Communication method and communication device
By merging CSI reports and optimizing resource utilization, the computational resource bottleneck in the CSI measurement and feedback process under the AI/ML background was resolved, achieving efficient CSI report processing and system performance improvement.
Patent Information
- Application Number
- PCT/CN2025/094046
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-05-09
- Publication Date
- 2026-11-12
AI Technical Summary
When AI/ML technology is applied to the CSI measurement and feedback process, existing protocols do not fully consider the differences in computing resources, resulting in processing bottlenecks on the user equipment side and affecting the timeliness and accuracy of CSI reporting.
By merging multiple CSI reports and/or associating a set of CSI resources with multiple reports, resource utilization is optimized, the number of CSI reports and reporting priorities are flexibly determined, computing resource allocation is dynamically adjusted, a multi-task shared measurement set mechanism is introduced, and the occupancy time of processing units and timing design under the hybrid architecture are defined.
It improves the processing efficiency and system performance of CSI reports, reduces resource consumption and overhead, ensures the efficient reporting of both AI-related and non-AI-related CSI reports, and improves system resource utilization and communication performance.
Smart Images

Figure CN2025094046_12112026_PF_FP_ABST
Abstract
Description
Communication methods and communication equipment Technical Field
[0001] This application relates to the field of mobile communication technology, specifically to a communication method and communication device. Background Technology
[0002] In massively multi-input multiple-output (MIMO) communication systems, the measurement and feedback of Channel State Information (CSI) is a crucial technology in the physical layer. Accurate CSI acquisition helps base stations select beamforming, modulation, and coding strategies, thereby improving spectrum utilization and system performance. Existing CSI feedback mechanisms are primarily designed and optimized based on traditional rules and statistical models, making it difficult to cope with the complex changes in dynamic wireless environments. With the development of Artificial Intelligence / Machine Learning (AI / ML) technologies, more and more communication systems are beginning to incorporate AI / ML mechanisms into CSI-related processing to achieve higher accuracy and lower latency CSI prediction and compression. However, AI / ML algorithms themselves place higher demands on computing resources, especially when CSI measurement, processing, and reporting are performed at the User Equipment (UE) end, consuming significant CSI Processing Unit (CPU) resources and processing time. Because existing protocols do not fully consider the differences in computational load for AI / ML-related CSI processing, this may lead to processing bottlenecks at the UE end, affecting the timeliness and accuracy of CSI reporting. Therefore, jointly optimizing the processing resources (such as CPU usage) and computational latency for CSI measurement and feedback processes in the context of AI / ML has become an important requirement for realizing intelligent communication systems. Summary of the Invention
[0003] This application provides a communication method and a communication device.
[0004] This application provides a communication method executed by a user equipment, including: reporting the capability information of the user equipment to a base station; receiving channel state information (CSI) report configuration and / or CSI resource configuration sent by the base station, wherein the CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signal configurations, the CSI report configuration includes a configuration for merging CSI reports corresponding to multiple tasks into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports.
[0005] Through the above technical solution, this communication method can reduce reporting overhead and resource consumption, and improve the system's computing and communication efficiency by merging CSI reports from multiple tasks into one report and / or associating a set of CSI resources with multiple CSI reports.
[0006] This application provides a communication method executed by a user equipment, comprising: reporting the capability information of the user equipment to a base station; receiving channel state information (CSI) report configuration and / or CSI resource configuration sent by the base station; determining the number of CSI reports that the user equipment can process, and generating and sending CSI reports according to the processing results.
[0007] Through the above technical solution, this communication method can flexibly determine the number of CSI reports based on the processing capacity of the user equipment, optimize resource utilization, and improve the processing efficiency of CSI reports and system performance.
[0008] This application provides a communication method executed by a user equipment, comprising: determining the reporting priority of multiple Channel State Information (CSI) reports; and reporting the CSI reports to a base station according to the reporting priority of the CSI reports. The CSI reports include AI-related CSI reports and non-AI-related CSI reports. The user equipment determines the reporting priority of the CSI reports based on whether the CSI reports carry Physical Layer Reference Signal Received Power (L1-RSRP) or Physical Layer Signal Interference-Noise Ratio (L1-SINR). The CSI reports are performance monitoring results or intermediate values regarding whether the AI-related CSI reports carry the L1-RSRP or L1-SINR, and / or based on the reporting time-domain behavior and bearer type of the CSI reports.
