physical broadcast channels.

[0048] In some embodiments, to save signaling overhead, multiple reference signal resources may be divided into multiple sets (e.g., CSI-RS resource set, CSI-IM resource set, SRS resource set). Each reference signal resource set includes at least one reference signal resource, and multiple reference signal resource sets may all come from the same reference signal resource setting (e.g., CSI-RS resource setting, SRS resource setting; here, CSI-RS resource setting may be merged with CSI-IM resource setting and both are called CSI-RS resource setting) to configure parameter information.

[0049] In some examples, to better transmit data or signals, the base station or terminal needs to acquire measurement parameters. These measurement parameters may include channel state information or other parameters used to characterize the channel. Here, the channel state information may include at least one of the following: Channel State Information - Reference Signal Resource Indicator (CSI-RS resource indicator, CRI), Synchronization Signal Block Resource Indicator (SSBRI), L1 Reference Signal Received Power (L1-RSRP or RSRP), Differential RSRP, L1 Signal to Interference Noise Ratio (L1-SINR or SINR), Differential L1-SINR, Reference Signal Received Quality (RSRQ), Channel Quality Indicator (CQI), Precoding Matrix Indicator (PMI), Layer Indicator (LI), Rank Indicator (RI), and precoding information. Precoding information includes first-type precoding information, such as codebook-based precoding information. Here, the precoding matrix indicates one type of codebook-based precoding information. The codebook includes at least Rel-15 Type I codebook, Rel-15 Type II codebook, Rel-16eType II codebook, Rel-16eType II port selection codebook, Rel-17feType II codebook, Rel-18doppler Type II codebook, Rel-18 CJT Type II codebook, and Rel-19enhanced Type I codebook. Precoding information also includes precoding information based on non-codebook implementations, such as second-type precoding information, or precoding information obtained based on advanced technologies such as artificial intelligence.

[0050] In some embodiments, the channel information includes at least a channel matrix, a channel coefficient matrix, a precoding matrix obtained by processing the channel matrix, eigenvectors, eigenvalues, precoding, a precoding matrix index, a precoding coefficient matrix, and a precoding basis vector matrix. Here, the channel matrix and the precoding matrix can be any combination of the following domains: spatial domain, frequency domain, time domain, transformed angular domain, time delay domain, and Doppler domain.

[0051] In some embodiments, to transmit channel state information at the physical layer, the terminal and the base station define a CSI report (or CSI report configuration). The CSI report defines at least one of the following parameters: time-frequency resources used for CSI feedback, report quantity for CSI feedback, time-domain category (report ConfigType) for CSI feedback, channel measurement resources, interference measurement resources, and measurement bandwidth. The CSI report can be transmitted on uplink transmission resources, including PUSCH and PUCCH. The CSI report also includes time-domain characteristics, including periodic CSI reports (P-CSI), aperiodic CSI reports (AP-CSI), and semi-persistent CSI reports (SP-CSI).

[0052] In some embodiments, the base station configures N_C CSI reports to be fed back to the terminal via higher-layer signaling and / or physical-layer signaling. Each CSI report has an identity (ID), called the CSI report ID. The terminal can select M_C CSI reports from the N_C CSI reports based on its computing or processing capabilities and the requirements of the base station. Based on the uplink feedback resources, the terminal feeds back at least one of the M_C CSI reports, where N_C and M_C are positive integers, and M_C <= N_C. In one example, M_C CSI reports need to be fed back, but the feedback resources for at least two of the M_C reports conflict. The conflict occurs when at least one symbol and / or at least one subcarrier is the same in the transmission resources (e.g., PUCCH or PUSCH) corresponding to the two reports. In some embodiments, feeding back CSI can also be called transmitting CSI, sending CSI, or reporting CSI, such as carrying channel state information on uplink transmission resources for feedback, transmission, or reporting. The uplink transmission resources and the corresponding CSI are both indicated by a channel state information report. In some embodiments, feeding back or transmitting a CSI report means feeding back the channel state information configured in the CSI report. In some embodiments, feeding back or transmitting a CSI report means transmitting the content configured to be transmitted in the CSI report through transmission resources.

[0053] In some embodiments, a beam includes a transmit beam, a receive beam, a receive beam, and a transmit beam pair, and a transmit beam and a receive beam pair. The beam index can be replaced with a resource index (e.g., a reference signal resource index) because the beam can be transmittedly bound to some time-frequency code resource. The beam can also be a transmission (transmit / receive) mode; said transmission mode may include spatial multiplexing, frequency domain / time domain diversity, beamforming, etc. In some embodiments, a beam pair includes a combination of a transmit beam and a receive beam.

[0054] In some embodiments, a beam is equivalent to beam state, quasi-co-location (QCL) state, transmission configuration indicator (TCI) state, spatial relation, spatial relation information, reference signal (RS), reference signal resource, spatial filter, and precoding. In some embodiments, a transmit beam is equivalent to QCL state, TCI state, spatial relation, spatial relation state, uplink / downlink reference signal (e.g., CSI-RS, SSB, DMRS, SRS, PRACH), reference signal resource, transmit spatial filter, and transmit precoding. In some embodiments, a receive beam is equivalent to QCL state, TCI state, spatial relation, spatial relation state, spatial receive parameters, spatial filter, receive spatial filter, and receive precoding. Here, the spatial filter is also called a spatial domain filter, and can be either from the base station side or the UE side.

[0055] In some embodiments, the base station may configure quasi-co-location (QCL) for two reference signals and inform the user terminal to describe channel characteristic assumptions. The quasi-co-location involves parameters including at least one of the following: Doppler spread, Doppler shift, delay spread, average delay, average gain, and spatial Rx parameter (or spatial parameter). Here, spatial parameters may include spatial reception parameters, angle information, spatial correlation of the received beam, average delay, and correlation of the time-frequency channel response (including phase information). Angle information may include at least one of the following: angle of arrival (AOA), angle of departure (AOD), zenith angle of departure (ZOD), and zenith angle of arrival (ZOA). Spatial filtering can be at least one of the following: a discrete Fourier transform (DFT) vector, a precoded vector, a DFT matrix, a precoded matrix, or a vector formed by a linear combination of multiple DFTs, or a vector formed by a linear combination of multiple precoded vectors. In some embodiments, vector and vector array are interchangeable concepts.

[0056] In some embodiments, artificial intelligence (AI) includes self-learning devices, components, software, and modules such as machine learning (ML), deep learning, reinforcement learning, transfer learning, deep reinforcement learning, meta-learning, and large language models (LLM). In some embodiments, AI is implemented through an AI network (or neural network), which includes multiple layers, each layer including at least one node. In one example, the neural network includes an input layer, an output layer, and at least one hidden layer. Here, each layer of the neural network includes, but is not limited to, using at least one of fully connected layers, dense layers, convolutional layers, transposed convolutional layers, directly connected layers, activation functions, normalization layers, pooling layers, and positional encoding layers. In some embodiments, each layer of the neural network may include a sub-neural network, such as a residual network block (or Resnet block), a dense network (Densenet Block), a fully convolutional network (FCN), a recurrent neural network (RNN), a long short-term memory network (LSTM), or a transform network (Transformer). Artificial intelligence networks can be implemented through models. Here, a model can include a neural network model, which includes the neural network model structure and / or neural network model parameters. The neural network model structure can be simply referred to as the model structure, and the neural network model parameters can be simply referred to as network parameters or model parameters. A model structure defines the number of layers in the neural network, the size of each layer, the activation function, the connection configuration, the convolutional kernel and stride, the convolution type (e.g., 1D convolution, 2D convolution, 3D convolution, hollow convolution, transposed convolution, separable convolution, grouped convolution, expanded convolution, etc.), the number of sub-network blocks (e.g., the number of Transformer blocks), and other network architectures. Network parameters are the weights and / or biases of each layer in the neural network model and their values. A single model structure can correspond to multiple sets of different neural network model parameter values ​​to adapt to different scenarios. Neural network model parameters are obtained through online or offline training. For example, by inputting at least one sample and label, the neural network model is trained to obtain the neural network model parameters.

[0057] In some embodiments, a model refers to multiple linear or nonlinear components through which the data flow from the original input of a sample to the output target passes. The model includes at least one of the following: a machine learning algorithm, a neural network model, a non-artificial intelligence module for processing information or a corresponding model thereof, or a functional component or function that maps input information to output information (this mapping includes both linear and nonlinear mappings).

[0058] In some embodiments, "model" is a general term used to describe a processing method, function, feature, or feature group that a terminal can perform. In some embodiments, "model" is equivalent to function or functionality, functional module, functional entity, processing method, information processing method, implementation, feature, feature group, configuration, configuration combination, configuration set, parameter, parameter set, dataset (e.g., a dataset used for model training), data-driven algorithm, signal processing algorithm, CSI reporting, non-CSI reporting, beam reporting, non-beam reporting, CSI reporting configuration, non-CSI reporting configuration, beam reporting configuration, non-beam reporting configuration, resource configuration, resource set configuration, etc.

[0059] In some embodiments, a model may be a data-driven algorithm that applies AI / ML techniques to generate a set of outputs from a set of inputs. Additionally, a model may be a linear or nonlinear algorithm, or a combination of both. Furthermore, a function may refer to the functionality enabled by the AI / ML model. Alternatively, a function may refer to a set of parameters or configurations for a particular feature. For example, a UE may use a recurrent neural network (RNN) model to predict CSI at future time points; the RNN model is the model, and the CSI prediction is the function. Different models and / or functions may be associated with different configurations (e.g., Radio Resource Control (RRC) configurations). Model activation refers to activating the configuration corresponding to the terminal. Similarly, model deactivation, switching, and rollback may refer to deactivating the corresponding configuration, switching the configuration, and rolling back to a configuration without a model, respectively. In some embodiments, a model is equivalent to a model group consisting of multiple models sharing commonalities within a scenario, case, function, feature, or scenario. In some embodiments, a model may be implicitly represented as or equal to an ID, such as a model ID, association ID, identifier ID, pairing ID, function ID, configuration ID, sub-configuration ID, or group ID.

[0060] It should be noted that Figure 1 is only an exemplary framework diagram, and the number of devices included in Figure 1 and the names of each device are not limited.

[0061] The application scenarios of the embodiments disclosed herein are not limited. The system architecture and business scenarios described in the embodiments of this disclosure are for the purpose of more clearly illustrating the technical solutions of the embodiments of this disclosure, and do not constitute a limitation on the technical solutions provided by the embodiments of this disclosure. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided by the embodiments of this disclosure are also applicable to similar technical problems.

[0062] The communication method provided in the embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0063] The communication method provided in this embodiment can be applied to the first node 101 in the communication system shown in FIG1. ​​FIG2 shows a flowchart of a communication method, which includes the following steps S201-S202:

[0064] S201. Receive configuration information from the second node.

[0065] The configuration information is used by the first node to send the first report.

[0066] S202, Send the first message to the second node.

[0067] The first information is used to indicate information related to the model processing unit of the first node.

[0068] It should be noted that, in order to manage the terminal-side models and resources (such as the storage resources they occupy), the resources occupied by the models are abstracted into model processing units. Note that MPU is a general term reflecting the resources occupied or required by a model; it can refer to the occupancy of one or more types of resources such as computing power and storage, and is not limited to a specific resource. The resources occupied by each model (i.e., the number of MPUs) and / or the total resources that the UE can simultaneously occupy (i.e., the total number of MPUs) are reported to the base station through the UE. Obviously, different models may occupy different numbers of MPUs, which is related to at least one of the following factors: the dimension of the model input / output, the type of model input / output parameters, the model structure, the number of model parameters, the model computational complexity, the UE capability indication value, the temporal characteristics of the reference signal resources, etc. Furthermore, the total number of MPUs is strongly correlated with the UE capability; the UE needs to indicate the total number of MPUs it can support in its UE capability report. Moreover, MPU is a general term in definition. It can be called, but is not limited to, AI processing unit (e.g., AIPU / APU), machine learning processing unit (e.g., ML PU), and for CSI-related use cases, it can also be called, but is not limited to, AI CSI processing unit (e.g., CPU_AI / AI_CPU / A-CPU / CPU-A), machine learning CSI processing unit (e.g., CPU_ML / ML_CPU / ML-CPU / CPU-ML), model CSI processing unit (CPU_Model / Model_CPU / M-CPU / CPU-M), etc. Here, the connector can be, but is not limited to, the following examples (e.g., - or _ or , or space), which have no actual physical meaning and are only used to distinguish different representations.

[0069] It should be understood that after receiving the configuration information for the first node to send the first report, the first node can send first information to the second node to notify the second node of information related to the first node's model processing unit. This allows the second node to adjust its configuration or reschedule based on the information related to the first node's model processing unit, ensuring the feasibility of the second node's configuration or scheduling of the first node. The model processing unit can abstract the resources occupied by the model in the first node; therefore, the first information, by indicating information related to the first node's model processing unit, can efficiently and easily report relevant information about the model's resource usage and efficiently and reasonably assist / support the second node in performing configuration or scheduling.

[0070] In some embodiments, if the first node is unable to reliably send a first report based on configuration information from the second node, it may send first information to the second node to trigger the second node to adjust its configuration, reconfigure, or reschedule.

