Device and method for communication, and computer readable storage medium

Abstract

Example embodiments relate to a terminal device, a network device, a method and a computer readable storage medium for communication. In an example method, a terminal device comprises: a processor; and a transceiver coupled to the processor, where the processor is configured to transmit, via the transceiver, an indication of a number of processing units for an artificial intelligence (AI) / machine learning (ML) model or a plurality of AI / ML models for an AI / ML function to a network device; and receiving, from the network device via the transceiver, a configuration for operation of the terminal device associated with the AI / ML model or the AI / ML function, wherein the configuration is determined based on the number of processing units or the plurality of AI / ML models. Therefore, the network equipment can flexibly trigger the AI / ML operation in the terminal equipment according to different use cases.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of communications, and more particularly, to a terminal device, a network device, a method, and a computer-readable storage medium for communications. Background Art

[0002] The 3rd Generation Partnership Project (3GPP) is investigating the potential benefits of employing artificial intelligence (AI) / machine learning (ML) models for the air interface in several use cases, such as using AI / ML models in beam prediction.

[0003] From a hardware perspective, each AI / ML model corresponds to a set of hardware resources, including at least memory and MACs (multipliers and adders). Therefore, each AI / ML model can only be used for one AI / ML prediction operation at a time. This requires some management during AI / ML model deployment. Summary of the Invention

[0004] In general, example embodiments of the present disclosure provide a solution for communicating with AI / ML models.

[0005] In a first aspect, a terminal device is provided. The terminal device includes a processor and a transceiver, the transceiver being coupled to the processor. The processor is configured to send, via the transceiver, to a network device an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function. The processor is further configured to receive, from the network device, via the transceiver, a configuration for operations of the terminal device associated with the AI / ML model or the AI / ML function, wherein the configuration is determined based on the number of processing units or the plurality of AI / ML models.

[0006] In a second aspect, a network device is provided. The network device includes a processor and a transceiver, the transceiver being coupled to the processor. The processor is configured to receive, from a terminal device via the transceiver, an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function. The processor is further configured to determine, based on the number of processing units or the plurality of AI / ML models, a configuration for operation of the terminal device associated with the AI / ML model or the AI / ML function. The processor is further configured to send the configuration to the terminal device via the transceiver.

[0007] In a third aspect, a method performed by a terminal device is provided. The method includes sending an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function, to a network device. The method also includes receiving, from the network device, a configuration for operations of the terminal device associated with the AI / ML model or the AI / ML function, wherein the configuration is determined based on the number of processing units or the plurality of AI / ML models.

[0008] In a fourth aspect, a method performed by a network device is provided. The method includes receiving, from a terminal device, an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function. The method also includes determining, based on the number of processing units or the plurality of AI / ML models, a configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function. The method also includes sending the configuration to the terminal device.

[0009] In a fifth aspect, a non-transitory computer-readable medium having program instructions stored thereon is provided. When executed by an apparatus, the program instructions cause the apparatus to at least: send an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function, to a network device; and receive from the network device a configuration for terminal device operations associated with the AI / ML model or the AI / ML function, wherein the configuration is determined based on the number of processing units or the plurality of AI / ML models.

[0010] In a sixth aspect, a non-transitory computer-readable medium having program instructions stored thereon is provided. When executed by an apparatus, the program instructions cause the apparatus to at least: receive from a terminal device an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function; determine, based on the number of processing units or the plurality of AI / ML models, a configuration for operations of the terminal device associated with the AI / ML model or the AI / ML function; and transmit the configuration to the terminal device.

[0011] It should be understood that this summary is not intended to identify the key or essential features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Some example embodiments will now be described with reference to the accompanying drawings, in which:

[0013] Figure 1An example communication system is described in which some embodiments of the present disclosure may be implemented;

[0014] Figure 2 describes example AI / ML model deployments in which some example embodiments of the present disclosure may be implemented;

[0015] Figure 3 Describes the process of communicating with an AI / ML model according to some example embodiments of the present disclosure;

[0016] Figure 4 Described are AI / ML management based on AI / ML models according to some example embodiments of the present disclosure;

[0017] Figure 5 Described are AI / ML management based on AI / ML functionality according to some example embodiments of the present disclosure;

[0018] Figure 6 Describes example AI / ML models or processing unit occupancy times according to some example embodiments of the present disclosure;

[0019] Figure 7A Another example AI / ML model or processing unit occupancy time according to some example embodiments of the present disclosure is described;

[0020] Figure 7B Described is yet another example AI / ML model or processing unit occupancy time according to some example embodiments of the present disclosure;

[0021] Figure 8 An example report having both a first type of processing unit and a second type of processing unit is described according to some example embodiments of the present disclosure;

[0022] Figure 9 An example report having a first type of processing unit is described according to some example embodiments of the present disclosure;

[0023] Figure 10 An example of a method implemented on a terminal device according to some example embodiments of the present disclosure is described;

[0024] Figure 11 describes examples of methods implemented on network devices according to some example embodiments of the present disclosure; and

[0025] Figure 12 A simplified block diagram of a device suitable for implementing embodiments of the present disclosure is depicted.

[0026] Throughout the drawings, the same or similar reference numbers represent the same or similar elements. DETAILED DESCRIPTION

[0027] The principles of the present disclosure will now be described with reference to some exemplary embodiments. It should be understood that these embodiments are described only for the purpose of illustrating and helping those skilled in the art understand and implement the present disclosure, and are not intended to limit the scope of the present disclosure. In addition to the methods described below, the disclosure described herein may also be implemented in various other ways.

[0028] In the following description and claims, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

[0029] References in this disclosure to "one embodiment," "an embodiment," "an example embodiment," and the like indicate that the described embodiment may include certain features, structures, or characteristics, but not every embodiment must include the certain features, structures, or characteristics. Furthermore, these phrases do not necessarily refer to the same embodiment. Furthermore, when a certain feature, structure, or characteristic is described in conjunction with one embodiment, whether or not explicitly described, those skilled in the art will appreciate that such feature, structure, or characteristic may be affected in conjunction with other embodiments.

[0030] It should be understood that although the terms "first" and "second" may be used herein to describe various elements, these elements should not be limited by these terms. These terms are merely used to distinguish between elements. For example, a first element can be referred to as a second element, and similarly, a second element can be referred to as a first element without departing from the scope of the example embodiments. As used herein, the term "and / or" includes any and all combinations of one or more of the listed terms.

[0031] The terms used herein are used only to describe specific embodiments and are not intended to limit the example embodiments. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should be further understood that when used herein, the terms "comprise," "include," "have," "contain," "include," and / or "include" specify the presence of the stated features, elements, and / or components, but do not preclude the presence or addition of one or more other features, elements, components, and / or combinations thereof.

[0032] As used herein, the term "communication network" refers to a network that complies with any suitable communication standard, such as fifth generation new radio (5G NR), long term evolution (LTE), advanced LTE (LTE-A), wideband code division multiple access (WCDMA), high speed packet access (HSPA), narrowband Internet of Things (NB-IoT), etc. In addition, the communication between the terminal device and the network equipment in the communication network can be performed according to any suitable generation of communication protocol, including but not limited to the fourth generation (4G), 4.5G, future fifth generation (5G) communication protocols and / or any other protocols currently known or developed in the future. The embodiments of the present disclosure can be applied to various communication systems. In view of the rapid development of communication technology, there will certainly be future types of communication technologies and systems that can embody the present disclosure. The scope of the present disclosure should not be considered to be limited to the above-mentioned systems.