[0009] Through the above technical solution, the communication method of this embodiment flexibly adjusts the transmission order of CSI reports based on the reporting priority, reporting time domain behavior and bearer type of CSI reports, ensuring that AI-related and non-AI-related CSI reports can be reported efficiently according to the importance of the task, thereby improving system resource utilization and communication performance.
[0010] This application provides a communication method executed by a base station, comprising: receiving multiple Channel State Information (CSI) reports reported by a user equipment, wherein the CSI report configuration indicates that the multiple CSI reports are associated with the same or partially the same set of measurement reference signal configurations, the CSI report configuration includes a configuration for merging multiple task-related CSI reports into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports.
[0011] Through the above technical solution, this communication method can reduce reporting overhead and resource consumption, and improve the system's computing and communication efficiency by merging CSI reports from multiple tasks into one report and / or associating a set of CSI resources with multiple CSI reports.
[0012] This application provides a communication method executed by a base station, comprising: receiving capability information of a user equipment reported by the user equipment; sending Channel State Information (CSI) report configuration and / or CSI resource configuration to the user equipment; and receiving a CSI report reported by the user equipment, wherein the CSI report is based on the number of CSI reports that the user equipment can process.
[0013] Through the above technical solution, this communication method can flexibly determine the number of CSI reports based on the processing capacity of the user equipment, optimize resource utilization, and improve the processing efficiency of CSI reports and system performance.
[0014] This application provides a communication method executed by a base station, comprising: receiving multiple Channel State Information (CSI) reports reported by a user equipment, wherein the CSI reports are based on a reporting priority. The CSI reports include AI-related CSI reports and non-AI-related CSI reports. The user equipment determines the reporting priority of the CSI reports based on whether the CSI reports carry Physical Layer Reference Signal Received Power (L1-RSRP) or Physical Layer Signal-to-Interference-Noise Ratio (L1-SINR). The CSI reports are performance monitoring results or intermediate values regarding whether the AI-related CSI reports carry the L1-RSRP or L1-SINR, and / or based on the reporting time-domain behavior and bearer type of the CSI reports.
[0015] Through the above technical solution, the communication method of this embodiment dynamically adjusts the transmission order of CSI reports based on the reporting priority of CSI reports, ensuring that AI-related and non-AI-related CSI reports can be reported efficiently according to the importance of the task, thereby improving system resource utilization and communication performance.
[0016] This application provides a communication device including a processor and a memory. The memory is used to store computer programs, and the processor is used to call and run the computer programs stored in the memory to perform the methods described above.
[0017] This application provides a user equipment including a processor and a memory. The memory stores computer programs, and the processor calls and runs the computer programs stored in the memory to perform the methods described above.
[0018] This application provides a base station including a processor and a memory. The memory stores computer programs, and the processor calls and runs the computer programs stored in the memory to perform the methods described above.
[0019] This application provides a network element including a processor and a memory. The memory stores computer programs, and the processor calls and runs the computer programs stored in the memory to perform the methods described above.
[0020] This application provides a chip for implementing the above-described method.
[0021] Specifically, the chip includes a processor for calling and running a computer program from a memory, causing a device on which the chip is installed to perform the methods described above.
[0022] This application provides a computer-readable storage medium for storing a computer program that causes a computer to perform the above-described method.
[0023] This application provides a computer program product including computer program instructions that cause a computer to execute the above-described method. Attached Figure Description
[0024] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0025] Figure 1 is a schematic diagram of a general AI / ML architecture in a wireless communication system;
[0026] Figure 2 is a schematic diagram of a wireless communication system architecture provided in an embodiment of this application;
[0027] Figure 3A is a flowchart illustrating a communication method executed by a user equipment according to an embodiment of this application;
[0028] Figure 3B is a schematic flowchart of a communication method performed by a base station according to an embodiment of this application;
[0029] Figure 3C is a schematic flowchart of a communication method performed by a user equipment according to an embodiment of this application;
[0030] Figure 3D is a flowchart illustrating a communication method executed by a base station according to an embodiment of this application;
[0031] Figure 3E is a flowchart illustrating a communication method provided in an embodiment of this application;
[0032] Figure 4A is a schematic diagram of a combined serial and parallel CSI report processing provided in an embodiment of this application;
[0033] Figure 4B is a schematic diagram of a combined serial and parallel CSI report processing provided in an embodiment of this application;
[0034] Figure 5A is a schematic diagram showing the time occupied by an artificial intelligence processing unit (AI Processing Unit, such as APU) according to an embodiment of this application, including data collection and model inference.