[0071] In some embodiments, the model processing unit of the first node includes a first type of model processing unit and a second type of model processing unit; relevant information of the first type of model processing unit has been reported to the second node, while relevant information of the second type of model processing unit has not been reported to the second node. Alternatively, the relevant information of the first type of model processing unit is visible or opaque to the second node, while the relevant information of the second type of model processing unit is invisible or transparent to the second node. In other words, when scheduling / configuring the behavior of the first node, the second node can or should consider the relevant information of the first type of model processing unit, but cannot or should not consider the relevant information of the second type of model processing unit. For example, the relevant information of the first type of model processing unit can be reported to the second node through capability reporting (or UE capability reporting), etc.

[0072] In some embodiments, the first information includes at least one of the following: the number of unoccupied model processing units in the model processing units of the first node at the current time, the number of occupied model processing units in the model processing units of the first node at the current time, the occupancy trend of the model processing units of the first node in a future time period, and the unoccupied trend of the model processing units of the first node in a future time period. This allows the second node to adjust its configuration or reschedule the actions that the first node can perform in the current or future time period (such as adjusting the content of the first report to complete the submission of the first report) based on the number of unoccupied and occupied model processing units in the model processing units of the first node at the current time, ensuring the feasibility of configuration scheduling. Alternatively, it allows the second node to configure or schedule the actions that the first node can perform in a future time period based on the occupancy and unoccupied trends of the model processing units of the first node in a future time period. In some embodiments, the quantity information may include quantity and / or percentage.

[0073] In some embodiments, the method further includes: reporting a first report to a second node based on configuration information; the first report is obtained based on the model processing unit of the first node.

[0074] In some embodiments, if the number of model processing units available in the first node is less than the number of model processing units required for the report to be scheduled and reported by the configuration information, the first report may include a portion of the report or part of the report content from the report to be scheduled and reported by the configuration information; or the first report may include a portion of the updated report content from the report to be scheduled and reported by the configuration information; or the first report may include a portion of the unupdated report content; or the first report may include a portion of the report content obtained based on the model processing units; or the first report may include a portion of the report content not obtained based on the model processing units. It should be understood that if the number of model processing units available in the first node is less than the number of model processing units required for the report to be scheduled and reported by the configuration information, it indicates that the first node does not have the capability to report a complete report corresponding to the configuration information. In this case, the first report reported by the first node may include a portion of the report or part of the report content from the report to be scheduled and reported by the configuration information, or may include a portion of the updated report content from the report to be scheduled and reported by the configuration information (or, the first report may contain some unupdated report content and some updated report content), or some content in the first report may be generated based on the model processing units (other content may be generated based on resources other than the model processing units). In this way, the first node can selectively report the report corresponding to the configuration information based on its own capabilities. In some embodiments, the content of the first report may also include any content reported by the first node, such as all 0 information or all 1 information, so as to ensure the integrity of the format or the content of the first report.

[0075] In some embodiments, the reports or report content in the first report are determined based on their priority. Priority is determined based on at least one of the following: the index of the report or report content, the type of the report or report content, the function or functional group of the report or report content, the use case of the report or report content, the characteristics or feature groups of the report or report content, and the number of model processing units required by the report or report content. That is, the reports or report content reported by the first node can be determined based on their priority. For reports or report content that configuration information requires to be reported, higher-priority reports or report content can be reported, while lower-priority reports or report content may not be reported or updated.

[0076] It should be noted that the report types, report entities (report quantity), report content types, functions, function groups, use cases, features, or feature groups mentioned in this disclosure can be used interchangeably.

[0077] In some embodiments, the method further includes: reporting second information to the second node; the second information includes at least one of the following: location information of a portion of the report or part of the report content, indication information of whether each report in the first report is based on the model processing unit, index of the report with abnormal content in the first report, index of the report with abnormal timeline occupation in the first report, time unit offset corresponding to the report with abnormal timeline occupation in the first report, and time unit offset caused by model switching.

[0078] It should be understood that, in the case of the first node reporting the first report, the first node may also report second information to notify / inform the second node of relevant information regarding a portion of the report or part of the report content in the first report. For example, location information of a portion of the report or part of the report content can enable the second node to know the location of updated or outdated reports or content in the first report. In some embodiments, the location information of a portion of the report or part of the report content may include the location of updated reports or content and / or the location of outdated reports or content.

[0079] The first report is reported to indicate whether each report is based on the model processing unit, so that the second node knows whether each report or the content of each report in the first report is based on the model processing unit.

[0080] The index of reports with abnormal content allows the second node to know about reports or report content that have not been updated or have not been generated based on the configuration information.

[0081] The index of reports with abnormal timeline occupancy in the first report is provided. These abnormal timeline occupancy may be due to the report not being based on model processing units, or not being based on a sufficient number of model processing units. When the number of model processing units is insufficient, abnormal timeline occupancy can be caused by a longer timeline occupancy or a larger number of time units (longer report or report content generation time). If the report is based on model processing units or has not been updated, abnormal timeline occupancy can be caused by a shorter timeline occupancy or a smaller number of time units (shorter report or report content generation time).

[0082] The time unit offset corresponding to the abnormally occupied timeline in the first report can reflect the length of time that the abnormally occupied timeline or the occupied time corresponding to the report content has become longer or shorter.

[0083] In the time unit offset caused by model switching, the model switching is to switch the model that generates the report or report content based on the configuration information or scheduling to another model.

[0084] In some embodiments, the second information may be partially or entirely included in the first report. That is, the second information may be entirely included in the first report. Alternatively, some of the second information may be reported independently, while other information may be included in the first report. In some embodiments, the second information may not be included in the first report and may be reported entirely independently.

[0085] In some embodiments, if the number of model processing units available in the first node is less than the number of model processing units required for multiple related functions, the report content corresponding to the multiple functions in the first report is not updated; or the report content corresponding to the multiple functions in the first report is not obtained based on model processing units; or the report content corresponding to the multiple functions in the first report is obtained based on a first number of model processing units, where the first number is less than the number of model processing units required for the multiple functions. It should be understood that the multiple related functions can be functions with collaborative relationships, binding relationships, or simultaneous operation. For example, the operation of some functions requires the output of another function, or the operation of some functions requires the assistance / support of another function. If the number of model processing units available in the first node is less than the number of model processing units required for multiple related functions, it indicates that the first node does not have the ability to fully activate / use / support multiple related functions. In this case, the report content corresponding to these multiple functions may not be updated, or the report content corresponding to these multiple functions may not be obtained based on model processing units, or the report content corresponding to these multiple functions may be obtained based on fewer model processing units, or only the report content corresponding to some functions may be updated.

[0086] In some embodiments, when the report content corresponding to multiple functions in the first report is obtained based on a first number of model processing units, the first number is determined based on the number of model processing units required for the multiple functions and a first offset; the first offset is determined based on one of the following: predefined or preconfigured; determined by the first node; indicated by the second node to the first node; or determined through negotiation between the first node and the second node. It should be understood that when multiple functions with a relationship need to complete the generation of report content through a first number of model processing units, the first number can be predefined or preconfigured, or it can be determined autonomously by the first node, indicated by the second node to the first node, or determined through negotiation between the first node and the second node.

[0087] In some embodiments, the first report includes third information and tag information corresponding to the third information; the third information is generated based on the model processing unit, and the tag information is generated based on the channel state information processing unit; the occupancy start time of the model processing unit that generates the third information is the same as the occupancy start time of the channel state information processing unit that generates the tag information, or is determined based on the occupancy start time of the channel state information processing unit that generates the tag information and a second offset, or is the last time unit of the occupancy time of the channel state information processing unit that generates the tag information, or is determined based on the last time unit of the occupancy time of the channel state information processing unit that generates the tag information and a third offset value.

[0088] It should be noted that since the third information is generated based on the model processing unit, it reflects the capability / performance of the model processing unit in the first node. The label information of the third information is used to indicate the input information or quantized input information (such as the input information after high-precision codebook quantization) of the model processing unit (or the model corresponding to the model processing unit) when generating the third information. The first report includes the label information and the third information. The second node can know the input and output information of the model processing unit of the first node, and thus can judge / calculate / determine / evaluate the capability / performance of the model processing unit of the first node. For example, assuming the third information is generated based on information 1, the label information of the third information can be information 1 or information 1 after processing. Here, information 1 after processing can be information 1 after quantization. In some embodiments, quantization can be scalar quantization, vector quantization, codebook quantization, or high-precision codebook quantization, etc.

[0089] The starting time of the occupancy of the channel state information processing unit that generates tag information is determined based on the second offset. That is, the starting time of the occupancy of the model processing unit that generates third information can be the sum of the starting time of the occupancy of the channel state information processing unit that generates tag information and the second offset.

[0090] The start time of the model processing unit for generating third information is determined based on the last time unit of the channel state information processing unit's occupation time for generating tag information and the third offset value. That is, the start time of the model processing unit for generating third information can be the sum of the last time unit of the channel state information processing unit's occupation time for generating tag information and the third offset value. In some embodiments, the second offset value or the third offset value can be 0. When the start time of the model processing unit for generating third information is the sum of the start time of the channel state information processing unit for generating tag information and the second offset value, the occupation time of the model processing unit and the occupation time of the channel state information processing unit can be simultaneous. When the start time of the model processing unit for generating third information can be the sum of the last time unit of the channel state information processing unit's occupation time for generating tag information and the third offset value, the occupation time of the model processing unit and the occupation time of the channel state information processing unit are closely connected or closely followed.

[0091] In some embodiments, the first report or the second information further includes at least one of the following: third information, tag information corresponding to the third information, fourth information, performance indicators of the function corresponding to the fourth information, the number of model processing units required to generate the fourth information, and a second quantity; the ability of the fourth information to determine the function of the first node by the second node; the second quantity includes the sum of the number of model processing units required to generate the third information and the number of model processing units required to generate the fourth information.

[0092] It should be understood that in some scenarios, the second node needs to determine the capabilities / performance of the functions in the first node. In this case, the first report or the second information may report the fourth information and / or the performance indicators of the functions corresponding to the fourth information. The second node may analyze and judge the performance / capabilities of the functions of the first node based on the fourth information, or it may obtain the performance / capabilities of the functions of the first node through the performance indicators of the functions corresponding to the fourth information.

[0093] Knowing the number of model processing units required to generate the fourth information allows the second node to know the number of model processing units required by the first node to generate the fourth information, facilitating the second node's configuration / scheduling.

[0094] The second quantity allows the second node to know the number of model processing units required by the first node to generate the third and fourth information, facilitating the second node's configuration / scheduling.

[0095] For example, suppose the third information is generated through function 1 of the first node, and the fourth information is obtained by processing the third information through function 2 of the first node or function 3 of the second node. In some scenarios, the first node can generate the third information and send it to the second node, which then generates the fourth information based on the third information. In this case, the first and second nodes collaborate to generate the fourth information. If the second node expects / needs the first node to directly generate the fourth information, the first node can generate the fourth information for the second node to judge, or provide the second node with the performance metrics of the function corresponding to generating the fourth information so that the second node knows the first node's ability to generate the fourth information. Therefore, the second node can schedule the first node to generate the fourth information or choose not to use the first node's method of generating the fourth information.

[0096] In some embodiments, when the number of available model processing units in the first node is greater than or equal to the second number, the first report includes fourth information and / or performance metrics of the function corresponding to the fourth information. Thus, the first node only notifies the second node of the fourth information and / or the performance metrics of the function corresponding to the fourth information when it has the capability to generate the fourth information. This ensures that when the second node configures / schedules the first node to generate the fourth information, the first node can reliably generate the fourth information.

[0097] In some embodiments, when the number of additional available model processing units in the first node is greater than or equal to the number of model processing units required to generate the fourth information, the first report includes the fourth information and / or performance metrics of the function corresponding to the fourth information. In some embodiments, the additional available processing units may include a second type of model processing unit. In this case, when the number of the second type of model processing units is greater than or equal to the number required to generate the fourth information, the first node has the capability to generate the fourth information. In this case, if the second node instructs / schedules / configures the first node to generate the fourth information, the first node can reliably generate the fourth information. In some embodiments, the additional available processing units do not include model processing units for generating the third information, i.e., the first node has reserved model processing units for generating the third information or the model processing units for generating the third information are already occupied. In this case, when the number of additional available model processing units in the first node is greater than or equal to the number of model processing units required to generate the fourth information, the first node has the capability to generate the fourth information. In this case, if the second node instructs / schedules / configures the first node to generate the fourth information, the first node can reliably generate the fourth information.

[0098] In some embodiments, the occupancy trend of the model processing units of the first node in the future time period includes: information on the number of model processing units that need to be occupied in each time unit in the future at least one time unit.

[0099] In some embodiments, the number of model processing units required for each time unit within at least one future time unit satisfies one of the following: an absolute value; the absolute value of the model processing unit corresponding to the initial time unit, and the difference value of the model processing unit corresponding to each time unit other than the initial time unit; the absolute value of the model processing unit corresponding to the initial time unit, and the level index of the range of the difference value of the model processing unit corresponding to each time unit other than the initial time unit.

[0100] Here, the difference value of the model processing unit corresponding to a time unit is obtained by the difference of the absolute value of the model processing unit corresponding to the initial time unit.

[0101] In some embodiments, the first node includes channel state information processing unit resources for generating reports. It should be understood that the first node can assist in generating reports through the channel state processing unit when the number of available model processing units is insufficient.