[0033] The term "network function" (NF) as used herein refers to a function in the 5G core network, including at least one of the following: network slice selection function (NSSF), network exposure function (NEF), network storage function (NRF), policy control function (PCF), unified data management (UDM), unified data repository (UDR), application function (AF), network data analysis function (NWDAF), trusted non-3GPP gateway function (TNGF), authentication server function (AUSF), access and mobility management function (AMF), session management function (SMF) and user plane function (UPF).

[0034] The term "terminal device" refers to any terminal device capable of wireless communication. By way of example and not limitation, a terminal device may also be referred to as a communication device, user equipment (UE), subscriber station (SS), portable subscriber station, mobile station (MS), or access terminal (AT). Terminal devices may include, but are not limited to, mobile phones, cellular phones, smart phones, voice over IP (VoIP), wireless local loop phones, tablet computers, wearable terminal devices, personal digital assistants (PDAs), laptop computers, desktop computers, image capture terminal devices (e.g., digital cameras), gaming terminal devices, music storage and playback devices, in-vehicle wireless terminal devices, wireless endpoints, mobile stations, laptop embedded devices (LEEs), laptop mounted devices (LMEs), USB dongles, smart devices, wireless customer premises equipment (CPEs), Internet of Things (IoT) devices, watches or other wearable devices, head-mounted displays (HMDs), vehicles, drones, medical devices and applications (e.g., remote surgery), industrial devices and applications (e.g., robots and / or other wireless devices operating in industrial and / or automated process chain environments), consumer electronic devices, devices operating on commercial and / or industrial wireless networks, etc. In the following description, the terms "terminal device", "communication device" and "terminal" may be used interchangeably.

[0035] 3GPP is studying the potential benefits of adopting AI / ML capabilities on the air interface for some identified use cases. One scenario is to use AI / ML capabilities on the UE side by using multiple AI / ML models, where the AI / ML models can be used for channel state information (CSI) or beam prediction.

[0036] Figure 1 An example communication system is described in which some embodiments of the present disclosure may be implemented. In the communication system 100, beam prediction between the network device 110 and the terminal device 120 may be implemented via an AI / ML model.

[0037] For example, an AI / ML model can be used to predict the best K beams or beam pairs from one beam set based on the measurement results of another beam set, where the number of beams or beam pairs in the predicted beam set is greater than the number of beams or beam pairs in the measured beam set. Some AI / ML models can be used to predict the CSI or the best beams or beam pairs for future time instances based on historical measurement results. From a hardware perspective, each AI / ML model corresponds to a set of hardware resources, including memory and MAC (multiplier and adder). Therefore, each AI / ML model can only be used for one AI / ML prediction operation at a time instance. The network and UE need to align the occupancy of all available AI / ML models to achieve more efficient AI / ML operations. Management of AI / ML models deployed on the UE side needs to be implemented based on the concept of AI / ML model occupancy. Those skilled in the art will understand that AI / ML model management is also required in other situations such as CSI compression and positioning.

[0038] In some embodiments of the present disclosure, a processing unit for CSI is introduced to achieve more efficient CSI triggering. The CSI processing standard will be described in detail below.

[0039] The terminal device 120 indicates to the network device 110 the number of simultaneous CSI calculations supported in the component carriers with the RRC parameter simultaneousCSI-ReportsPerCC CPU If the terminal device 120 supports N CPU Synchronous CSI calculation, it is said to have N for processing CSI reports CPU If computing the CSI report in a given symbol (e.g., an OFDM symbol) takes L processing units, then the terminal device 120 has N processing units. CPU -L unoccupied processing units. If N CPU- On the same symbol (e.g., Orthogonal Frequency Division Multiplexing (OFDM) symbol) not occupied by L processing units, N CSI report calculations start to occupy their own processing units. Each CSI report (n = 0, ..., N-1) corresponds to Or calculate occupancy per CSI report The terminal device 120 does not need to update the NM requested CSI reports with the lowest priority, where 0≤M≤N is such that The maximum value of .

[0040] It is expected that the terminal device 120 will not be configured to include more than N CPU The CSI reporting is configured with a non-periodic CSI trigger state. The processing of the CSI report calculation occupies several processing units of several symbols, as shown below. For CSI reports with the CSI-ReportConfig parameter reportQuantity set to "none" and the CSI-RS-ResourceSet configured with the higher-level parameter trs-Info, CPu = 0. For CSI reports of CSI-ReportConfig with the higher layer parameter reportQuantity set to "cri-RSRP", "ssb-Index-RSRP", "cri-SINR", "ssb-Index-SINR", "cri-RSRP-Index", "ssb-Index-RSRP-Index", "cri-SINR-Index", "ssb-Index-SINR-Index" or "none", O CPu = 1. CSI-RS-ResourceSet with the higher layer parameter trs-Info is not configured. In some embodiments of the present disclosure, "cri" corresponds to a CSI-RS resource indicator, which is related to a beam indicator.

[0041] For CSI reporting with a CSI-ReportConfig where the higher layer parameter reportQuantity is set to "cri-RI-PMI-CQI", "cri-RI-i1", "cri-RI-i1-CQI", "cri-RI-CQI", or "cri-RI-LI-PMI-CQI", the following cases may occur. If max{μPDCCH, μCSI-RS, μUL}≤3, and when L=0 processing units are occupied, if CSI reporting is triggered aperiodically without sending a physical uplink shared channel (PUSCH) with transport blocks or hybrid automatic repeat request acknowledgement (HARQ-ACK) or both, where the CSI corresponds to a single CSI with wideband frequency granularity and up to 4 channel state information reference signal (CSI-RS) ports without CRI reporting in a single resource, and where codebookType is set to "typeI-SinglePanel" or where reportQuantity is set to "cri-RI-CQI", then O CPu =N CPU If CSI-ReportConfig is configured with codebookType set to "typeI-SinglePanel", and the corresponding CSI-RS resource set for channel measurement is configured with two resource groups and N resource pairs, then CPU =X·N+M, where x is the number of processing units occupied by a pair of codec mode requests (CMRs) for mTRP-CSI-numprocessing unit-r17. Otherwise, O CPU =K s , where Ks is the number of CSI-RS resources in the CSI-RS resource set used for channel measurement.

[0042] For CSI reports with a CSI-ReportConfig where the higher-layer parameter reportQuantity is not set to "None", one or more processing units are occupied for a number of OFDM symbols as shown below. Periodic or semi-persistent CSI reporting (excluding the initial semi-persistent (SP) CSI report on the PUSCH following the PDCCH that triggers the report) occupies one or more processing units, starting from the first symbol of the earliest of each CSI-RS / Channel State Information-Interference Measurement (CSI-IM) / Synchronization Signal and PBCH Block (SSB) resource used for channel or interference measurement, with the respective latest CSI-RS / CSI-IM / SSB opportunity no later than the corresponding CSI reference resource, until the last symbol of the configured Physical Uplink Shared Channel (PUSCH) or Physical Uplink Control Channel (PUCCH) carrying the report. Aperiodic CSI reporting occupies one or more processing units, starting from the first symbol following the Physical Downlink Control Channel (PDCCH) that triggers the CSI report, until the last symbol of the scheduled PUSCH carrying the report. When a PDCCH reception includes two PDCCH candidates from two respective search space sets, the PDCCH candidate that ends later in time is used for determining the processing unit occupancy duration. The initial semi-persistent CSI report on the PUSCH after a PDCCH trigger occupies one or more processing units, starting from the first symbol after the PDCCH until the last symbol of the scheduled PUSCH carrying the report. When a PDCCH reception includes two PDCCH candidates from two respective search space sets, the PDCCH candidate that ends later in time is used for determining the processing unit occupancy duration.