[0035] Figure 5B is a schematic diagram of an APU occupancy time that only includes model inference, provided by an embodiment of this application;
[0036] Figure 5C is a schematic diagram of an APU occupancy time that only includes model inference, provided by an embodiment of this application;
[0037] Figure 5D is a schematic diagram showing that the APU occupancy time only includes model inference, according to an embodiment of this application.
[0038] Figure 6A is a schematic diagram showing a CSI report located after the prediction results according to an embodiment of this application;
[0039] Figure 6B is a schematic diagram showing a CSI report located before the prediction results according to an embodiment of this application;
[0040] Figure 7A is a flowchart illustrating a communication method executed by a user equipment according to an embodiment of this application;
[0041] Figure 7B is a schematic flowchart of a communication method performed by a base station according to an embodiment of this application;
[0042] Figure 8 is a schematic structural diagram of a wireless communication device provided in an embodiment of this application;
[0043] Figure 9 is a schematic structural diagram of a chip according to an embodiment of this application;
[0044] Figure 10 is a schematic block diagram of a wireless communication system provided in an embodiment of this application. Detailed Implementation
[0045] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.
[0046] In the framework of massive MIMO communication systems, CSI measurement and feedback is one of the core technologies in the physical layer. Accurate CSI measurement and feedback can help the base station select appropriate data modulation and coding methods, thereby improving the spectral efficiency of the communication system.
[0047] With the rapid development of AI technology, the 3rd Generation Partnership Project (3GPP) began researching AI-based CSI prediction technology in Release 18 (Rel-18). However, due to time constraints, Rel-18 was unable to complete the research. Therefore, in Rel-19, 3GPP continued to advance research on AI-based CSI prediction technology, and finally decided at the RAN#105 meeting to upgrade it from a Study Item (SI) to a Work Item (WI).
[0048] Based on the results of the Rel-18 discussion, a general AI / ML architecture for wireless communication systems is shown in Figure 1. As described in Figure 1, the management function module includes monitoring capabilities for models / functionalities. In Rel-18, the main functions considered for implementation via AI / ML include CSI prediction and compression, beam management, and localization. Currently, regarding the application of AI / ML in these functions, the standard has reached the following consensus on monitoring AI / ML models / functionalities: the purpose of researching AI / ML model monitoring should include at least the following aspects: model activation, deactivation, selection, switching, rollback, and updating (including retraining).
[0049] Rel-18 investigated AI / ML-based beam management, CSI prediction, and CSI compression technologies, and plans to promote their standardization in Rel-19 and subsequent versions. AI / ML-based beam management can be further divided into two categories: one is spatial beam prediction based on measurement results of beam set B to predict the optimal beam in beam set A; the other is temporal beam prediction based on historical measurement results of beam set B to predict the optimal beam in beam set A. The overall process of AI / ML-based beam management uses the measurement results of beam set B as model input, and outputs data such as the probability of each beam in beam set A becoming the optimal beam or the predicted Reference Signal Received Power (RSRP) value to predict one or more optimal beams in beam set A.
[0050] AI / ML-based CSI prediction can effectively prevent channel aging. The process uses historical CSI information as model input, and the model can output predicted CSI at multiple future time points, thereby obtaining channel status in advance and improving the system's predictive capability and communication reliability.
[0051] AI / ML-based CSI compression improves CSI accuracy while reducing the air interface resource overhead of CSI reporting. Its space-frequency domain compression process is based on an encoder-decoder model architecture. The encoder is deployed on the UE side, using space-frequency domain channel information (e.g., precoding matrix, feature vector) obtained from channel measurements as model input and outputting compressed channel information. The decoder is deployed on the base station side, using the compressed channel information as input and outputting restored complete channel information.
[0052] A quantizer can be deployed at the back end of the encoder to further reduce the transmission burden of CSI feedback, forming a complete CSI generation part together with the encoder; correspondingly, a dequantizer can be deployed at the front end of the decoder to restore the compressed channel information before quantization, forming a CSI reconstruction part together with the decoder.