[0102] In some embodiments, the first information further includes at least one of the following: unoccupied channel state information processing unit resources in the first node at the current time; additionally occupied channel state information processing unit resources; unoccupied channel state information processing unit resources in the first node in a future time period; and the mapping relationship between the channel state information processing unit resources of the first node and the model processing unit of the first node.

[0103] It should be understood that the mapping relationship between the channel state information processing unit resources of the first node and the model processing unit of the first node reflects the capability of the channel state information processing unit resources of the first node relative to the model processing unit of the first node, and can assist the second node in better scheduling and configuring the resources of the first node. For example, assume that the capability of the three channel state information processing units of the first node is equivalent to the capability of one model processing unit of the first node.

[0104] In some embodiments, the first information includes at least one of the following:

[0105] (1) The number of unoccupied model processing units in the first type of model processing unit at the current time.

[0106] (2) The number of unoccupied model processing units in the second type of model processing unit at the current time.

[0107] (3) The sum of the number of unoccupied model processing units in the first type of model processing unit at the current time and the number of unoccupied model processing units in the second type of model processing unit at the current time.

[0108] (4) The number of unoccupied model processing units in the first type of model processing units in the future time period.

[0109] (5) The number of unoccupied model processing units in the second type of model processing units in the future time period.

[0110] (6) The sum of the number of unoccupied model processing units in the first type of model processing units in the future time period and the number of unoccupied model processing units in the second type of model processing units in the future time period.

[0111] In some embodiments, the first type of model processing unit belongs to at least one first unit set, and the second type of model processing unit belongs to at least one second unit set; the first information includes at least one of the following:

[0112] (1) The number of unoccupied model processing units in each first unit set at the current time.

[0113] (2) The number of unoccupied model processing units in each second unit set at the current time.

[0114] (3) The sum of the number of unoccupied model processing units in at least one first unit set at the current time and the number of unoccupied model processing units in at least one second unit set at the current time.

[0115] (4) The number of unoccupied model processing units in each first unit set during the future time period.

[0116] (5) The number of unoccupied model processing units in each first unit set during the future time period.

[0117] (6) The sum of the number of unoccupied model processing units in at least one first unit set in the future time period and the number of unoccupied model processing units in at least one second unit set in the future time period.

[0118] It should be understood that by reporting information related to the second unit set, the second node can obtain information about model processing units that it was not originally aware of, enabling the second node to perform reliable scheduling / configuration.

[0119] In some embodiments, the second type of model processing unit corresponds to at least one set of model processing units; different sets of model processing units correspond to different report types, report content types, report entities, functions, function groups, use cases, features, or feature groups. That is, model processing units can be classified by report type, report content type, report entity, function, function group, use case, feature, or feature group to obtain different sets of model processing units. This allows the model processing units in the first node to be classified according to different needs.

[0120] In some embodiments, the first information includes at least one of the following:

[0121] (1) The number of model processing units supported by each set of model processing units.

[0122] (2) The report type or report content type or report entity or function or function group or use case or feature or feature group corresponding to each model processing unit set.

[0123] (3) Priority of each model processing unit set.

[0124] (4) Resource occupancy rules corresponding to each model processing unit set.

[0125] In some embodiments, different report types, report content categories, report entities, functions, function groups, use cases, features, or feature groups each have different priorities. For example, at least one set of model processing units obtained based on report types may have their respective priorities determined based on the priority of the report type corresponding to each model processing unit set.

[0126] In some embodiments, resource occupancy rules include one of the following:

[0127] Resources in a high-priority set of model processing units are not allowed to be allocated to lower-priority report types, report content types, report entities, functions, function groups, use cases, features, or feature groups. Some or all resources in a low-priority set of model processing units are allowed to be allocated to higher-priority report types, report content types, report entities, functions, function groups, use cases, features, or feature groups. In this way, the capabilities or requirements of the report types, report content types, report entities, functions, function groups, use cases, features, or feature groups corresponding to the high-priority set of model processing units can be guaranteed.

[0128] Resources in the highest-priority set of model processing units cannot be allocated to lower-priority report types, report content types, report entities, functions, function groups, use cases, features, or feature groups. Report types, report content types, report entities, functions, function groups, use cases, features, or feature groups corresponding to lower-priority set of model processing units share some or all of the resources in the lower-priority set of model processing units. In this way, the capabilities or requirements of the report types, report content types, report entities, functions, function groups, use cases, features, or feature groups corresponding to the highest-priority set of model processing units can be guaranteed, allowing report types, report content types, report entities, functions, function groups, use cases, features, or feature groups corresponding to other-priority set of model processing units to share the resources of the model processing units, ensuring flexibility.

[0129] Different priority sets of model processing units can share some or all of the resources in their respective sets of model processing units, whether it be the type of report content, report entity, function, function group, use case, feature, or feature group. This ensures that all priority sets of model processing units can share the resources of the model processing units, guaranteeing flexibility.

[0130] Different priority sets of model processing units, corresponding to different report types, report content categories, report entities, functions, function groups, use cases, features, or feature groups, share some or all resources within the corresponding set of model processing units. The order of resource sharing within a set of model processing units is positively correlated with the number of available resources within that set. This ensures that all priority sets of model processing units can share resources according to the desired sharing order, guaranteeing flexibility. Specifically, the more available resources a set of model processing units contains, the earlier it is shared.

[0131] In some embodiments, the resource occupancy rule further includes: for a first set of model processing units with a priority lower than a priority threshold, if the number of occupied model processing units in the first set of model processing units is less than or equal to a first threshold value, then some model processing units in the first set of model processing units are not activated, or it is determined that some model processing units in the first set of model processing units are not allowed to be occupied. This reduces the number of activated or allowed model processing units, thereby reducing power consumption.

[0132] In some embodiments, the number of shared resources is determined by one of the following methods: pre-configuration; a predetermined value; determination by the first node itself; indication by the second node; or negotiation between the first node and the second node.

[0133] In some embodiments, sending first information to the second node includes: sending first information to the second node based on the relationship between a preset threshold value and the parameters corresponding to the preset threshold value. This allows the reporting of first information to be triggered only when certain conditions are met, reducing the frequency of first information reporting and the amount of information that the second node needs to process.

[0134] In some embodiments, the preset threshold value includes at least one of the following:

[0135] (1) The number threshold value corresponding to the number of available model processing units.

[0136] (2) The ratio threshold value corresponding to the ratio of available model processing units to all model processing units.

[0137] (3) The number threshold value corresponding to the number of occupied model processing units.

[0138] (4) The ratio threshold value corresponding to the ratio of occupied model processing units to all model processing units.

[0139] (5) The number threshold value corresponding to the number of occupied model processing units in the future time period occupancy trend of model processing units.

[0140] (6) The ratio threshold value corresponding to the ratio of occupied model processing units to all model processing units in the future time period occupancy trend.

[0141] (7) The ratio threshold value corresponding to the ratio of occupied model processing units to available model processing units at the current time in the future time period occupancy trend of model processing units.

[0142] (8) The number threshold value corresponding to the number of unoccupied model processing units in the unoccupied trend of model processing units in the future time period.

[0143] (9) The ratio threshold value corresponding to the ratio of unoccupied model processing units to all model processing units in the unoccupied trend of model processing units in the future time period.

[0144] (10) The ratio threshold value corresponding to the ratio of unoccupied model processing units to available model processing units at the current time in the unoccupied trend of model processing units in the future time period.

[0145] In some embodiments, the first information may further include the relationship between the preset threshold value and the parameter corresponding to the preset threshold value;

[0146] The parameters corresponding to the preset threshold value include at least one of the following:

[0147] (1) The number of units that can be processed by the available model.

[0148] (2) The ratio of available model processing units to all model processing units.

[0149] (3) The number of model processing units already occupied.

[0150] (4) The ratio of occupied model processing units to all model processing units.

[0151] (5) The number of model processing units occupied in the future time period.

[0152] (6) The ratio of occupied model processing units to all model processing units in the future time period.

[0153] (7) The ratio of occupied model processing units to available model processing units at the current time in the future time period occupancy trend of model processing units.

[0154] (8) The number of unoccupied model processing units in the unoccupied trend of model processing units in the future time period.

[0155] (9) The ratio of unoccupied model processing units to all model processing units in the unoccupied trend of model processing units in the future time period.

[0156] (10) The ratio of unoccupied model processing units to available model processing units at the current time in the unoccupied trend of model processing units in the future time period.

[0157] The following are exemplary descriptions provided by embodiments of this disclosure:

[0158] Considering that the terminal will perform some AI / ML enhancement functions visible to the base station (e.g., AI-based CSI compression, CSI prediction), and some AI / ML enhancement functions transparent / invisible to the base station (e.g., AI channel estimation), the terminal may use a dedicated MPU to handle model inference reporting for these functions (different from the CPU resources used for traditional CSI processing). If the terminal executes use cases invisible to the base station and consumes MPU resources, the base station, unaware of this behavior, may schedule reports exceeding the terminal's currently available processing resources. This can lead to some scheduled reports failing to provide accurate information. To help the base station achieve efficient and reasonable model scheduling, several solutions are as follows:

[0159] Example 1:

[0160] Assume the terminal has only one MPU pool, and the total number of supported MPUs is N. Since the terminal may perform K model inference reports simultaneously (here, K is an integer greater than or equal to 1), and the total number of MPUs used simultaneously is M = ∑ i K =1 X i Let i = 1, 2, ..., K, where Xi is the number of MPUs used in each model inference. It should be noted that K model inferences can correspond to, but are not limited to, K models, or fewer than K models (i.e., greater than 0 and less than or equal to K). Preferably, K model inferences correspond to K models. K model inferences can also correspond to, but are not limited to, K reports (e.g., CSI reports), or fewer than K reports (i.e., greater than 0 and less than or equal to K). Preferably, K model inferences correspond to K reports. This description also applies to other examples in this document.

[0161] The terminal needs to report at least one of the following information to the base station, or the standard protocol needs to define at least one of the following information, or the base station and the UE need to negotiate at least one of the following information, or the base station needs to indicate at least one of the following information:

[0162] (1) Number of unoccupied MPU resources at the current moment:

[0163] For example, if a UE can support a maximum of N MPUs, and at a certain moment M PPUs are occupied (here, M1 MPUs are occupied in a way visible to the base station, and M2 MPUs are occupied in a way that is transparent to the base station; for example, depending on the UE's implementation, M = M1 + M2), then the UE has NM unoccupied MPUs. Since the base station is unaware that M2 MPUs are also occupied, it considers there to be N-M1 unoccupied MPUs remaining. Therefore, the scheduled L CSI reports may not be fully activated using the L models for inference because the remaining NM MPU resources cannot be fully utilized. The UE can then ignore the scheduling or use of the lower-priority L1 (0 ≤ L1 ≤ L) models among these L models (e.g., not activating L1 models, not switching L1 models, not updating CSI report content related to L1 models, not using L1 models to generate related content, etc.) to ensure that the total number of occupied MPUs at that moment does not exceed N, or to ensure that the total number of occupied MPUs for the L-L1 models does not exceed NM. Here, L1 is the minimum number of MPUs occupied at any given time that does not exceed N, or L-L1 is the maximum number of MPUs occupied at any given time that does not exceed N, or L1 is the minimum number of remaining total MPUs occupied at any given time that does not exceed Nm, or L-L1 is the maximum number of remaining total MPUs occupied at any given time that does not exceed Nm. In a specific example, the priority relationship of the L models is consistent with the priority of the corresponding L CSI reports; for example, beam prediction CSI reports have higher priority than CSI generation reports. In another specific example, the priority order of the L models is consistent with the order of the number of MPUs they occupy. For example, models occupying more MPUs have higher priority because resources need to ensure the normal generation of complex CSI report content. In yet another specific example, the priority order of the L models is reversed compared to the order of the number of MPUs they occupy. For example, models occupying more MPUs have lower priority because resources need to complete the normal generation of as many CSI report contents as possible.

[0164] Based on the above examples, when considering the time-domain characteristics of CSI reports, in a specific instance, the number of MPUs used by periodic / semi-persistent reports should be prioritized to be sufficient and should not be affected by the number of MPUs used by non-periodic reporting inference or UE-side transparent functions. Specifically, when the number of MPUs (PM / SM) used by periodic / semi-persistent reports plus the number of MPUs (AM) used by non-periodic reporting inference is less than the currently available NM resources, then when the content of periodic / semi-persistent reports is normally inferred and reported, the remaining NM-PM / SM MPU resources are allocated to non-periodic reporting processing. If there are insufficient resources to process the reports, abnormal processing is performed.

[0165] In a specific example, the number of MPUs that can be occupied is related to the length of the time unit. Specifically, the longer the time unit, the fewer MPUs can be occupied. For example, the first time slot model allows 5 MPU resources to be occupied, and the second time slot allows 2 MPU resources to be occupied.