[0043] For CSI reporting with a CSI-ReportConfig with the higher-layer parameter reportQuantity set to "None" and a CSI-RS-ResourceSet without the higher-layer parameter trs-Info configured, one or more processing units are occupied for a number of OFDM symbols as shown below. A semi-persistent CSI report (excluding the initial semi-persistent CSI report on the PUSCH following a PDCCH that triggers the report) occupies one or more processing units starting from the first symbol of the earliest of each transmission opportunity of the periodic or semi-persistent CSI-RS / CSI-IM / SSB resource used for channel measurement for L1-RSRP calculation, and ending Z′3 symbols after the last symbol of the latest of the CSI-RS / CSI-IM / SSB resource used for channel measurement for L1-RSRP calculation in each transmission opportunity. Aperiodic CSI reporting occupies one or more processing units, starting from the first symbol after the PDCCH that triggers the CSI report, and ending with the last symbol between Z3 symbols after the first symbol after the PDCCH that triggers the CSI report and Z′3 symbols after the last symbol of the latest of each CSI-RS / CSI-IM / SSB resource for channel measurement used for L1-RSRP calculation. (Z3, Z′3) is defined in Table 5.4-2 of 3GPP specification 38.213.

[0044] In any time slot, the terminal device 110 is not expected to have more active CSI-RS ports or active CSI-RS resources in the active bandwidth part (BWP) than the reported capability. No Zero Power (NZP) CSI-RS resources are active for the time period defined below. For aperiodic CSI-RS, the occupied time starts at the end of the PDCCH containing the request and ends at the end of the scheduled PUSCH containing the report associated with this aperiodic CSI-RS. When a PDCCH candidate is associated with a search space set configured with search Space Linking, in order to determine the NZP CSI-RS resource active duration, the PDCCH candidate that ends at the later time of the two linked PDCCH candidates will be used. For Semi-Persistent (SP) CSI-RS, the occupied time starts at the end when the activation command is applied and ends at the end when the deactivation command is applied. For periodic CSI-RS, the occupied time starts when the periodic CSI-RS is configured by higher layer signaling and ends when the periodic CSI-RS configuration is released. If a CSI-RS resource is referenced N times by one or more CSI report settings, the CSI-RS resource and the CSI-RS port in the CSI-RS resource are counted N times. For a CSI-RS resource set for channel measurement configured with two resource groups and N resource pairs, if a CSI-RS resource is referenced X times by one of the M CSI-RS resources and / or one or two resource pairs, the CSI-RS resource and the CSI-RS port in the CSI-RS resource are counted X times.

[0045] In some embodiments, research can be implemented using a lifecycle management (LCM) program based on the AI / ML model having a model ID with associated information and / or model capabilities for at least some AI / ML operations.

[0046] In some embodiments, at least for the model selection, activation, deactivation, switching, and fallback of the UE-side model and the bilateral model, the following mechanisms may be used for research. If the model selection is determined by the network device 110, it may be initiated by the network. The model selection may also be initiated by the UE, with a request to the network. If the model selection is determined by the terminal device 120, the network device 110 may be configured as event-triggered, and the decision of the terminal device 120 may be reported to the network device 110. It may be UE-autonomous, with the decision of the terminal device 120 reported to the network device 110. It may be UE-autonomous, with the decision of the terminal device 120 not reported to the network device 110.

[0047] In some embodiments, the study of potential regulatory implications enables the development of a set of specific models, such as scenario- or configuration-specific and site-specific models, compared to a unified model. User data privacy needs to be protected. The provision of auxiliary information may require consideration of the feasibility of disclosing proprietary information to the other party.

[0048] In some embodiments, the study of the impact of specifications may support multiple AI / ML models for the same functionality, including at least the following aspects: procedures and auxiliary signaling for AI / ML model switching and / or selection.

[0049] In some embodiments, for the UE-side model, the study can be implemented through the following mechanism for the LCM procedure. For the function-based LCM procedure, the indication of activation / deactivation / switching / fallback is based on a separate AI / ML function. The terminal device 120 can have one AI / ML model for a function, or the terminal device 120 can have multiple AI / ML models for a function. It is necessary to determine whether or how to indicate the AI / ML function. For the model ID-based LCM procedure, the indication of model selection / activation / deactivation / switching / fallback is based on a separate model ID.

[0050] In some embodiments, for model identification, a process or method for identifying an AI / ML model may be communicated between network device 110 and terminal device 120. In some embodiments, for function identification, a process or method for identifying an AI / ML function may be communicated between network device 110 and terminal device 120.

[0051] In some embodiments, both single-sided and dual-sided AI / ML models can be explored. For example, single-sided AI / ML models can be used for beam management and positioning scenarios, while dual-sided AI / ML models can be used for CSI compression. For single-sided AI / ML models, the AI / ML model can be deployed at either the terminal device 120 or the network device 110. For dual-sided AI / ML models, a pair of AI / ML models is deployed at both the UE and the network.

[0052] Figure 2 An example AI / ML model deployment is described in which some example embodiments of the present disclosure may be implemented. In deployment 200, an AI / ML model on the terminal side or in the terminal device 120 (e.g., AI / ML encoders 210, 230, 250) is used for CSI compression, while an AI / ML model on the network side or in the network device 110 (e.g., AI / ML decoders 220, 240, 260) is used for CSI decompression.

[0053] In some embodiments, when AI / ML models are deployed only on the network side, AI / ML inference is performed by the network device 110, and the network device 110 can manage the AI / ML models with little or no impact on the specifications. In some embodiments, when AI / ML models are deployed on the terminal side, AI / ML model management is required to ensure that the terminal device 120 and the network device 110 have a common understanding of the occupancy of all AI / ML models.

[0054] In some embodiments, a single AI / ML model can correspond to a set of dedicated hardware resources, while different AI / ML models can correspond to separate hardware resources. High-performance terminal devices can deploy multiple AI / ML models for the same or different purposes to enable efficient AI / ML operations. In some embodiments, when hardware resources for an AI / ML model are in use at a given time instance, they cannot be used for other operations.

[0055] In some embodiments, different AI / ML models can be used for different use cases in different scenarios. For example, an AI / ML model used for spatial beam prediction may not be used for temporal beam prediction, and an AI / ML model may not be applicable to both low-speed and high-speed scenarios. Two AI / ML model management approaches are proposed: an AI / ML model-based approach and an AI / ML function-based approach.

[0056] Figure 3 The process of communicating with an AI / ML model according to some example embodiments of the present disclosure is described.

[0057] In process 300, the terminal device 120 sends (302) an indication of the number of processing units of the AI / ML model, or a plurality of AI / ML models of the AI / ML function 305, to the network device 110. Thus, the network device 110 can obtain basic information from the terminal device 120, such as the processing unit capabilities of the AI / ML model or the structure of the AI / ML model of the AI / ML function. Then, the network device 110 sends (308) a configuration 310 for operation to the terminal device 120. The operation can be providing CSI reporting, performing beam prediction, performing CSI compression, performing CSI prediction, or positioning. In this way, the network device 110 can trigger AI / ML operations in the terminal device 120, such as CSI reporting or beam prediction, and the operation can be flexible according to different use cases.

[0058] Figure 4This section describes AI / ML model-based AI / ML management according to some exemplary embodiments of the present disclosure. In this solution, different AI / ML models may have different input requirements for the same or different purposes. At least when training an AI / ML model on a terminal device 120, necessary information regarding the AI / ML model's input / output, usage, or applicable scenarios is reported in capability reports or registration information.