[0166] (2) The position index of L1 reports in the total L reports scheduled:

[0167] Because some lower-priority L1 reports may not be able to generate CSI reports properly based on model inference due to insufficient resources, one possibility is that the content of these CSI reports remains the same as the last reported content and is not updated, or traditional CPU resources are used for calculation. However, the base station cannot know which reports are obtained without model inference. Therefore, the UE needs to indicate to the base station which L1 reports did not use model inference. In a specific example, the position index of L1 in L is indicated in the form of a bitmap. For example, at a certain moment, the base station schedules 4 CSI reports that require model inference, but the UE only updates the 2 highest-priority reports. The remaining 2 reports cannot be updated due to insufficient MPU. The UE uses a bitmap indicator 1010 to indicate that the first and third reports are normally using MPU for inference. In a specific example, the position index of L1 in L is indicated using binary encoding, such as dynamic bit width (i.e., B is related to the number of L). (This is for rounding up), or a fixed-bit index. In a specific example, each report uses an extra bit to indicate whether the content was obtained using the MPU; for example, "1" indicates that the report content was obtained using MPU inference, "0" indicates that the report content was obtained without using MPU inference, and vice versa.

[0168] (3) MPU occupancy or trend in the near future:

[0169] Assuming that the UE can predict the possible MPU occupancy or trend in the future based on the transparent use cases of MPU occupancy in the past period, the UE can instruct the base station to provide this information so that the base station can make reasonable and efficient scheduling in the future.

[0170] In a specific example, the UE reports the additional number of MPUs required in the current and / or future T time units, M2 (an integer where T ≥ 0 and M2 ≥ 0). In a specific instance, M2 is an absolute value. For example, the UE reports the additional number of MPUs required in the next 4 time units as 2, 2, 1, 1. In another specific instance, M2 is the absolute value of the current time unit T0 and the difference between the next T-1 time units and time T0. For example, based on the above example, the UE reports the additional number of MPUs required in the next 4 time units as 2, 0, -1, -1. It should be noted that for absolute values, they need to be quantized into binary representations, which can be, but are not limited to, dynamic bit widths (i.e., B is related to the absolute value Abs). (For rounding up), or a fixed-bit-width index (i.e., B is related to the maximum absolute value Abs_max). (This is for rounding up). Similarly, for differential values, they also need to be quantized into binary representations, specifically a 1-bit sign indicator and a B-bit absolute difference indicator. The absolute difference indicator can, but is not limited to, a dynamic bit width (i.e., B is related to the absolute value of the difference value Diff). (For rounding up), or a fixed-bit-width index (i.e., B is related to the absolute value of the largest difference value Diff). (For the rounding up operation), the specific quantization method applies to all use cases of this patent, and will not be elaborated further. Based on the above difference example, the difference value -1 can be indicated as 101 using 3 bits, where 1 represents the negative sign and 01 represents the absolute value of the difference.

[0171] In another specific example, M2 is the index of the absolute value of the current time unit T0 and the differential value range level of the next T-1 time points relative to time T0. Here, the differential value range level can be obtained by looking up a table, which may be, but is not limited to, a table defined by a standard protocol, negotiated between the base station and the UE, configured by the base station for the UE, or indicated by the UE to the base station. The absolute value still needs to be quantized using B bits of binary value. For example, if the UE reports that the additional number of MPUs needed for the next four time points is 2, 3, 2, 1, then the corresponding differential values ​​are 10 (representing the absolute value 2), 1, 0, -1, corresponding to indices 2, 01, 00, 11 in the table below. For example, as shown in Table 1:

[0172] Table 1

[0173] In one example, the UE only reports the trend of the number of additional MPUs used at the current time T0 and / or the number of additional MPUs used in the next T-1 time units (M2) (i.e., it does not show the specific MPU values ​​used), only indicating the relationship between the number of additional MPUs used in the T-1 time unit and the number used at time T0. A specific example uses a bitmap to represent this relationship; for instance, "1" indicates that the number of MPUs used is not less than the number used at time T0, and "0" indicates that the number of MPUs used is less than the number used at time T0. For example, if the UE's additional MPU needs for the next four time units are 2, 3, 2, and 1, then the corresponding reporting indication would be 10 (representing the absolute value 2)|110.

[0174] It should be noted that the above quantification method also applies to other examples in this article, which will not be elaborated upon here.

[0175] Example 2:

[0176] Based on the case of an MPU pool in Example 1, considering that if a large number of UEs use this frequent dynamic reporting of available MPU resources, it will bring huge challenges to the service complexity of UE scheduling and management on the base station side, the following method can be considered to solve this problem:

[0177] The base station and the UE need to negotiate at least one of the following information, or the base station needs to indicate at least one of the following information, or the standard protocol needs to define at least one of the following information, to trigger the terminal to report the relevant MPU occupancy status to the base station:

[0178] An MPU occupancy threshold includes at least one of the following:

[0179] (1) Threshold for the number of remaining available MPUs: THR#1 (threshold, THR)).

[0180] When the remaining available MPUs NM of the UE are less than or equal to THR#1, the UE may, but is not limited to, report a feedback value to the base station, i.e., a 1-bit reply information (e.g., "1" indicates below or not above the threshold, "0" indicates not below or above the threshold). Furthermore, the base station can perform secondary configuration based on the reply information, configuring the UE to report the relevant number of MPUs in use (which can be considered in conjunction with other examples); the remaining available MPUs NM reported to the base station; the total number of currently used MPUs M reported to the base station; the number of additionally used MPUs M2 reported to the base station; and the number of differential MPUs below the threshold reported to the base station, i.e., THR-N+M. This example applies to other threshold values, which will not be elaborated upon further.

[0181] (2) The percentage threshold THR#1' of the remaining available MPUs is used to determine whether to trigger reporting. That is, compare (NM) / N with the threshold THR#1'.

[0182] For example: 30%, 20%, 15%, etc.

[0183] (3) The threshold value THR#2 of the total number of MPUs occupied is used to determine whether to trigger reporting.

[0184] (4) The percentage threshold THR#2' of the total number of MPUs currently in use relative to the total number of MPUs supported by the UE, i.e., comparing M / N with the threshold THR#2' to determine whether to trigger reporting.

[0185] For example: 70%, 80%, 85%, etc.

[0186] (5) Additional MPU number threshold value THR#3, that is, compare M2 with the threshold value THR#3 to determine whether to trigger reporting.

[0187] (6) The percentage threshold THR#3' of the number of additional MPUs occupied relative to the total number of MPUs supported by the UE, i.e., comparing M2 / N with the threshold THR#3' to determine whether to trigger reporting.

[0188] For example: 30%, 20%, 15%, etc.

[0189] (7) The percentage threshold THR#3” of the number of extra MPUs occupied relative to the number of NM remaining in the UE at the current time, that is, comparing M2 / (NM) with the threshold THR#3” to determine whether to trigger reporting.

[0190] For example: 70%, 60%, 50%, etc.

[0191] For example, the percentage threshold of the number of additional MPUs occupied M2 to the number of MPUs occupied M1 for non-transparent processing at the base station. That is, compare M2 / M1 with the threshold value. To determine whether a report has been triggered.

[0192] It should be noted that when the MPU occupancy on the UE side is below the threshold, there is no need to report MPU-related information. The base station only needs to assume that the UE is capable of processing the scheduling report and does not need to change the scheduling policy.

[0193] Example 3:

[0194] Considering that if a terminal has only one MPU pool, all processing involving the model will compete for MPU resources, and the base station may encounter situations where the scheduling report content is abnormal because the base station cannot know that the UE is running a transparent function that consumes some resources. However, dividing one MPU pool into P pools (P being an integer greater than 1) may solve the problem of competing for resources.

[0195] Assume the terminal has P MPU pools (P being an integer greater than 1), and the total number of supported MPUs is N. Then, the total number of MPUs in each pool is {N1, N2, ..., NP}, where N1 + N2 + ... + NP = N. Since the terminal may simultaneously perform K model inference reports (where K is an integer greater than or equal to 1), and the total number of MPUs used simultaneously... Here, Qj is the total number of MPUs used for model inference in the j-th pool (where Q is an integer greater than or equal to 0).

[0196] Preferably, in one example, the terminal has two MPU pools, including a first MPU pool and a second MPU pool (here, "first" and "second" are only used to distinguish different names and do not involve any order, priority or other factors).

[0197] In a concrete example, the MPU resources in the first MPU pool are not transparent to the base station (i.e., the MPU resources that the base station can schedule), while the MPU resources in the second MPU pool are from an extended MPU pool, meaning the base station cannot schedule MPU resources unless indicated by the UE. When the remaining available MPU resources in the first MPU pool are less than the resources required for scheduling reports / models, the available resources in the second MPU pool can be added to the first pool as additional MPU resources to complete normal model inference.

[0198] The terminal needs to report at least one of the following information to the base station, or the standard protocol needs to define at least one of the following information, or the base station and the UE need to negotiate at least one of the following information, or the base station needs to indicate at least one of the following information:

[0199] (1) The number of unoccupied MPUs in the first MPU pool at the current time.

[0200] (2) The percentage of unoccupied MPUs in the first MPU pool at the current time relative to the total number of MPUs in that pool.

[0201] (3) The number of unoccupied MPUs in the second MPU pool at the current time.

[0202] (4) The percentage of the number of MPUs available in the second MPU pool at the current time relative to the total number of MPUs in the first MPU pool.

[0203] (5) The total number of MPUs that can be occupied in the first MPU pool and the second MPU pool at the current time.

[0204] (6) The number of MPUs available in the first MPU pool in the next T time units.

[0205] (7) The percentage of unoccupied MPUs in the first MPU pool in the next T time units relative to the total number of MPUs in that pool.

[0206] (8) The minimum / maximum / average / weighted average number of MPUs occupied in the first MPU pool in the next T time units (the weight is related to the time unit index, for example, it decreases as the time unit index increases).

[0207] (9) The number of MPUs available in the second MPU pool in the next T time units.

[0208] (10) The percentage of the number of MPUs available in the second MPU pool in the next T time units relative to the total number of MPUs in the first MPU pool.

[0209] (11) The minimum / maximum / average / weighted average number of MPUs available in the second MPU pool in the next T time units (the weight is related to the time unit index, for example, it decreases as the time unit index increases).

[0210] (12) The total number of MPUs that can be occupied in the first MPU pool and the second MPU pool in the next T time units.

[0211] (13) The minimum / maximum / average / weighted average of the total number of MPUs available in the first MPU pool and the second MPU pool in the next T time units (the weight is related to the time unit index, for example, it decreases as the time unit index increases).

[0212] Example 4:

[0213] Based on the assumptions of Example 3, assume the terminal has P MPU pools (P being an integer greater than 1). These P MPU pools are associated with functions / use cases / report types / report content categories or report entities; that is, different pools are used only to process reports for their corresponding functions. The MPU occupancy method in different pools is as follows:

[0214] Assume that there are k MPU pools that are not transparent to the base station, meaning that the base station can schedule the UE to use MPU resources from the k pools.

[0215] The terminal needs to report at least one of the following information to the base station, or the standard protocol needs to define at least one of the following information, or the base station and the UE need to negotiate at least one of the following information, or the base station needs to indicate at least one of the following information:

[0216] (1) The total number of MPUs supported by each MPU pool is {N1, N2, ..., Nk}.

[0217] (2) Which different use cases / functions / features / feature groups / report types are the MPU resources in different pools used for?

[0218] For example, the first MPU pool is used to process CSI and beam-related reports, while the second MPU pool is used to process location-related reports.

[0219] Here, the classification of different use cases / functions / features / feature groups / report types can be based on different priorities. For a specific example, the priority is inversely proportional to the report latency; for example, reports with low latency have higher priority than reports with high latency.

[0220] (3) Priority information for different MPU pools:

[0221] For example, if the first MPU pool has more resources and stronger processing capabilities, while the second MPU pool has fewer resources and weaker processing capabilities, then the first pool has a higher priority than the second pool. Therefore, the base station can prioritize allocating resources from the first MPU pool to process reports.

[0222] (4) MPU resource usage rules for different MPU pools:

[0223] In a specific example, resources from the MPU pool handling high-priority functions are not allowed to be allocated to low-priority function processing, while Z MPU resources from the MPU pool handling low-priority functions are allowed to be additionally allocated to high-priority function processing. Here, Z is an integer greater than 0 and less than or equal to the total number of MPUs supported by that pool. It should be noted that the value of Z can be indicated by the UE to the base station, by the base station to the UE, by negotiation between the two parties, or by a value defined in the protocol.

[0224] For example, suppose the first MPU pool is used to process CSI reports, and the second MPU pool is used to process location reports. If the first MPU pool is one MPU short of processing a CSI report, it can borrow an MPU from the second MPU pool to update the report content. Specifically, if there are no spare MPU resources in the second pool, no MPU will be borrowed.

[0225] In a specific example, the resources of the MPU pool that handles the highest priority function are not allowed to be allocated to lower priority function processing, while Z MPU resources of the MPU pool that handles lower priority functions are allowed to be allocated to function processing with high / low / same priority, where Z is an integer greater than 0 and less than or equal to the total number of MPUs supported by the pool.

[0226] In a specific example, resources of MPU pools that handle functions of different priorities can be shared among themselves, but Z MPUs are shared among MPU pools that handle functions from low to high priority.

[0227] In a specific example, resources of MPU pools that handle different priority functions can be shared with each other, but Z MPUs are shared among pools in descending order of the number of remaining available MPUs.

[0228] In a specific example, Z MPUs are shared in pools ordered from low to high based on the number of use cases / features / feature groups / report types that simultaneously consume MPU processing within each pool.

[0229] Based on the above example, further, within t time units, if the number V of occupied MPUs in the MPU pool is greater than or equal to a certain threshold value Threshold, then the resources in that MPU pool will not be allocated.

[0230] Here, the threshold value (Threshold) can be indicated by the UE to the base station, by the base station to the UE, by negotiation between the two parties, or as defined in the protocol. For example, Threshold = 3 / 4 * Ni or 2 / 3 * Ni, etc., where Ni is the total number of MPUs supported by the i-th pool.

[0231] Considering the energy-saving issue on the UE side, if some MPU resources of the UE remain idle and unoccupied for a period of time but are still running, they will consume power due to continuous operation. Therefore, in a specific example, within t time units, if the number V of occupied MPUs in the lower priority MPU pool is less than or equal to a certain threshold, then some MPU resources are shut down to save energy, or some MPU resources are not allowed to be occupied, where t is an integer greater than 0.

[0232] Here, the threshold value can be indicated by the UE to the base station, by the base station to the UE, by negotiation between the two parties, or as defined in the protocol. For example: Ni is the total number of MPUs supported by the i-th pool.

[0233] It should be noted that the MPU pool resource usage solutions mentioned in the article can also be applied to, but are not limited to, traditional CPU pools, and potentially future memory processing unit pools.

[0234] Example 5:

[0235] This example primarily considers the impact on model inference time that may result from hardware resource preemption, i.e., inference time increases when resources are scarce.

[0236] Assume the terminal may simultaneously perform K model inference reports (where K is an integer greater than or equal to 1), and the time units occupied simultaneously are {t1, t2, ..., tK}, typically the number of OFDM symbols. Specifically, the timeline occupied under aperiodic reporting can be from the first symbol of the PDCCH to the last symbol of the scheduled bearer report PUSCH, or under periodic / semi-persistent reporting, from the first symbol of the earliest reference signal resource in the most recent CSI-RS / CSI-IM / SSB before the triggered CSI reference resource to the last symbol of the configured report-carrying PUSCH / PUCCH.

[0237] When the remaining available MPU resources are less than the MPU resources required for the scheduled L reports, the terminal needs to report at least one of the following information to the base station, or the standard protocol needs to define at least one of the following information, or the base station and UE need to negotiate at least one of the following information, or the base station needs to indicate at least one of the following information:

[0238] (1) All reports have an added time unit offset value Δt:

[0239] For example, the normal model inference timeline occupies t1 OFDM symbols. Due to limited hardware resources, all reports are increased by an offset value Δt, that is, the total time occupied is t1+Δt.

[0240] (2) Increased time unit offset value Δt' for some low-priority reports:

[0241] For example, due to limited resources, but in order to ensure that high-priority reports occupy the model inference timeline normally, i.e., occupy t1 OFDM symbols, this causes low-priority reports to have an additional offset value Δt', i.e., the total time occupied is t2+Δt'.

[0242] Furthermore, because some low-priority reporting timelines have become longer, the UE also needs to indicate the reporting index occupied by the abnormal timelines. The index indication can be in bitmap format or binary value representation; see Example 1 for a specific example.

[0243] (3) The time unit offset value Δs added during model switching:

[0244] For example, a UE may have multiple models when implementing the same function / use case / report type / report content. Different models may use different numbers of MPUs due to their varying complexity. When MPU resources are insufficient, switching to a smaller model can be considered to reduce the number of MPU resources used. However, the accuracy of the inference results may be lower than that of the complex model. Therefore, when a model switch occurs, the inference timeline length needs to be increased by an additional Δs.

[0245] Considering that if a model exists in different storage locations on the terminal, it can also be mapped to different states (e.g., whether the model is inactive, active, or inference state), the different delays Δs for loading / switching a new model are bound to different states, i.e., they correspond to different values.

[0246] It should be noted that the time unit used in the normal model inference timeline can be, but is not limited to, Z / Z' as defined in the TS38.214 protocol.

[0247] Example 6:

[0248] Considering the issue of MPUs coexisting with traditional CPUs, when MPU resources are insufficient, it is advisable to allocate some resources from the CPU pool for its use.

[0249] In a concrete example, suppose there is one MPU pool and one CPU pool. If the currently available MPU resources are insufficient to fully schedule L model inferences, then some of the available O CPU resources can be drawn from the CPU pool, where O is an integer greater than or equal to 1. It should be noted that borrowing MPU resources also applies when CPU resources are insufficient.

[0250] The terminal needs to report at least one of the following information to the base station, or the standard protocol needs to define at least one of the following information, or the base station and the UE need to negotiate at least one of the following information, or the base station needs to indicate at least one of the following information:

[0251] (1) The number of MPU resources that are not currently in use.

[0252] (2) The number of MPU resources that will not be used in the future.

[0253] (3) The number of CPU resources that are not currently in use.

[0254] (4) The number of additional CPU resources used.

[0255] (5) The number of CPU resources that will not be used in the future.

[0256] (6) Mapping relationship between CPU and MPU.

[0257] It should be noted that the number of resources includes at least one of the following: an absolute value, a percentage of the total number of CPUs used, etc.

[0258] Regarding the mapping relationship between CPU and MPU, for example, an MPU may be a more powerful dedicated processor, while a CPU is a less powerful traditional general-purpose processor. Therefore, if a CPU is introduced to handle the corresponding functions / reports, the mapping relationship between the two needs to be defined, that is, 1 MPU usage is equivalent to x CPU usage, for example, x = 2, 3, etc.

[0259] When MPU resources are insufficient to borrow CPU usage, in order to ensure that reports processed using traditional CPU usage are not affected, some rules can be preset or negotiated between the parties:

[0260] In a concrete example, when available CPU resources are insufficient to support the MPU resources required to support the model, additional CPU resources are not allowed to be called to process reports that cannot reason normally due to insufficient MPU resources. For example, when available CPU resources are less than or equal to a certain threshold, additional CPU resources are not allowed to be called to process these reports. Here, the threshold can be indicated by the UE to the base station, by the base station to the UE, by negotiation between the two parties, or by a value defined in the protocol. For example:

[0261] For some existing use cases, it may be necessary to utilize both CPU and MPU simultaneously. For a specific example, in the CSI compression use case, the channel measurement and preprocessing parts are the same as traditional methods and can be handled using CPU. However, CSI compression uses an AI model, which consumes MPU. Therefore, how to rationally schedule and plan these two components requires further consideration. Assume that a certain function / use case / report type / report content requires 1 CPU and 2 MPUs respectively.

[0262] In a specific example, if the number of resources in at least one of the CPU and MPU is insufficient, the UE will ignore the scheduling or use of this model (e.g., not activating the model, not switching the model, not updating the CSI report content related to the model, not using the model to generate related content, etc.). In another specific example, if the number of resources in at least one of the CPU and MPU is insufficient, the UE can additionally call another available resource to assist in completing inference, provided that there are spare resources available for use.

[0263] It should be noted that for reports of certain special functions (excluding reports involving UE-side model inference), such as training data collection, model inference, and performance monitoring in base station-side model scenarios, since model inference is performed on the base station side, the UE side only needs to report some measurement values. Alternatively, for training data collection and performance monitoring in UE-side model scenarios, it may, but is not limited to, occupying only CPU resources or only MPU resources. For occupying only MPU resources, a specific example is that the number of MPU resources occupied, O_MPU, is less than the number of CPU resources, O_CPU, where O_MPU is a real number greater than 0. It should be noted that a real number less than 1 means that one MPU can complete the processing of multiple of the above reports; for example, O_MPU = 0.5, meaning one MPU can complete the processing of two special function reports.

[0264] Example 7:

[0265] For some use cases, the UE side may jointly run m models to obtain the content of the report, where m is an integer greater than or equal to 1. Assume that the number of MPUs occupied by the m models is {MPU_1, MPU_2, ..., MPU_m} respectively, and the number of MPUs occupied by the inference of the m models is MPU_total = MPU_1 + MPU_2 + ... + MPU_m. Typically, m=2.

[0266] In a specific example, the m models are regarded as a whole, that is, when the remaining number of MPUs N-M on the UE side is less than MPU_total, none of the m models can be activated, that is, the scheduling or use of these m models is ignored (for example, the models are not activated, the models are not switched, the CSI report content related to the models is not updated, the models are not used to generate related content, or the system switches back to the conventional method to occupy the CPU for processing, etc.).

[0267] In another example, the m models are regarded as m independent modules, that is, when the remaining number of MPUs N-M on the UE side is less than MPU_total, but N-M is greater than the sum MPU_total' of the MPUs occupied by m' models in {MPU_1, MPU_2, ..., MPU_m}, that is, 0<m ′ <m, then the MPUs can be occupied to run the m' models therein, but the MPUs cannot be occupied to run the remaining m-m' models, that is, the scheduling or use of the m' models is ignored (for example, the models are not activated, the models are not switched, the CSI report content related to the models is not updated, the models are not used to generate related content, or the system switches back to the conventional method to occupy the CPU for processing, etc.).

[0268] It should be noted that, here, m' is the maximum number that the total number of occupied MPUs at this moment does not exceed N, or m-m' is the minimum number that the total number of occupied MPUs at this moment does not exceed N, or m' is the maximum number that the remaining total number of occupied MPUs at this moment does not exceed N-m, or m-m' is the minimum number that the remaining total number of occupied MPUs at this moment does not exceed N-m.

[0269] Typically, applicable functions / use cases / features / feature groups may be, but are not limited to, a combination of at least one of the following modules: AI-based channel estimation, channel equalization, CSI prediction, CSI compression, source coding, channel coding, modulation, etc.

[0270] For the timeline, considering that some modules use AI models to replace traditional algorithms, an additional time unit offset value Δm needs to be introduced based on the conventional timeline. The offset value may be indicated by the UE to the base station, or indicated by the base station to the UE, or a result negotiated by both parties, or predefined in the protocol. For example, Δm is y OFDM symbols, which is associated with the model complexity and depends on the UE's capability report.

[0271] For some use cases, the UE side may run n models to implement the functions of m modules and obtain the reported content, where n is an integer, 1≤n<m. Assume that the number of MPUs occupied by the n models is {MPU_1, MPU_2, ..., MPU_n} respectively, and the number of MPUs occupied by the inference of the n models is MPU_total2 = MPU_1 + MPU_2 + ... + MPU_n. Typically, n=1.

[0272] In a specific example, MPU_total2 is less than MPU_total, or the MPU resources occupied by MPU_total2 are equal to MPU_total minus an offset value. Here, the offset value may be indicated by the UE to the base station, or indicated by the base station to the UE, or a result negotiated by both parties, or predefined in the protocol. The offset value is associated with the saved model complexity and depends on the UE's capability report.

[0273] For the timeline, considering that one model implements the functions of multiple modules mentioned above, an additional time unit offset value Δm' needs to be subtracted from the timeline based on the above example. The offset value may be indicated by the UE to the base station, or indicated by the base station to the UE, or a result negotiated by both parties, or predefined in the protocol. For example, Δm' is z OFDM symbols, which is associated with the saved model complexity and depends on the UE's capability report.

[0274] Further, a typical scenario here is a bilateral model performance monitoring scenario (that is, the first model and the second model deployed on the UE and the base station cooperate to complete the inference).

[0275] Exemplarily, as shown in FIG. 3, which is a schematic flowchart of bilateral model performance monitoring provided according to some embodiments, the process includes inputting information 1 into a first model to obtain information 2; inputting information 2 into a second model to obtain information 3; inputting the output of the first model into a third model to obtain information 4. Here, the first model and the third model are terminal-side models, and the second model is a base station-side model.

[0276] For base station-side performance monitoring, the terminal needs to report model output information 2 and high-precision information 1 as tags (e.g., information 1 or high-precision codebook quantized information 1). In a specific instance, the model output information and high-precision tag information reported by the terminal are reported separately, i.e., the former occupies MPU resources, and the latter occupies CPU resources. Specifically, if a model inference report and a performance monitoring report are triggered in the same time unit, the available MPU and CPU resources of the current UE must simultaneously meet the MPU and CPU resource requirements of both reports; otherwise, these two reports are ignored (e.g., not reported, reported invalid information such as all 0s or all 1s, or not updating the CSI report content related to this model, or not using this model to generate related content, etc.). If the available resources meet at least one of the requirements, the report that does not meet the resource requirement is ignored. Furthermore, the base station can configure an offset value for the UE. This offset value indicates the starting position of the time unit occupying the MPU's resources. It can be consistent with the start time of CPU occupancy, or it can be an offset time unit position after the start time unit of CPU occupancy. Here, it can be, but is not limited to, the last time unit occupied by the CPU, or the time unit after the last time unit occupied by the CPU (i.e., the two are closely related). This offset value can also indicate that the starting position of the time unit occupying the MPU's resources is an offset time unit position after the last time unit occupied by the CPU. Specifically, this offset value can be carried in higher-layer signaling or physical-layer signaling, such as DCI signaling.

[0277] In a specific instance, if the model output information and high-precision tag information reported by the terminal are reported using the same report, then this CSI report will simultaneously occupy the MPU and CPU. Here, the MPU's occupation time can start from, but is not limited to, the last symbol of the DCI / PDCCH, or be based on an offset value, such as z OFDM symbols. The offset value can be pre-configured, pre-defined, configured by the base station to the UE, indicated by the UE to the base station, a result negotiated between the two parties, or defined in the protocol.