[0059] To save signaling overhead, the terminal device 120 may have multiple AI / ML processing units to operate the deployed AI / ML models, such as Figure 4 As described. Utilizing the concept of processing units for AI / ML reasoning, the terminal device 120 may further report the number of processing units used for AI / ML reasoning for each identified AI / ML model, each processing unit may be independently used for AI / ML reasoning, and all processing units may be simultaneously used for AI / ML reasoning. For example, in 400, the AI / ML model 410 requires AI / ML processing units 415, 420, 425, and the terminal device 120 may report to the network device 110 one of the following items of the model 410: the model input format, the model output format, the number of processing units of the AI / ML model 410, the purpose, or the application scenario. The AI / ML model 430 requires AI / ML processing units 435, 440, 445, and the terminal device 120 may report to the network device 110 one of the following items of the model 430: the model input format, the model output format, the number of processing units of the AI / ML model 430, the purpose, or the application scenario. In this way, the terminal device can report detailed capability and structure information of the AI / ML model or AI / ML function, making the processing unit resource management more efficient.

[0060] When terminal device 120 is configured with CSI measurements and / or CSI reporting associated with an AI / ML model, terminal device 120 can use any processing unit used for AI / ML inference to generate the corresponding CSI. This allows processing units to be used more efficiently. To more efficiently manage AI / ML models, terminal device 120 and network device 110 can have a shared understanding of the occupancy of all processing units used for AI / ML inference.

[0061] Figure 5 AI / ML management based on AI / ML functions according to some example embodiments of the present disclosure is described.

[0062] In some embodiments, different AI / ML models may correspond to different or the same AI / ML functions, such as Figure 5As described. The terminal device 120 can report one or more information items for each AI / ML function: purpose, applicable scenarios, input / output formats, and AI / ML models that may belong to the function. For example, in 500, there are AI / ML functions 510 and 530. AI / ML models 515, 520, and 525 belong to AI / ML function 510, while AI / ML models 535, 540, and 545 belong to AI / ML function 530. The terminal device 120 can report the model input format, model output format, purpose (e.g., CSI compression), applicable scenarios, and AI / ML models 515, 520, and 525 used for AI / ML function 510. The terminal device 120 can also report the model input format, model output format, purpose (e.g., CSI compression), applicable scenarios, and AI / ML models 535, 540, and 545 used for AI / ML function 530. In this way, network device 110 can obtain detailed capability and structure information for AI / ML functions from terminal device 120 and manage processing units more efficiently. In some embodiments, all AI / ML models belonging to the same function identifier have the same purpose, applicable scenarios, and the same input / output formats. This allows compatibility between AI / ML models within an AI / ML function.

[0063] In some embodiments, the CSI reporting configuration can be associated with the AI / ML function ID to inform the terminal device 120 that CSI is generated by AI / ML operation, and any AI / ML model belonging to the same AI / ML function can be used for CSI reasoning. This can achieve flexibility for the terminal device 120. Alternatively, the network device 110 can directly associate the AI / ML model ID with the CSI report to achieve the same purpose. In this way, the network device 110 can control all terminal devices in the cell, thereby performing cell-level optimization in the network device 110. In this case, the AI / ML model ID can only be associated with the CSI reporting configuration. The terminal device 120 and the network device 110 can have the same understanding of the occupancy of all AI / ML models deployed by the terminal device 120.

[0064] Regardless of the previous solution, the terminal device 120 and the network device 110 have the same knowledge of the occupancy of the AI / ML model and the processing unit. When there is no AI / ML model or processing unit available in the terminal device 120, the UE behavior can be specified.

[0065] In some embodiments, the number of AI / ML models or processing units occupied by the CSI report calculation can be analyzed as follows. Whether it is beam management or PMI / CQI reporting, the number of AI / ML models or processing units occupied by the CSI report can be different for different CSI types. The number of occupied processing units can be as follows. For CSI reports used for beam management and beam reports associated with AI / ML models or functions, an AI / ML model or processing unit with OAPU=1 is occupied. The CSI report of the CSI-ReportConfig with the high-level parameter reportQuantity set to "cri-RSRP", "ssb-Index-RSRP", "cri-SINR", "ssb-Index-SINR", "cri-RSRP-Index", "ssb-Index-RSRP-Index", "cri-SINR-Index", "ssb-Index-SINR-Index" corresponds to the CSI report for beam reporting. In this way, beam management is relatively simple in computation and can reduce the processing unit or AI / ML model occupancy. For CSI reports used for PMI / CQI reporting, the AI / ML model or processing unit occupies OAPU=Ks, where Ks is the number of CSI-RS resources in the CSI-RS resource set used for channel measurement. The CSI report of the CSI-ReportConfig with the higher-layer parameter reportQuantity set to "cri-RI-PMI-CQI", "cri-RI-i1", "cri-RI-i1-CQI", "cri-RI-CQI" or "cri-RI-LI-PMI-CQI" corresponds to the CSI report used for PMI / CQI reporting. As such, the PMI / CQI report calculation is computationally relatively complex, and accurate calculation can improve the efficiency of the processing unit or AI / ML model management.

[0066] In some embodiments, AI / ML models or processing unit occupancy criteria may be defined for different CSI report types.

[0067] Figure 6 Example AI / ML models or processing unit occupancy times according to some example embodiments of the present disclosure are described.

[0068] In 600, for periodic CSI reporting associated with AI / ML operation, one or more processing units or AI / ML models are occupied starting from the first symbol of the earliest one of each CSI-RS / CSI-IM / SSB resource used for channel measurement or interference measurement, respectively, the latest CSI-RS / CSI-IM / SSB timing no later than the corresponding CSI reference resource, which corresponds to the DL time slot in which the reference signal for the CSI report is received, until the last symbol of the configured PUSCH / PUCCH carrying the CSI report 630. The CSI-RS / CSI-IM / SSB resource can be the CSI-RS 610 corresponding to the nth CSI report in the CSI reference resource, 620 is the downlink time slot as the CSI reference resource, and the PUCCH carrying the beam report 630 can be the nth CSI report. The duration that the AI / ML model or processing unit is occupied for CSI calculation can be 640. Figure 6 The embodiment in

[15] provides an explanation of processing unit occupancy and uses "PUSCH / PUCCH carrying CSI reporting." This allows for allocation of processing units or AI / ML models when CSI-RS / CSI-IM / SSB resources arrive, avoiding allocation before CSI-RS / CSI-IM / SSB resources, thereby improving the efficiency of processing unit and AI / ML model usage.

[0069] In some embodiments, for semi-persistent (SP) CSI reports carried by the PUCCH activated by the medium access control control element (MAC CE) and SP CSI reports other than the initial SP CSI report on the PUSCH, one or more processing units are occupied starting from the first symbol of the earliest one used for channel measurement or interference measurement in each CSI-RS / CSI-IM / SSB resource, and the corresponding latest CSI-RS / CSI-IM / SSB opportunity is no later than the corresponding CSI reference resource (which corresponds to the DL time slot in which the reference signal for the CSI report is received), until the last symbol of the CSI report carried in the configured PUCCH, such as Figure 6 As shown. This improves the utilization efficiency of processing units and AI / ML models. Those skilled in the art will appreciate that the report can be any CSI report, such as PMI or RI, or CSI prediction, or positioning, etc. This makes reporting more flexible.

[0070] Figure 7A Another example AI / ML model or processing unit occupancy time according to some example embodiments of the present disclosure is described.

[0071] In some embodiments, in 700, for aperiodic CSI reporting associated with AI / ML operation, one or more processing units are occupied starting from the first symbol after the PDCCH 710 that triggers the CSI report until the last symbol of the scheduled PUSCH carrying the report 730. The CSI-RS / CSI-IM / SSB resources (e.g., CSI-RS 720) are located within the duration 740 that the AI / ML model or processing unit is occupied. In this way, the processing unit or AI / ML model can be allocated as soon as possible to meet the high priority of the aperiodic CSI reporting.

[0072] Figure 7B Further example AI / ML models or processing unit occupancy times according to some example embodiments of the present disclosure are described.