[0278] For terminal-side performance monitoring, it is assumed that the terminal uses a third model to assist in recovering information 4. It should be noted that the third model can be the same as or different from the second model. The output information 4 should be as similar as possible to information 3. The performance monitoring index between information 4 and information 1 is calculated to determine whether the performance of the AI ​​model meets the requirements.

[0279] In a specific example, if the third model is not transparent to the base station (i.e., it requires scheduling and management by the base station), the actual number of MPUs used by the UE when calculating performance monitoring metrics is the sum of the number used by the first and third models, U. When the UE is triggered to report performance monitoring results, it is necessary to ensure that the number of available MPUs is greater than or equal to the required number U; otherwise, the report should be ignored, or other monitoring methods and reporting content should be triggered.

[0280] In a specific example, if the third model is invisible to the base station—that is, it's a model implemented internally on the UE side, unknown to the base station and not required to manage it—the actual number of MPUs used by the UE side during performance monitoring metric calculation remains the sum of the first and third models's usage, U'. When the UE is triggered to report performance monitoring results, it needs to report the value of U' required for the UE-side model monitoring metric calculation for subsequent scheduling by the base station, or report the additional number of MPUs required by the third model, depending on the UE's capabilities. The UE-side model monitoring performance metric calculation must ensure that the number of available MPUs is greater than or equal to the required number U'; otherwise, the report is ignored, or other monitoring methods and reporting content are triggered.

[0281] For UE-side performance monitoring scenarios, the overall time unit length T_monitoring adds a third-model related calculation process compared to the time unit length T_inference occupied by model inference. In a specific example, an offset value δp is added to the time unit length T_inference occupied by inference, i.e., T_monitoring = T_inference + δp. The offset value δp may be related to the UE's capabilities, and may be configured by the base station for the UE, indicated by the UE to the base station, or be the result of negotiation between the two parties, or defined in the protocol.

[0282] In another example, the UE first feeds back information 2, compressed from the first model, to the base station. The base station then recovers information 3 using the second model, applies information 3 to a reference signal (e.g., CSI-RS, DMRS), and sends the applied reference signal back to the UE. The UE measures this reference signal and a normal reference signal without information 3 to obtain the fifth and sixth information, and calculates the relevant performance monitoring indicators. In one example, this process can be viewed as a two-step operation: the UE feeding back information 2 requires MPU resources, and the UE measuring the reference signal and calculating the relevant performance indicators requires CPU resources. Considering that the two steps are a whole, the UE needs to report the number of MPU resources and CPU resources used by this method. The monitoring results of this model can only be reported when the available MPU resources and CPU resources are both greater than the MPU and CPU resources required for monitoring; otherwise, the report is ignored, or other monitoring methods and reporting content are triggered.

[0283] This example may be applicable, but is not limited to, the following use cases: AI-based downlink CSI compression feedback, AI-based CSI feedback with joint source-channel coding, AI-based CSI feedback with joint source-channel coding and modulation, and AI-based uplink SRS channel-assisted downlink CSI feedback in frequency division duplex (FDD) or time division duplex (TDD) scenarios.

[0284] It should be noted that the configuration signaling and indication signaling in this article can be physical layer signaling (e.g., DCI or UCI) or higher layer signaling (e.g., RRC or MAC CE).

[0285] It should be noted that the instructions / reports mentioned in the text may be, but are not limited to, UE capability report reporting, RRC reconfiguration report reporting, UE assistance information (UAI) report reporting, UE applicability report reporting, or CSI report reporting, etc.

[0286] It should be noted that different examples can be combined to achieve a complete solution.

[0287] The communication method provided in this embodiment can be applied to the second node 102 in the communication system shown in FIG1. ​​FIG4 shows a flowchart of another communication method, which includes the following steps S401-S402:

[0288] S401, Send configuration information to the first node.

[0289] The configuration information is used by the first node to send the first report.

[0290] S402, Receive the first information from the first node.

[0291] The first piece of information is used to indicate information related to the model processing unit.

[0292] In some embodiments, the model processing unit of the first node includes a first type of model processing unit and a second type of model processing unit; the relevant information of the first type of model processing unit has been reported to the second node, while the relevant information of the second type of model processing unit has not been reported to the second node.

[0293] In some embodiments, the first information includes at least one of the following: the number of unoccupied model processing units in the model processing units of the first node at the current time, the number of occupied model processing units in the model processing units of the first node at the current time, the occupancy trend of the model processing units of the first node in a future time period, and the unoccupied trend of the model processing units of the first node in a future time period.

[0294] In some embodiments, the method further includes:

[0295] Receive the first report reported by the first node; the first report is obtained based on the model processing unit of the first node.

[0296] In some embodiments, if the number of model processing units available in the first node is less than the number of model processing units required for the report reported by the configuration information scheduling, the first report includes a portion of the report or a portion of the report content reported by the configuration information scheduling; or the first report includes a portion of the updated report content reported by the configuration information scheduling; or the first report includes a portion of the unupdated report content; or the first report includes a portion of the report content obtained based on the model processing units; or the first report includes a portion of the report content not obtained based on the model processing units.

[0297] In some embodiments, the report or report content in the first report is determined based on the priority of the report or report content; the priority is determined based on at least one of the following: the index of the report or the index of the report content, the type of the report or report content, the function or functional group of the report or report content, the use case of the report or report content, the feature or feature group of the report or report content, and the number of model processing units required by the report or report content.

[0298] In some embodiments, the method further includes:

[0299] Receive second information reported by the first node; the second information includes at least one of the following:

[0300] Location information of partial reports or partial report content, indication information of whether each report in the first report is based on the model processing unit, index of reports with abnormal content in the first report, index of reports with abnormal timeline occupation in the first report, time unit offset corresponding to reports with abnormal timeline occupation in the first report, and time unit offset caused by model switching.

[0301] In some embodiments, the second information is included in part or all of the first report.

[0302] In some embodiments, if the number of model processing units available in the first node is less than the number of model processing units required for multiple related functions, the report content corresponding to multiple functions in the first report is not updated; or the report content corresponding to multiple functions in the first report is not obtained based on model processing units; or the report content corresponding to multiple functions in the first report is obtained based on a first number of model processing units, where the first number is less than the number of model processing units required for multiple functions.

[0303] In some embodiments, when the report content corresponding to multiple functions in the first report is obtained based on a first number of model processing units, the first number is determined based on the number of model processing units required for multiple functions and a first offset; the first offset is determined based on one of the following: predefined or preconfigured; determined by the first node; indicated by the second node to the first node; or determined through negotiation between the first node and the second node.

[0304] In some embodiments, the first report includes third information and tag information corresponding to the third information; the third information is generated based on the model processing unit, and the tag information is generated based on the channel state information processing unit; the occupancy start time of the model processing unit that generates the third information is the same as the occupancy start time of the channel state information processing unit that generates the tag information, or is determined based on the occupancy start time of the channel state information processing unit that generates the tag information and a second offset, or is the last time unit of the occupancy time of the channel state information processing unit that generates the tag information, or is determined based on the last time unit of the occupancy time of the channel state information processing unit that generates the tag information and a third offset value.

[0305] In some embodiments, the first report or the second information further includes at least one of the following: third information, tag information corresponding to the third information, fourth information, performance indicators of the function corresponding to the fourth information, the number of model processing units required to generate the fourth information, and a second quantity; the ability of the fourth information to determine the function of the first node by the second node; the second quantity includes the sum of the number of model processing units required to generate the third information and the number of model processing units required to generate the fourth information.

[0306] In some embodiments, when the number of available model processing units in the first node is greater than or equal to the second number, the first report includes fourth information and / or performance metrics of the function corresponding to the fourth information.

[0307] In some embodiments, the occupancy trend of the model processing units of the first node in the future time period includes: information on the number of model processing units that need to be occupied in each time unit in the future at least one time unit.

[0308] In some embodiments, the number of model processing units required for each time unit satisfies one of the following: an absolute value; the absolute value of the model processing unit corresponding to the initial time unit, and the difference value of the model processing units corresponding to each time unit other than the initial time unit; the absolute value of the model processing unit corresponding to the initial time unit, and the level index of the range of the difference values ​​of the model processing units corresponding to each time unit other than the initial time unit; here, the difference value of the model processing unit corresponding to a time unit is obtained by differing from the absolute value of the model processing unit corresponding to the initial time unit;

[0309] In some embodiments, the first node includes channel state information processing unit resources for generating reports.

[0310] In some embodiments, the first information further includes at least one of the following:

[0311] The unused channel state information processing unit resources in the first node at the current moment;

[0312] Additional channel state information processing unit resources occupied;

[0313] Unused channel state information processing unit resources in the first node during the future time period;

[0314] The mapping relationship between the channel state information processing unit resources of the first node and the model processing unit of the first node.

[0315] In some embodiments, the first information includes at least one of the following:

[0316] The number of unoccupied model processing units in the first type of model processing unit at the current moment;

[0317] The number of unoccupied model processing units in the second type of model processing units at the current moment;

[0318] The sum of the number of unoccupied model processing units in the first type of model processing units at the current time and the number of unoccupied model processing units in the second type of model processing units at the current time;

[0319] The number of unoccupied model processing units in the first type of model processing units within the future time period;

[0320] The number of unoccupied model processing units in the second type of model processing units within the future time period;

[0321] The sum of the number of unoccupied model processing units in the first type of model processing units during the future time period and the number of unoccupied model processing units in the second type of model processing units during the future time period.

[0322] In some embodiments, the first type of model processing unit belongs to at least one first unit set, and the second type of model processing unit belongs to at least one second unit set; the first information includes at least one of the following:

[0323] The number of unoccupied model processing units in each first unit set at the current moment;

[0324] The number of unoccupied model processing units in each second unit set at the current moment;

[0325] The sum of the number of unoccupied model processing units in at least one first unit set at the current time and the number of unoccupied model processing units in at least one second unit set at the current time;

[0326] The number of unused model processing units in each first unit set within the future time period;

[0327] The number of unused model processing units in each first unit set within the future time period;

[0328] The sum of the number of unoccupied model processing units in at least one first unit set within a future time period and the number of unoccupied model processing units in at least one second unit set within a future time period.

[0329] In some embodiments, the second type of model processing unit corresponds to at least one set of model processing units; different sets of model processing units correspond to different report types or types of report content or report entities or functions or function groups or use cases or features or feature groups.

[0330] In some embodiments, the first information includes at least one of the following:

[0331] The number of model processing units supported by each set of model processing units;

[0332] Each set of model processing units corresponds to a report type, report content category, report entity, function, function group, use case, feature, or feature group;

[0333] The priority of each model processing unit set;

[0334] Resource usage rules corresponding to each model processing unit set.

[0335] In some embodiments, different report types, report content categories, report entities, functions, function groups, use cases, features, or feature groups have different priorities.

[0336] In some embodiments, resource occupancy rules include one of the following:

[0337] Resources in a high-priority set of model processing units are not allowed to be allocated to a low-priority function, while some or all of the resources in a low-priority set of model processing units are allowed to be allocated to a high-priority function.

[0338] Resources in the highest priority model processing unit set are not allowed to be allocated to lower priority functions. Functions corresponding to the non-highest priority model processing unit set share some or all of the resources in the non-highest priority model processing unit set.

[0339] The functions corresponding to model processing unit sets with different priorities share some or all of the resources in their respective model processing unit sets;

[0340] Different priority sets of model processing units share some or all of the resources in the corresponding set of model processing units; the order of resource sharing in the set of model processing units is positively correlated with the number of available resources included in the set of model processing units.

[0341] In some embodiments, the resource occupancy rules also include:

[0342] For a first set of model processing units with a priority lower than the priority threshold, if the number of occupied model processing units in the first set of model processing units is less than or equal to the first threshold value, then some model processing units in the first set of model processing units will not be activated, or it will be determined that some model processing units in the first set of model processing units are not allowed to be occupied.

[0343] In some embodiments, the number of shared resources is determined in one of the following ways:

[0344] The first node is determined by itself;

[0345] Second node indication;

[0346] The first node and the second node negotiate and determine the outcome.

[0347] In some embodiments, the first node sends first information to the second node based on the relationship between a preset threshold value and the parameters corresponding to the preset threshold value.

[0348] In some embodiments, the preset threshold value includes at least one of the following:

[0349] The number threshold value corresponding to the number of available model processing units;

[0350] The ratio threshold value corresponding to the ratio of available model processing units to all model processing units;

[0351] The number threshold value corresponding to the number of occupied model processing units;

[0352] The ratio threshold value corresponding to the ratio of occupied model processing units to all model processing units;

[0353] The number threshold value corresponding to the number of occupied model processing units in the future time period's model processing unit occupancy trend;

[0354] The ratio threshold value corresponding to the ratio of occupied model processing units to all model processing units in the future time period's model processing unit occupancy trend;

[0355] The ratio threshold value corresponding to the ratio of occupied model processing units to available model processing units at the current time in the future time period occupancy trend;

[0356] The number threshold value corresponding to the number of unoccupied model processing units in the unoccupied trend of model processing units in the future time period;

[0357] The ratio threshold value corresponding to the ratio of unoccupied model processing units to all model processing units in the unoccupied trend of model processing units in the future time period;

[0358] The threshold value corresponding to the ratio of unoccupied model processing units to available model processing units at the current time in the future unoccupied trend of model processing units.

[0359] In some embodiments, the first information may further include the relationship between the preset threshold value and the parameter corresponding to the preset threshold value;

[0360] The parameters corresponding to the preset threshold value include at least one of the following:

[0361] The number of available model processing units;

[0362] The ratio of available model processing units to all model processing units;

[0363] The number of model processing units already in use;

[0364] The ratio of occupied model processing units to all model processing units;

[0365] The number of model processing units occupied in the future time period;

[0366] The proportion of occupied model processing units to all model processing units in the future time period;

[0367] The ratio of occupied model processing units to available model processing units at the current time in the future time period occupancy trend;

[0368] The number of unoccupied model processing units in the future time period's unoccupied trend;

[0369] The ratio of unoccupied model processing units to all model processing units in the future time period's unoccupied trend;

[0370] The ratio of unoccupied model processing units to available model processing units at the current time in the future unoccupied trend of model processing units.

[0371] It should be noted that the explanation of the embodiment of the communication method applied to the second node 102 in the communication system shown in FIG1 can be referred to the explanation of the embodiment of the communication method of the first node 101 in the communication system shown in FIG1.

[0372] The disclosed embodiments can divide the communication device into functional modules according to the above method embodiments. For example, each function can be divided into a separate functional module, or two or more functions can be integrated into one functional module. The integrated module can be implemented in hardware or software. It should be noted that the module division in this disclosed embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods. The following description uses the example of dividing each functional module according to each function.

[0373] Figure 5 is a block diagram of a communication device according to some embodiments, which can execute the communication method provided in the above-described method embodiments. As shown in Figure 5, the communication device includes a receiving module 501 and a transmitting module 502.

[0374] The receiving module 501 is used to receive configuration information from the second node; the configuration information is used by the first node to send the first report.

[0375] The sending module 502 is used to send first information to the second node; the first information is used to indicate information related to the model processing unit of the first node.

[0376] In some embodiments, the model processing unit of the first node includes a first type of model processing unit and a second type of model processing unit; the relevant information of the first type of model processing unit has been reported to the second node, while the relevant information of the second type of model processing unit has not been reported to the second node.

[0377] In some embodiments, the first information includes at least one of the following: the number of unoccupied model processing units in the model processing units of the first node at the current time, the number of occupied model processing units in the model processing units of the first node at the current time, the occupancy trend of the model processing units of the first node in a future time period, and the unoccupied trend of the model processing units of the first node in a future time period.

[0378] In some embodiments, the sending module 502 is further configured to report a first report to the second node based on configuration information; the first report is obtained based on the model processing unit of the first node.

[0379] In some embodiments, if the number of model processing units available in the first node is less than the number of model processing units required for the report reported by the configuration information scheduling, the first report includes a portion of the report or a portion of the report content reported by the configuration information scheduling; or the first report includes a portion of the updated report content reported by the configuration information scheduling; or the first report includes a portion of the unupdated report content; or the first report includes a portion of the report content obtained based on the model processing units; or the first report includes a portion of the report content not obtained based on the model processing units.

[0380] In some embodiments, the report or report content in the first report is determined based on the priority of the report or report content; the priority is determined based on at least one of the following: the index of the report or the index of the report content, the type of the report or report content, the function or functional group of the report or report content, the use case of the report or report content, the feature or feature group of the report or report content, and the number of model processing units required by the report or report content.

[0381] In some embodiments, the sending module 502 is further configured to report second information to the second node; the second information includes at least one of the following:

[0382] Location information of partial reports or partial report content, indication information of whether each report in the first report is based on the model processing unit, index of reports with abnormal content in the first report, index of reports with abnormal timeline occupation in the first report, time unit offset corresponding to reports with abnormal timeline occupation in the first report, and time unit offset caused by model switching.

[0383] In some embodiments, the second information is included in part or all of the first report.

[0384] In some embodiments, if the number of model processing units available in the first node is less than the number of model processing units required for multiple related functions, the report content corresponding to multiple functions in the first report is not updated; or the report content corresponding to multiple functions in the first report is not obtained based on model processing units; or the report content corresponding to multiple functions in the first report is obtained based on a first number of model processing units, where the first number is less than the number of model processing units required for multiple functions.

[0385] In some embodiments, where the report content corresponding to multiple functions in the first report is obtained based on a first number of model processing units, the first number is determined based on the number of model processing units required for the multiple functions and a first offset; the first offset is determined based on one of the following:

[0386] Predefined or preconfigured;

[0387] The first node is determined;

[0388] The second node indicates to the first node;

[0389] The first node and the second node negotiate and determine the outcome.

[0390] In some embodiments, the first report includes third information and tag information corresponding to the third information; the third information is generated based on the model processing unit, and the tag information is generated based on the channel state information processing unit; the occupancy start time of the model processing unit that generates the third information is the same as the occupancy start time of the channel state information processing unit that generates the tag information, or is determined based on the occupancy start time of the channel state information processing unit that generates the tag information and a second offset, or is the last time unit of the occupancy time of the channel state information processing unit that generates the tag information, or is determined based on the last time unit of the occupancy time of the channel state information processing unit that generates the tag information and a third offset value.

[0391] In some embodiments, the first report or the second information further includes at least one of the following: third information, tag information corresponding to the third information, fourth information, performance indicators of the function corresponding to the fourth information, the number of model processing units required to generate the fourth information, and a second quantity; the ability of the fourth information to determine the function of the first node by the second node; the second quantity includes the sum of the number of model processing units required to generate the third information and the number of model processing units required to generate the fourth information.

[0392] In some embodiments, when the number of available model processing units in the first node is greater than or equal to the second number, the first report includes fourth information and / or performance metrics of the function corresponding to the fourth information.

[0393] In some embodiments, the occupancy trend of the model processing units of the first node in the future time period includes: information on the number of model processing units that need to be occupied in each time unit in the future at least one time unit.

[0394] In some embodiments, the number of model processing units required for each time unit satisfies one of the following:

[0395] Absolute value;

[0396] The absolute value of the model processing unit corresponding to the initial time unit, and the difference value of the model processing unit corresponding to each time unit other than the initial time unit;

[0397] The absolute value of the model processing unit corresponding to the initial time unit, and the level index of the range of the difference values ​​of the model processing units corresponding to each time unit other than the initial time unit;

[0398] Here, the difference value of the model processing unit corresponding to a time unit is obtained by the difference of the absolute value of the model processing unit corresponding to the initial time unit;

[0399] In some embodiments, the first node includes channel state information processing unit resources for generating reports.

[0400] In some embodiments, the first information further includes at least one of the following:

[0401] The unused channel state information processing unit resources in the first node at the current moment;

[0402] Additional channel state information processing unit resources occupied;

[0403] Unused channel state information processing unit resources in the first node during the future time period;

[0404] The mapping relationship between the channel state information processing unit resources of the first node and the model processing unit of the first node.

[0405] In some embodiments, the first information includes at least one of the following:

[0406] The number of unoccupied model processing units in the first type of model processing unit at the current moment;

[0407] The number of unoccupied model processing units in the second type of model processing units at the current moment;

[0408] The sum of the number of unoccupied model processing units in the first type of model processing units at the current time and the number of unoccupied model processing units in the second type of model processing units at the current time;

[0409] The number of unoccupied model processing units in the first type of model processing units within the future time period;

[0410] The number of unoccupied model processing units in the second type of model processing units within the future time period;

[0411] The sum of the number of unoccupied model processing units in the first type of model processing units during the future time period and the number of unoccupied model processing units in the second type of model processing units during the future time period.

[0412] In some embodiments, the first type of model processing unit belongs to at least one first unit set, and the second type of model processing unit belongs to at least one second unit set; the first information includes at least one of the following:

[0413] The number of unoccupied model processing units in each first unit set at the current moment;

[0414] The number of unoccupied model processing units in each second unit set at the current moment;

[0415] The sum of the number of unoccupied model processing units in at least one first unit set at the current time and the number of unoccupied model processing units in at least one second unit set at the current time;

[0416] The number of unused model processing units in each first unit set within the future time period;

[0417] The number of unused model processing units in each first unit set within the future time period;

[0418] The sum of the number of unoccupied model processing units in at least one first unit set within a future time period and the number of unoccupied model processing units in at least one second unit set within a future time period.

[0419] In some embodiments, the second type of model processing unit corresponds to at least one set of model processing units; different sets of model processing units correspond to different report types or types of report content or report entities or functions or function groups or use cases or features or feature groups.

[0420] In some embodiments, the first information includes at least one of the following:

[0421] The number of model processing units supported by each set of model processing units;

[0422] Each set of model processing units corresponds to a report type, report content category, report entity, function, function group, use case, feature, or feature group;

[0423] The priority of each model processing unit set;

[0424] Resource usage rules corresponding to each model processing unit set.

[0425] In some embodiments, different report types, report content categories, report entities, functions, function groups, use cases, features, or feature groups have different priorities.

[0426] In some embodiments, resource occupancy rules include one of the following:

[0427] Resources in a high-priority set of model processing units are not allowed to be allocated to a low-priority function, while some or all of the resources in a low-priority set of model processing units are allowed to be allocated to a high-priority function.

[0428] Resources in the highest priority model processing unit set are not allowed to be allocated to lower priority functions. Functions corresponding to the non-highest priority model processing unit set share some or all of the resources in the non-highest priority model processing unit set.

[0429] The functions corresponding to model processing unit sets with different priorities share some or all of the resources in their respective model processing unit sets;

[0430] Different priority sets of model processing units share some or all of the resources in the corresponding set of model processing units; the order of resource sharing in the set of model processing units is positively correlated with the number of available resources included in the set of model processing units.

[0431] In some embodiments, the resource occupancy rules also include:

[0432] For a first set of model processing units with a priority lower than the priority threshold, if the number of occupied model processing units in the first set of model processing units is less than or equal to the first threshold value, then some model processing units in the first set of model processing units will not be activated, or it will be determined that some model processing units in the first set of model processing units are not allowed to be occupied.

[0433] In some embodiments, the number of shared resources is determined in one of the following ways:

[0434] The first node is determined by itself;

[0435] Second node indication;

[0436] The first node and the second node negotiate and determine the outcome.

[0437] In some embodiments, the sending module 502 is specifically used to send first information to the second node based on the relationship between a preset threshold value and the parameters corresponding to the preset threshold value.

[0438] In some embodiments, the preset threshold value includes at least one of the following:

[0439] The number threshold value corresponding to the number of available model processing units;

[0440] The ratio threshold value corresponding to the ratio of available model processing units to all model processing units;

[0441] The number threshold value corresponding to the number of occupied model processing units;

[0442] The ratio threshold value corresponding to the ratio of occupied model processing units to all model processing units;

[0443] The number threshold value corresponding to the number of occupied model processing units in the future time period's model processing unit occupancy trend;

[0444] The ratio threshold value corresponding to the ratio of occupied model processing units to all model processing units in the future time period's model processing unit occupancy trend;

[0445] The ratio threshold value corresponding to the ratio of occupied model processing units to available model processing units at the current time in the future time period occupancy trend;

[0446] The number threshold value corresponding to the number of unoccupied model processing units in the unoccupied trend of model processing units in the future time period;

[0447] The ratio threshold value corresponding to the ratio of unoccupied model processing units to all model processing units in the unoccupied trend of model processing units in the future time period;

[0448] The threshold value corresponding to the ratio of unoccupied model processing units to available model processing units at the current time in the future unoccupied trend of model processing units.

[0449] In some embodiments, the first information may further include the relationship between the preset threshold value and the parameter corresponding to the preset threshold value;

[0450] The parameters corresponding to the preset threshold value include at least one of the following:

[0451] The number of available model processing units;

[0452] The ratio of available model processing units to all model processing units;

[0453] The number of model processing units already in use;

[0454] The ratio of occupied model processing units to all model processing units;

[0455] The number of model processing units occupied in the future time period;

[0456] The proportion of occupied model processing units to all model processing units in the future time period;

[0457] The ratio of occupied model processing units to available model processing units at the current time in the future time period occupancy trend;

[0458] The number of unoccupied model processing units in the future time period's unoccupied trend;

[0459] The ratio of unoccupied model processing units to all model processing units in the future time period's unoccupied trend;

[0460] The ratio of unoccupied model processing units to available model processing units at the current time in the future unoccupied trend of model processing units.

[0461] Figure 6 is a block diagram of another communication device according to some embodiments, which can perform the communication method provided in the above-described method embodiments. As shown in Figure 6, the communication device includes a transmitting module 601 and a receiving module 602.

[0462] Sending module 602 is used to send configuration information to the first node; the configuration information is used by the first node to send the first report;

[0463] The receiving module 602 is used to receive first information from the first node; the first information is used to indicate information related to the model processing unit.

[0464] In implementing the functions of the integrated modules described above in hardware, this disclosure provides another structure for the communication device involved in the above embodiments. As shown in FIG7, the communication device includes a processor 702 and a bus 704. In some embodiments, the communication device may further include a memory 701. In some embodiments, the communication device may further include a communication interface 703.

[0465] Processor 702 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with embodiments of this disclosure. Processor 702 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with embodiments of this disclosure. Processor 702 may also be a combination that implements computing functions, for example, including one or more microprocessor combinations, a combination of a digital signal processor (DSP) and a microprocessor, etc.