[0073] In some embodiments, at 750, for an initial SP CSI report carried on a PUSCH triggered by a DCI, one or more processing units are occupied starting from the first symbol after the PDCCH 760 until the last symbol of the scheduled PUSCH carrying the report 780. The CSI-RS / CSI-IM / SSB resources (e.g., CSI-RS 770) are within the duration 790 that the AI / ML model or processing unit is occupied. This allows for quick allocation of processing units or AI / ML models to meet the high priority and randomness of the initial SP CSI report. Those skilled in the art will appreciate that the report can be any CSI report, such as PMI or RI, or CSI prediction, or positioning, etc. This allows for more flexible reporting.

[0074] In some embodiments, the processing units are first-class processing units used for AI / ML reasoning of AI / ML models, and second-class processing units of the terminal device are occupied to provide input to the AI / ML model. Using different types of processing units, the generation of AI / ML model input and AI / ML reasoning can be performed more efficiently. When the terminal device 120 reports that the number of first-class processing units occupied for AI / ML model i is N APU,i , and when L first-type processing units are occupied for CSI inference in a given symbol (e.g., OFDM symbol), the terminal device 120 has N for the AI / ML model i APU,i -L unoccupied first-class processing units. If N CSI reports associated with AI / ML model i are in N APU,i -L APUs start occupying their respective processing units on the same OFDM symbol where each CSI report = 0, ..., N-1 corresponds to And 0≤M≤N is the maximum value, so The terminal device 120 then does not update the CSI reports for the NM requests with the lowest priority. Alternatively, when there are available second-class processing units for CSI reporting, the terminal device 120 reports the CSI reports for the NM requests with the lowest priority without performing AI / ML inference. When a CSI report is associated with an AI / ML model and there are available first-class processing units for the CSI report, but the terminal device 120 cannot provide the required AI / ML model input format, the terminal device 120 does not update the corresponding CSI, and no first-class processing units are occupied for the triggered CSI report.

[0075] Figure 8 An example report having both a first type of processing unit and a second type of processing unit is described according to some example embodiments of the present disclosure.

[0076] In some embodiments, CSI reporting is triggered for spatial beam prediction. Terminal device 120 is instructed to predict the best beam from the prediction beam set based on the measurement results of the measurement beam set. As described in 800, DCI 810 triggers aperiodic CSI reporting, and terminal device 120 receives CSI-RS resources 820 for the triggered CSI report. During CSI report calculation, terminal device 120 first obtains the L1-RSRP of all received CSI-RS resources in 830 using the second type of processing unit to obtain AI / ML model input. Each CSI-RS resource represents a beam. Using the input provided by the second type of processing unit, terminal device 120 further obtains the required CSI, namely the best K beams in the prediction beam set, and calculates the best K beam IDs and corresponding L1-RSRPs in the prediction beam set using AI / ML reasoning in 840 using the first type of processing unit. The beam report is carried in the PUSCH in 850 and sent to network device 110.

[0077] In some embodiments, the following potential situations exist. If there are available second-class processing units and available first-class processing units for the triggered CSI, the terminal device 120 can report the required CSI through AI / ML reasoning. The CSI report can occupy one or more second-class processing units. If there are no second-class processing units available for the CSI report, even if there are available first-class processing units, the terminal device 120 cannot obtain the AI / ML input required for AI / ML reasoning. The terminal device 120 will not update the corresponding CSI, and therefore the CSI report calculation may not occupy the first-class processing unit. In this way, when the second-class processing unit is unavailable, the resources of the first-class processing unit can be saved to improve the occupancy efficiency of the first-class processing unit and avoid waste.

[0078] In some embodiments, if a second processing unit is available for the triggered CSI, but no first processing unit is available, the following two options may be considered. In the first option, the terminal device 120 may report a set of CSI based on the output of the second processing unit, or CSI without AI / ML inference. The corresponding CSI report is updated with non-AI / ML CSI, and the CSI report occupies one or more second processing units. In the second option, the terminal device 120 does not update the corresponding CSI. The corresponding CSI report is not updated, and the CSI report calculation does not occupy the second processing unit. This allows flexibility when the first processing unit is not available.

[0079] Figure 9 An example report having a first type of processing unit is described according to some example embodiments of the present disclosure.

[0080] In some embodiments, Figure 9 An example of a CSI report with only AI / ML inference is described in . The same CSI report is triggered by DCI 910 and is associated with an AI / ML model. After receiving CSI-RS resources 920, the terminal device 120 calculates the CSI report 960 for PUSCH carrying using only one or more first-type processing units without using any second-type processing units. Figure 8 The AI / ML model in the AI / ML is different. The preparation of the AI / ML model input, or obtaining the L1 RSRP of each CSI-RS 940 is part of the AI / ML model and occupies the first type of processing unit, called the AI / ML pre-processing unit. The implementation of 940 may be different from Figure 8 830 in to match the structural differences between the first and second type processing units. In 950, the first type processing unit obtains the CSI for reporting via AI / ML reasoning. This can be the main processing unit. In other words, the AI / ML model can directly receive the transmitted CSI-RS or CSI / IM or SSB resources for AI / ML reasoning. The CSI report does not occupy any second type processing unit and only occupies one or more first processing units, depending on the CSI content. Therefore, if there is no available first processing unit for the triggered CSI report, the terminal device 120 does not update the corresponding CSI. In this way, the CSI report calculation can be performed without the resources of the second type processing unit and improve flexibility.

[0081] Figure 10 An example of a method implemented at the terminal device 120 according to some example embodiments of the present disclosure is described.

[0082] At 1010, the processor in the terminal device 120 sends an indication of the number of processing units used for the AI / ML model, or a plurality of AI / ML models used for the AI / ML function, to the network device 110 via the transceiver. At 1020, the processor in the terminal device 120 receives a configuration of terminal device operation associated with the AI / ML model or the AI / ML function from the network device via the transceiver, wherein the configuration is determined based on the number of processing units or the plurality of AI / ML models.

[0083] In some embodiments, a processing unit is defined for an AI / ML model. Multiple AI / ML models used for AI / ML functions have the same AI / ML input format, the same AI / ML output format, the same purpose, or the same applicable scenario. A processing unit for an AI / ML model can also be used for AI / ML reasoning of the AI / ML model.

[0084] In some embodiments, the operation includes one of: providing a channel state information (CSI) report, performing beam prediction, performing CSI compression, performing CSI prediction, or positioning. In some embodiments, when the CSI report is configured as a beam report, one processing unit for the AI / ML model, or one AI / ML model for the AI / ML function is occupied. In some embodiments, when the CSI report is configured to report at least one of a precoding matrix indicator (PMI) or a channel quality indicator (CQI), the number of occupied processing units for the AI / ML model or the number of occupied AI / ML models for the AI / ML function is the number of channel state information reference signal (CSI-RS) resources configured in the CSI resource set for channel measurement of the CSI report.

[0085] In some embodiments, from the first symbol of the earliest resource used for channel measurement or interference measurement to the last symbol of the physical uplink control channel (PUCCH) or physical uplink shared channel (PUSCH) carrying the CSI report, at least one processing unit is occupied when the CSI report is one of the following: periodic CSI report, semi-persistent (SP) CSI report carried by PUCCH, or SPCSI report other than the initial SP CSI report on the physical uplink shared channel (PUSCH). In some embodiments, from the first symbol after the PDCCH that triggers the CSI report to the last symbol of the PUSCH carrying the CSI report, at least one processing unit is occupied when the CSI report is one of the following: aperiodic CSI report or an initial SP CSI report triggered by downlink control information (DCI).