[0466] The communication interface 703 is used to connect with other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.

[0467] The memory 701 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0468] In some embodiments, the memory 701 may exist independently of the processor 702. The memory 701 may be connected to the processor 702 via a bus 704 and is used to store instructions or program code. When the processor 702 calls and executes the instructions or program code stored in the memory 701, it can implement the methods provided in the embodiments of this disclosure.

[0469] In other embodiments, the memory 701 may also be integrated with the processor 702.

[0470] Bus 704 can be an extended industry standard architecture (EISA) bus, etc. Bus 704 can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in Figure 7, but this does not mean that there is only one bus or one type of bus.

[0471] Some embodiments of this disclosure provide a computer-readable storage medium (e.g., a non-transitory computer-readable storage medium) storing computer program instructions that, when executed on a computer, cause the computer to perform the methods described in any of the above embodiments.

[0472] Exemplary examples show that the aforementioned computer-readable storage media may include, but are not limited to: magnetic storage devices (e.g., hard disks, floppy disks, or magnetic tapes), optical disks (e.g., compact disks (CDs), digital versatile disks (DVDs), etc.), smart cards, and flash memory devices (e.g., erasable programmable read-only memory (EPROMs), cards, sticks, or key drives, etc.). The various computer-readable storage media described in this disclosure may represent one or more devices for storing information and / or other machine-readable storage media. The term "machine-readable storage media" may include, but is not limited to, wireless channels and various other media capable of storing, containing, and / or carrying instructions and / or data.

[0473] This disclosure provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the methods described in any of the above embodiments.

[0474] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any changes or substitutions within the technical scope disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A communication method, wherein, The method is applied to a first node, and the method comprises: receiving configuration information from a second node; the configuration information is used for the first node to send a first report; sending first information to the second node; the first information is used for indicating information related to a model processing unit of the first node.

2. The method of claim 1, wherein the model processing unit of the first node comprises a first type of model processing unit and a second type of model processing unit; the related information of the first type of model processing unit has been reported to the second node, and the related information of the second type of model processing unit has not been reported to the second node.

3. The method of claim 2, wherein, the first information comprises at least one of the following: quantity information of model processing units not occupied in the model processing unit of the first node at a current time, quantity information of model processing units occupied in the model processing unit of the first node at the current time, occupation trend of the model processing unit of the first node in a future time period, or unoccupied trend of the model processing unit of the first node in the future time period.

4. The method of claim 3, further comprising: reporting a first report to the second node based on the configuration information; the first report is obtained based on the model processing unit of the first node.

5. The method of claim 4, wherein in a case where the quantity information of the model processing units available in the first node includes a quantity less than a quantity of model processing units required for a report scheduled to be reported by the configuration information, the first report includes part of the report or part of the report content in the report scheduled to be reported by the configuration information; or the first report includes part of the updated report content in the report scheduled to be reported by the configuration information; or the first report includes part of the unupdated report content; or the first report includes part of the report content obtained based on the model processing unit; or the first report includes part of the report content not obtained based on the model processing unit.

6. The method of claim 5, wherein, the report or the report content in the first report is determined based on a priority of the report or the report content; the priority is determined based on at least one of the following: index of the report or index of the report content, index of a reporting entity, type of the report or the report content, function or function group of the report or the report content, use case of the report or the report content, feature or feature group of the report or the report content, or quantity of model processing units required for the report or the report content.

7. The method of claim 5, further comprising: reporting second information to the second node; the second information comprises at least one of the following: location information of the part of the report or the part of the report content, indication information of whether each report in the first report is obtained based on the model processing unit, index of a report with abnormal report content in the first report, index of a report with abnormal timeline occupation in the first report, time unit offset corresponding to the report with abnormal timeline occupation in the first report, or time unit offset caused by model switching.

8. The method of claim 7, wherein, the second information is partially or entirely contained in the first report.

9. The method of claim 4, wherein, in a case where the number information of the model processing units available in the first node includes a number smaller than a number of model processing units required by the plurality of functions having the association relationship, the report content corresponding to the plurality of functions in the first report is not updated; or the report content corresponding to the plurality of functions in the first report is not obtained based on the model processing units; or the report content corresponding to the plurality of functions in the first report is obtained based on a first number of model processing units, the first number being smaller than the number of model processing units required by the plurality of functions.

10. The method of claim 9, wherein, in a case where the report content corresponding to the plurality of functions in the first report is obtained based on a first number of model processing units, the first number is determined based on the number of model processing units required by the plurality of functions and a first offset; the first offset is determined based on one of the following: predefined or preconfigured; determined by the first node; indicated by the second node to the first node; determined by negotiation between the first node and the second node.

11. The method of claim 4, wherein, the first report includes third information and label information corresponding to the third information; the third information is generated based on the model processing units, and the label information is generated based on the channel state information processing units; an occupation start time of the model processing unit generating the third information is the same as an occupation start time of the channel state information processing unit generating the label information, or is determined based on the occupation start time of the channel state information processing unit generating the label information and a second offset, or is the last time unit of the occupation time of the channel state information processing unit generating the label information, or is determined based on the last time unit of the occupation time of the channel state information processing unit generating the label information and a third offset value.

12. The method of claim 7, wherein, the first report or the second information further includes at least one of the following: third information, label information corresponding to the third information, fourth information, a performance indicator of a function corresponding to the fourth information, a number or a second number of model processing units required for generating the fourth information; the fourth information is used by the second node to determine the capability of the function of the first node; the second number includes a sum of the number of model processing units required for generating the third information and the number of model processing units required for generating the fourth information.

13. The method of claim 12, wherein, in a case where the number of model processing units available in the first node is greater than or equal to the second number, the first report includes at least one of the fourth information and the performance indicator of the function corresponding to the fourth information.

14. The method of claim 3, wherein, the occupation trend of the model processing units of the first node in the future time period includes number information of model processing units required for occupation in each of at least one future time unit.

15. The method of claim 14, wherein, the number information of model processing units required for occupation in each of the time units satisfies one of the following: an absolute value; an absolute value of model processing units corresponding to an initial time unit and a differential value of model processing units corresponding to each of the time units other than the initial time unit; an absolute value of model processing units corresponding to an initial time unit and a differential value of model processing units corresponding to each of the time units other than the initial time unit. an absolute value of a model processing unit corresponding to an initial time unit, a range index of a differential value of a model processing unit corresponding to each of the time units other than the initial time unit; wherein the differential value of the model processing unit corresponding to one of the time units is obtained by differentiating the absolute value of the model processing unit corresponding to the initial time unit.

16. The method of claim 3, wherein, the first node comprises a channel state information processing unit resource for generating the report.

17. The method of claim 16, wherein, the first information further comprises at least one of: a number of unoccupied channel state information processing unit resources in the first node at a current time; an additional occupied channel state information processing unit resource; a number of unoccupied channel state information processing unit resources in the first node in a future time period; a mapping relationship between the channel state information processing unit resources of the first node and the model processing units of the first node.

18. The method of claim 2, wherein, the first information comprises at least one of: a number of unoccupied model processing units of the first type of model processing units at a current time; a number of unoccupied model processing units of the second type of model processing units at a current time; a sum of a number of unoccupied model processing units of the first type of model processing units at a current time and a number of unoccupied model processing units of the second type of model processing units at a current time; a number of unoccupied model processing units of the first type of model processing units in a future time period; a number of unoccupied model processing units of the second type of model processing units in a future time period; a sum of a number of unoccupied model processing units of the first type of model processing units in a future time period and a number of unoccupied model processing units of the second type of model processing units in a future time period.

19. The method of claim 2, wherein, the first type of model processing units belongs to at least one first unit set, and the second type of model processing units belongs to at least one second unit set; the first information comprises at least one of: a number of unoccupied model processing units in each of the first unit sets at a current time; a number of unoccupied model processing units in each of the second unit sets at a current time; a sum of a number of unoccupied model processing units in the at least one first unit set at a current time and a number of unoccupied model processing units in the at least one second unit set at a current time; a number of unoccupied model processing units in each of the first unit sets in a future time period; a number of unoccupied model processing units in each of the first unit sets in a future a sum of a number of unoccupied model processing units in the at least one first unit set in a future time period and a number of unoccupied model processing units in the at least one second unit set in a future time period.

20. The method of claim 2, wherein, the second type of model processing units corresponds to at least one model processing unit set; different model processing unit sets correspond to different report types or different categories of report contents or different report entities or different functions or different function groups or different use cases or different features or different feature groups.

21. The method of claim 20, wherein, the first information comprises at least one of: A number of model processing units supported by each of the model processing unit sets; A reporting type or a reporting content category or a reporting entity or a function or a function group or a use case or a feature or a feature group corresponding to each of the model processing unit sets; A priority of each of the model processing unit sets; A resource occupation rule corresponding to each of the model processing unit sets.

22. The method of claim 21, wherein, The priorities of different reporting types or reporting content categories or reporting entities or functions or function groups or use cases or features or feature groups are different.

23. The method of claim 21, wherein, The resource occupation rule includes one of the following: Resources in a model processing unit set with a high priority are not allowed to be allocated to functions with a lower priority, and part or all of the resources in the model processing unit set with a low priority are allowed to be allocated to functions with a higher priority. Resources in a model processing unit set with the highest priority are not allowed to be allocated to functions with a lower priority, and functions corresponding to model processing unit sets with a non-highest priority share part or all of the resources in the model processing unit set with the non-highest priority. Functions corresponding to model processing unit sets with different priorities share part or all of the resources in the respective model processing unit sets. Functions corresponding to model processing unit sets share part or all of the resources in the model processing unit sets corresponding to the functions, and the order of resource sharing of the model processing unit sets is positively correlated with the number of available resources included in the model processing unit sets.

24. The method of claim 23, wherein, The resource occupation rule further includes: For a first model processing unit set with a priority lower than a priority threshold, if the number of occupied model processing units in the first model processing unit set is less than or equal to a first threshold value, part of the model processing units in the first model processing unit set are not activated, or it is determined that part of the model processing units in the first model processing unit set are not allowed to be occupied.

25. The method of claim 23, wherein, The number of shared resources is determined in one of the following manners: Determined by the first node itself; Indicated by the second node; Negotiated and determined by the first node and the second node.

26. The method of claim 1, wherein, The sending of the first information to the second node includes: Based on a size relationship between a preset threshold value and a parameter corresponding to the preset threshold value, the first information is sent to the second node.

27. The method of claim 26, wherein, The preset threshold value includes at least one of the following: A number threshold value corresponding to the number of available model processing units; A ratio threshold value corresponding to a ratio of available model processing units to all model processing units; A number threshold value corresponding to the number of occupied model processing units; A ratio threshold value corresponding to a ratio of occupied model processing units to all model processing units; A number threshold value corresponding to the number of occupied model processing units in an occupation trend of model processing units in a future time period; A ratio threshold value corresponding to a ratio of occupied model processing units to all model processing units in an occupation trend of model processing units in a future time period; A ratio threshold value corresponding to a ratio of occupied model processing units to available model processing units at a current time in an occupation trend of model processing units in a future time period; a quantity threshold corresponding to a quantity of unoccupied model processing units in an unoccupied trend of the model processing units in the future time period; a ratio threshold corresponding to a ratio of unoccupied model processing units to all model processing units in an unoccupied trend of the model processing units in the future time period; a ratio threshold corresponding to a ratio of unoccupied model processing units to available model processing units at the current time in an unoccupied trend of the model processing units in the future time period.

28. The method of claim 26, wherein, the first information further comprises a size relationship between the preset threshold and a parameter corresponding to the preset threshold; the parameter corresponding to the preset threshold comprises at least one of: a quantity of available model processing units; a ratio of available model processing units to all model processing units; a quantity of occupied model processing units; a ratio of occupied model processing units to all model processing units; a quantity of occupied model processing units in an occupied trend of the model processing units in the future time period; a ratio of occupied model processing units to all model processing units in an occupied trend of the model processing units in the future time period; a ratio of occupied model processing units to available model processing units at the current time in an occupied trend of the model processing units in the future time period; a quantity of unoccupied model processing units in an unoccupied trend of the model processing units in the future time period; a ratio of unoccupied model processing units to all model processing units in an unoccupied trend of the model processing units in the future time period; a ratio of unoccupied model processing units to available model processing units at the current time in an unoccupied trend of the model processing units in the future time period.

29. A communication method, wherein, The method applied to the second node comprises: sending configuration information to the first node; the configuration information is used for the first node to send the first report; receiving first information from the first node; the first information is used for indicating information related to model processing units.

30. A communications device comprising: a memory and a processor; the memory and the processor are coupled; the memory is used for storing instructions executable by the processor; the processor executes the instructions to perform the method in any one of claims 1-28, or to perform the method in claim 29.

31. A computer readable storage medium, wherein, The computer readable storage medium stores computer instructions, when the computer instructions run on a computer, make the computer execute the method in any one of claims 1-28, or perform the method in claim 29.

32. A computer program product, wherein, The computer program product comprises computing technology program instructions, when the computing technology program instructions are executed by a processor, the method in any one of claims 1-28 is implemented, or the method in claim 29 is performed. The computer program product comprises computing technology program instructions, when the computing technology program instructions are executed by a processor, the method in any one of claims 1-28 is implemented, or the method in claim 29 is performed.