[0086] In some embodiments, when the number N of requested CSI reports is greater than the number M of unoccupied processing units, the processor in the terminal device 120 may skip updating the NM requested CSI reports with the lowest priority. Alternatively, when there are available CSI processing units for all NM requested CSI reports, the processor in the terminal device 120 may report the NM requested CSI reports with the lowest priority without performing AI / ML inference of the AI / ML model.

[0087] In some embodiments, the processing unit is a first type of processing unit for AI / ML inference of an AI / ML model; and a second type of processing unit occupied by the terminal device is used to provide input to the AI / ML model.

[0088] In some embodiments, when the second type of processing unit is available and the first type of processing unit is unavailable, the processor in the terminal device 120 performs the operation without AI / ML reasoning or skips performing the operation.

[0089] In some embodiments, if the second type of processing unit is unavailable, the processor in the terminal device 120 skips executing the operation. In some embodiments, the processing unit in the processing unit includes: a main processing unit for AI / ML inference of the AI / ML model; and a pre-processing unit for providing input to the AI / ML model.

[0090] In some embodiments, if the processing unit is unable to perform an operation, the processor in the terminal device 120 skips performing the operation. In some embodiments, the processor in the terminal device 120 sends the model input format, model output format, purpose, or applicable scenario of the AI / ML model or AI / ML function to the network device via the transceiver.

[0091] Figure 11 An example of a method implemented at the network device 110 according to some example embodiments of the present disclosure is described.

[0092] At 1110 , a processor in the network device 110 receives an indication of a number of processing units for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models for an AI / ML function, from a terminal device via a transceiver.

[0093] At 1120 , the processor in the network device 110 determines a configuration of the operation of the terminal device associated with the AI / ML model or AI / ML function based on the number of processing units or the plurality of AI / ML models.

[0094] At 1130 , the processor in the network device 110 sends the configuration to the terminal device via the transceiver.

[0095] In some embodiments, processing units are defined for AI / ML models. Multiple AI / ML models have the same AI / ML input format, the same AI / ML output format, the same purpose, or the same applicable scenarios. A processing unit for an AI / ML model can also be used for AI / ML reasoning of the AI / ML model.

[0096] In some embodiments, the operation includes one of providing a channel state information (CSI) report, performing beam prediction, performing CSI compression, performing CSI prediction, or positioning.

[0097] In some embodiments, where CSI reporting is configured for beam reporting, one processing unit of one AI / ML model for an AI / ML model or AI / ML function is occupied.

[0098] In some embodiments, when the CSI report is configured to report at least one of a precoding matrix indicator (PMI) report or a channel quality indicator (CQI) report, the number of occupied processing units for the AI / ML model or the number of occupied AI / ML models for the AI / ML function is the number of channel state information reference signal (CSI-RS) resources configured in the CSI resource set for channel measurement of the CSI report.

[0099] In some embodiments, at least one processing unit is occupied from the first symbol of the earliest resource used for channel measurement or interference measurement to the last symbol of the physical uplink control channel (PUCCH) or physical uplink shared channel (PUSCH) carrying the CSI report, when the CSI report is one of the following: periodic CSI reporting, semi-persistent (SP) CSI reporting carried by PUCCH, or SP CSI reporting other than the initial SP CSI report on the physical uplink shared channel (PUSCH).

[0100] In some embodiments, at least one processing unit is occupied from the first symbol after the PDCCH that triggers the CSI report to the last symbol of the PUSCH carrying the CSI report when the CSI report is one of the following: aperiodic CSI report, or initial SP CSI report triggered by downlink control information (DCI).

[0101] In some embodiments, the processor in the network device 110 receives a model input format, a model output format, a purpose, or an applicable scenario of an AI / ML model or an AI / ML function from a terminal device via a transceiver.

[0102] Figure 12A simplified block diagram of a device 1200 suitable for implementing embodiments of the present disclosure is depicted. The device 1200 may be viewed as Figure 1 and Figure 2 A further example implementation of the terminal device 120 and the network device 110 is shown. Thus, the device 1200 may be implemented on the terminal device 120 or the network device 110, or at least as a part of the terminal device 120 or the network device 110.

[0103] As shown, device 1200 includes a processor 1210, a memory 1220 coupled to processor 1210, a suitable transmitter (TX) and receiver (RX) 1240 coupled to processor 1210, and a communication interface coupled to TX / RX 1240. Memory 1210 stores at least a portion of program 1230. TX / RX 1240 is configured for bidirectional communication. TX / RX 1240 has at least one antenna to facilitate communication, although in practice, access nodes as described in this disclosure may have multiple antennas. The communication interface may represent any interface required for communication with other network elements, such as an X2 interface for bidirectional communication between eNBs, an S1 interface for communication between a Mobility Management Entity (MME) / Serving Gateway (S-GW) and an eNB, a Un interface for communication between an eNB and a relay node (RN), or a Uu interface for communication between an eNB and a terminal device.

[0104] Assume that the program 1230 includes program instructions that, when executed by the associated processor 1210, enable the device 1200 to operate in accordance with the embodiments of the present disclosure, as described herein with reference to Figure 1-11 The embodiments herein may be implemented by computer software, hardware, or a combination of software and hardware executable by the processor 1210 of the device 1200. The processor 1210 may be configured to implement various embodiments of the present disclosure. In addition, the combination of the processor 1210 and the memory 1220 may form a processing component 1250 suitable for implementing various embodiments of the present disclosure.

[0105] Memory 1220 can be of any type suitable for the local technology network and can be implemented using any suitable data storage technology, such as non-transitory computer-readable storage media, semiconductor-based storage devices, magnetic storage devices and systems, optical storage devices and systems, fixed memory, and removable memory, as non-limiting examples. Although only one memory 1220 is shown in device 1200, there may be multiple physically distinct memory modules in device 1200. Processor 1210 can be of any type suitable for the local technology network and can include one or more of a general-purpose computer, a special-purpose computer, a microprocessor, a digital signal processor (DSP), and a processor based on a multi-core processor architecture, as non-limiting examples. Device 1200 can have multiple processors, such as application-specific integrated circuit chips that are time-slave to a clock synchronized with a main processor.

[0106] In summary, the embodiments of the present disclosure can provide the following solutions.

[0107] Clause 1. A terminal device comprising: a processor; and a transceiver coupled to the processor, wherein the processor is configured to: send an indication of the number of processing units for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models for an AI / ML function, to a network device via the transceiver; and receive a configuration for operation of the terminal device associated with the AI / ML model or the AI / ML function from the network device via the transceiver, wherein the configuration is determined based on the number of processing units or the plurality of AI / ML models.

[0108] Clause 2. In the terminal device according to Clause 1, the processing unit is defined for an AI / ML model; multiple AI / ML models used for AI / ML functions have the same AI / ML input format, the same AI / ML output format, the same purpose, or the same applicable scenario; or the processing unit used for the AI / ML model can be simultaneously used for AI / ML inference of the AI / ML model.

[0109] Clause 3. The terminal device of clause 1, wherein the operation comprises one of: providing channel state information (CSI) reporting, performing beam prediction, performing CSI compression, performing CSI prediction, or positioning.

[0110] Clause 4. A terminal device according to clause 3, wherein, in case CSI reporting is configured for beam reporting, a processing unit for an AI / ML model or an AI / ML model for an AI / ML function is occupied.

[0111] Clause 5. For a terminal device according to Clause 3, when the CSI report is configured to report at least one of a precoding matrix indicator (PMI) or a channel quality indicator (CQI), the number of occupied processing units for the AI / ML model or the number of occupied AI / ML models for the AI / ML function is the channel measurement for the CSI report, the number of channel state information reference signal (CSI-RS) resources configured in the CSI resource set.

[0112] Clause 6. A terminal device according to clause 3, wherein at least one of the processing units is occupied from the first symbol of the earliest resource used for channel measurement or interference measurement to the last symbol of the physical uplink control channel (PUCCH) or physical uplink shared channel (PUSCH) carrying the CSI report, when the CSI report is one of the following: a periodic CSI report, a semi-persistent (SP) CSI report carried by the PUCCH, or an SP CSI report other than an initial SP CSI report on the physical uplink shared channel (PUSCH).

[0113] Clause 7. A terminal device according to clause 3, wherein at least one of the processing units is occupied from the first symbol after the PDCCH that triggers the CSI report to the last symbol of the PUSCH carrying the CSI report when the CSI report is one of the following: a non-periodic CSI report, or an initial SPCSI report triggered by downlink control information (DCI).

[0114] Clause 8. A terminal device according to clause 3, wherein the processor is further configured to: skip updating NM requested CSI reports with the lowest priority when the number of requested CSI reports N is greater than the number of unoccupied processing units M; or report the NM requested CSI reports with the lowest priority when there are available CSI processing units for all NM requested CSI reports in the absence of AI / ML inference of the AI / ML model.

[0115] Clause 9. A terminal device according to clause 3, wherein the processor is further configured to: skip updating the CSI report if the processing unit is available for CSI reporting but the terminal device is unable to provide input to the AI / ML model.

[0116] Clause 10. The terminal device according to clause 1, wherein the processing unit is a first type of processing unit for AI / ML inference of the AI / ML model; and a second type of processing unit occupied by the terminal device for providing input to the AI / ML model.

[0117] Clause 11. The terminal device according to clause 10, wherein the processor is further configured to: perform the operation without AI / ML reasoning or skip performing the operation when the second type of processing unit is available and the first type of processing unit is unavailable.

[0118] Clause 12. The terminal device according to clause 10, wherein the processor is further configured to: skip executing the operation if the second type of processing unit is unavailable.

[0119] Clause 13. A terminal device according to clause 1, wherein a processing unit in the processing unit includes: a main processing unit for AI / ML inference of an AI / ML model; and a preprocessing unit for providing input to the AI / ML model.

[0120] Clause 14. The terminal device according to clause 13, wherein the processor is further configured to: skip performing the operation if the processing unit is not available for the operation.

[0121] Clause 15. A terminal device according to clause 1, wherein the processor is further configured to: send a model input format, a model output format, a purpose or an applicable scenario of the AI / ML model or AI / ML function to a network device via a transceiver.

[0122] Clause 16. A network device comprising: a processor; and a transceiver coupled to the processor, wherein the processor is configured to: receive an indication of the number of processing units for an artificial intelligence (AI) / machine learning (ML) model or a plurality of AI / ML models for an AI / ML function from a terminal device via the transceiver; determine a configuration for operation of the terminal device associated with the AI / ML model or the AI / ML function based on the number of processing units or the plurality of AI / ML models; and send the configuration to the terminal device via the transceiver.

[0123] Clause 17. In a network device according to Clause 16, the processing unit is defined for an AI / ML model; multiple AI / ML models have the same AI / ML input format, the same AI / ML output format, the same purpose, or the same applicable scenario; or the processing unit for the AI / ML model can also be used for AI / ML inference of the AI / ML model.

[0124] Clause 18. The network device of clause 16, wherein the operation comprises one of providing channel state information (CSI) reporting, performing beam prediction, performing CSI compression, performing CSI prediction, or positioning.

[0125] Clause 19. A network device according to clause 18, wherein, in case CSI reporting is configured for beam reporting, one processing unit for an AI / ML model or one AI / ML model for an AI / ML function is occupied.

[0126] Clause 20. A network device according to clause 18, wherein, in a case where the CSI report is configured to report at least one of a precoding matrix indicator (PMI) report or a channel quality indicator (CQI) report, the number of occupied processing units for the AI / ML model or the number of occupied AI / ML models for the AI / ML function is the number of channel state information reference signal (CSI-RS) resources configured in the CSI resource set for channel measurement of the CSI report.

[0127] Clause 21. A network device according to clause 18, in the case where the CSI report is one of the following: a periodic CSI report, a semi-persistent (SP) CSI report carried by a PUCCH, or an SP CSI report other than an initial SP CSI report on a physical uplink shared channel (PUSCH), at least one processing unit is occupied from the first symbol of the earliest resource used for channel measurement or interference measurement to the last symbol of the physical uplink control channel (PUCCH) or physical uplink shared channel (PUSCH) carrying the CSI report.

[0128] Clause 22. A network device according to clause 18, wherein at least one of the processing units is occupied from the first symbol after the PDCCH that triggers the CSI report to the last symbol of the PUSCH that carries the CSI report, in case the CSI report is one of the following: an aperiodic CSI report, or an initial SP CSI report triggered by downlink control information (DCI).

[0129] Clause 23. The network device according to Clause 16, wherein the processor is further configured to: receive a model input format, a model output format, a purpose, or an applicable scenario of an AI / ML model or an AI / ML function from a terminal device via a transceiver.

[0130] Clause 24. A method performed by a terminal device, comprising: sending an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function, to a network device; and receiving from the network device a configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function, wherein the configuration is determined based on the number of processing units or the plurality of AI / ML models.

[0131] Clause 25. A method performed by a network device, comprising: receiving from a terminal device an indication of a number of processing units for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models for an AI / ML function; determining, based on the number of processing units or the plurality of AI / ML models, a configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function; and sending the configuration to the terminal device.

[0132] Clause 26. A non-transitory computer-readable medium having program instructions stored thereon, which, when executed by an apparatus, causes the apparatus to at least: send an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function, to a network device; and receive from the network device a configuration for operation of a terminal device associated with the AI / ML model or the AI / ML function, wherein the configuration is determined based on the number of processing units or the number of the plurality of AI / ML models.

[0133] Clause 27. A non-transitory computer-readable medium having program instructions stored thereon, which, when executed by an apparatus, causes the apparatus to at least: receive from a terminal device an indication of the number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function; determine, based on the number of processing units or the plurality of AI / ML models, a configuration for the operation of the terminal device associated with the AI / ML model or the AI / ML function; and send the configuration to the terminal device.

[0134] The solution disclosed herein introduces a signaling mechanism that enables terminal devices to indicate the usage status of TOs to network devices, enabling the network devices to adjust subsequent resource allocations, avoiding waste of unused TOs and improving communication performance. The special design of the indication information format disclosed herein helps reduce signaling overhead, for example, by using a limited number of bits to indicate the usage of TOs associated with one or more CG configurations.

[0135] In general, the various embodiments of the present disclosure may be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. Some aspects may be implemented in hardware, while other aspects may be implemented in firmware or software, which may be executed by a controller, microprocessor, or other computing device. Although various aspects of the embodiments of the present disclosure are illustrated and described in terms of block diagrams, flow charts, or other graphical representations, it will be understood that, as non-limiting examples, the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuits or logic, general-purpose hardware or a controller or other computing device, or some combination thereof.

[0136] The present disclosure also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as those contained in program modules, which are executed in a device on a target real or virtual processor to perform the above-described process or method. Generally speaking, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or split between program modules as needed. The machine-executable instructions of the program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.

[0137] The program code for executing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that, when executed by the processor or controller, the program code implements the functions / operations specified in the flow chart and / or block diagram. The program code can be executed entirely on the machine, partially on the machine, as a separate software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0138] The program code described above may be embodied on a machine-readable medium, which may be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include an electrical connection having one or more conductors, a portable computer floppy disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0139] In addition, although the operations are described in a particular order, this should not be construed as requiring that the operations be performed in the particular order shown or in a sequential order, or that all of the described operations be performed to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. Similarly, although the above discussion includes some specific implementation details, these details should not be construed as limitations on the scope of this disclosure, but rather as descriptions of features that may be unique to a particular embodiment. Certain features described in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented separately in multiple embodiments or in any suitable subcombination.

[0140] Although the present disclosure is described in language specific to structural features and / or methodological acts, it should be understood that the present disclosure defined in the appended claims is not necessarily limited to the specific features or acts described above. Instead, the specific features and acts described above are disclosed as example forms of implementing the claims.

Claims

1. A terminal device, comprising: processor; as well as a transceiver coupled to the processor, wherein the processor is configured to: sending, via the transceiver, to a network device an indication of a number of processing units for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models for an AI / ML function; and A configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function is received from the network device via the transceiver, wherein the configuration is determined based on the number of the processing units or the plurality of AI / ML models.

2. The terminal device according to claim 1, wherein: The processing unit is defined for the AI / ML model; The multiple AI / ML models used for the AI / ML function have the same AI / ML input format, the same AI / ML output format, the same purpose, or the same applicable scenario; or The processing unit used for the AI / ML model can also be used for AI / ML reasoning of the AI / ML model.

3. The terminal device according to claim 1, wherein the operation comprises one of the following: Provides channel state information (CSI) reporting, Perform beam prediction, Perform CSI compression, Perform CSI prediction, or position.

4. The terminal device according to claim 3, wherein, when the CSI report is configured for beam reporting, one processing unit for the AI / ML model or one AI / ML model for the AI / ML function is occupied.

5. The terminal device according to claim 3, wherein, when the CSI report is configured to report at least one of a precoding matrix indicator (PMI) or a channel quality indicator (CQI), the number of processing units occupied by the AI / ML model or the number of AI / ML models occupied by the AI / ML function is a channel measurement for the CSI report and a number of channel state information reference signal (CSI-RS) resources configured in a CSI resource set.

6. The terminal device of claim 3 , wherein, when the CSI report is one of the following, at least one of the processing units is occupied from a first symbol of an earliest resource used for channel measurement or interference measurement to a last symbol of a physical uplink control channel (PUCCH) or a physical uplink shared channel (PUSCH) carrying the CSI report: Periodic CSI reports, Semi-persistent (SP) CSI reporting carried by the PUCCH, or SP CSI reporting in addition to the initial SP CSI reporting on the Physical Uplink Shared Channel (PUSCH).

7. The terminal device according to claim 3, wherein when the CSI report is one of the following, at least one of the processing units is occupied from the first symbol after the PDCCH that triggers the CSI report to the last symbol of the PUSCH that carries the CSI report: Aperiodic CSI reporting, or Initial SP CSI reporting triggered by Downlink Control Information (DCI).

8. The terminal device according to claim 3, wherein the processor is further configured to: In case the number N of requested CSI reports is greater than the number M of unoccupied processing units, Skip updating the CSI reports for the NM requests with the lowest priority; or When there are available CSI processing units for all of the NM requested CSI reports, reporting the NM requested CSI reports with the lowest priority without AI / ML inference of the AI / ML model.

9. The terminal device according to claim 3, wherein the processor is further configured to: In case the processing unit is available for the CSI report but the terminal device is unable to provide input to the AI / ML model, updating the CSI report is skipped.

10. The terminal device according to claim 1, wherein: The processing unit is a first type of processing unit used for AI / ML inference of the AI / ML model; and The second type of processing unit of the terminal device is occupied to provide input to the AI / ML model.

11. The terminal device according to claim 10, wherein the processor is further configured to: In a case where the second type of processing unit is available and the first type of processing unit is unavailable, performing the operation without the AI / ML inference or skipping the operation.

12. The terminal device according to claim 10, wherein the processor is further configured to: In case the second type of processing unit is not available, performing the operation is skipped.

13. The terminal device according to claim 1, wherein one of the processing units comprises: A main processing unit for AI / ML inference of the AI / ML model; as well as A preprocessing unit, configured to provide input to the AI / ML model.

14. The terminal device according to claim 13, wherein the processor is further configured to: In case the processing unit is not available for the operation, execution of the operation is skipped.

15. The terminal device according to claim 1, wherein the processor is further configured to: The AI / ML model or the model input format, model output format, purpose, or applicable scenario of the AI / ML function is sent to the network device via the transceiver.

16. A network device comprising: processor; as well as a transceiver coupled to the processor, wherein the processor is configured to: receiving, via the transceiver, from a terminal device an indication of a number of processing units for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models for an AI / ML function; Determining, based on the number of processing units or the plurality of AI / ML models, a configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function; and The configuration is sent to the terminal device via the transceiver.

17. The network device according to claim 16, wherein: The processing unit is defined for the AI / ML model; The multiple AI / ML models have the same AI / ML input format, the same AI / ML output format, the same purpose, or the same applicable scenario; or The processing unit used for the AI / ML model can also be used for AI / ML reasoning of the AI / ML model.

18. The network device of claim 16, wherein the operation comprises one of: Provides channel state information (CSI) reporting, Perform beam prediction, Perform CSI compression, Perform CSI prediction, or position.

19. The network device of claim 18, wherein, in a case where the CSI report is configured for beam reporting, one processing unit for the AI / ML model or one AI / ML model for the AI / ML function is occupied.

20. The network device according to claim 18, wherein, when the CSI report is configured to report at least one of a precoding matrix indicator (PMI) report or a channel quality indicator (CQI) report, the number of processing units occupied by the AI / ML model or the number of AI / ML models occupied by the AI / ML function is a channel measurement for the CSI report, a number of channel state information reference signal (CSI-RS) resources configured in a CSI resource set.

21. The network device of claim 18 , wherein, if the CSI report is one of the following, at least one of the processing units is occupied from a first symbol of an earliest resource used for channel measurement or interference measurement to a last symbol of a physical uplink control channel (PUCCH) or a physical uplink shared channel (PUSCH) carrying the CSI report: Periodic CSI reports, Semi-persistent (SP) CSI reporting carried by the PUCCH, or SP CSI reporting in addition to the initial SP CSI reporting on the Physical Uplink Shared Channel (PUSCH).

22. The network device of claim 18 , wherein, if the CSI report is one of the following, at least one of the processing units is occupied from a first symbol after a PDCCH triggering the CSI report to a last symbol of a PUSCH carrying the CSI report: Aperiodic CSI reporting, or Initial SP CSI reporting triggered by Downlink Control Information (DCI).

23. The network device of claim 16, wherein the processor is further configured to: The AI / ML model or the model input format, model output format, purpose, or applicable scenario of the AI / ML function is received from the terminal device via the transceiver.

24. A method performed by a terminal device, comprising: Sending an indication to a network device of a number of processing units to use for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models to use for an AI / ML function; and A configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function is received from the network device, wherein the configuration is determined based on the number of the processing units or the plurality of AI / ML models.

25. A method performed by a network device, comprising: receiving, from the terminal device, an indication of a number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function; Determining, based on the number of processing units or the plurality of AI / ML models, a configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function; and Send the configuration to the terminal device.

26. A non-transitory computer-readable medium having program instructions stored thereon, which, when executed by a device, causes the device to at least: Sending an indication to a network device of a number of processing units to use for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models to use for an AI / ML function; and A configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function is received from the network device, wherein the configuration is determined based on the number of the processing units or the plurality of AI / ML models.

27. A non-transitory computer-readable medium having program instructions stored thereon, which, when executed by a device, causes the device to at least: receiving, from the terminal device, an indication of a number of processing units used for an artificial intelligence (AI) / machine learning (ML) model, or a plurality of AI / ML models used for an AI / ML function; Determining, based on the number of processing units or the plurality of AI / ML models, a configuration for an operation of the terminal device associated with the AI / ML model or the AI / ML function; and Send the configuration to the terminal device.

Images