[0053] The introduction of AI / ML will have a significant impact on CSI design. To ensure consistency between base stations and UEs in CSI measurement and reporting behavior, the design of CPU resource allocation and CSI computation time for AI / ML needs to be considered simultaneously.
[0054] This study investigates CSI processing criteria and timing based on AI / ML-based CSI prediction use cases. Starting with the traditional CSI feedback mechanism "typeII-Doppler-r18" in the terminal-side model inference scenarios, it examines whether CSI processing criteria and timing need modification to assess the potential impact on specifications. The study then extends to AI / ML beam management use cases, exploring whether traditional CSI reporting and AI / ML-based CSI reporting should share CPU resources. Using the existing CPU mechanism as a starting point, it further investigates whether traditional CSI and AI / ML functions should be counted separately or share CPU resources in the terminal-side model, and whether this mechanism is applicable to Case 1 and Case 2 in beam management use cases.
[0055] An APU (Average Processing Unit) is introduced as the basic unit for measuring AI-related CSI processing resources within the general AI / ML framework. An APU is used to quantify the parallel processing capability of at least multiple AI-related CSI reports, applicable to functions such as CSI compression, CSI prediction, spatial beamforming, and temporal beamforming. Furthermore, in the CSI prediction use case, research directions for CSI processing criteria and timing are further clarified, including: whether traditional CSI and AI / ML CSI share CPUs; whether different AI / ML functions share processing units; whether model inference requires the introduction of new processing timings; and whether existing CSI-RS resources and port counting frameworks can be reused.
[0056] The paper discusses CSI processing guidelines and timing in more detail, focusing on use cases such as AI / ML beam management, CSI prediction, and CSI compression. At least one processing resource configuration option is discussed: first, using only the APU without using the CPU; second, using only traditional CPU resources; and third, using both APU and traditional CPU resources. UEs can report the supported options through capability information, and also need to report the total number of supported APUs. These mechanisms provide a directional foundation for future standardized CSI processing under AI / ML.
[0057] Regarding the CPU, the UE indicates the amount of parallel CSI computation N that can be supported on a single carrier via the parameter simultaneousCSI-ReportsPerCC or across carriers via the parameter simultaneousCSI-ReportsAllCC. CPU If the UE supports N CPU Parallel CSI computational complexity means that N UEs exist on a given OFDM symbol. CPU There are L CSI processing units used to process CSI reports. If, on a given OFDM symbol, L CPUs are occupied for CSI report calculations, the UE will have N CPUs remaining. CPU -L unused CPUs. If N CSIs report that they intend to occupy N CPUs on the same OFDM symbol. CPU -L unused CPUs, where each CSI reports n=0,…,N-1 corresponding to one CPU utilization. Then the UE does not require the NM lowest priority CSI reports to be updated, where 0≤M≤N is satisfied. The maximum value.
[0058] The CPU usage of CSI reporting processing on an OFDM symbol segment is shown below:
[0059] If the high-level parameter reportQuantity is set to 'none' and CSI-RS-ResourceSet is configured as 'trs-Info', O CPU =0.
[0060] If the high-level parameter `reportQuantity` is 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' and `CSI-RS-ResourceSet` is not configured as 'trs-Info', O CPU =1.
[0061] If the high-level parameter `reportQuantity` is set to 'cri-RI-PMI-CQI' and `codebookType` is set to 'typeII-Doppler-r18' or 'typeII-Doppler-PortSelection-r18': If the CSI-RS resource set used for channel measurements is aperiodic and K CSI-RS resources are configured, then when K=12, O CPU =8, K<12 when O CPU =Y1·K, where Y1∈{1,2,3} is indicated by UE capability; if the CSI-RS resource set used for channel measurement is periodic or semi-persistent and one CSI-RS resource is configured, when N4=1, O CPU =4, O when N4>1 CPU =max(Y2·N4,4), where N4 is configured by the higher-level parameter N4, and Y2∈{1,2,3} is indicated by the UE capability.
[0062] If the enhanced Type II codebook or enhanced Type II port selection codebook is reported, O CPU =K s K s This refers to the number of CSI-RS resources in the CSI-RS resource set used for channel measurements.
[0063] Regarding the CSI processing timeline, if Downlink Control Information (DCI) triggers CSI transmission on the Physical Uplink Shared Channel (PUSCH), the UE will perform the nth CSI report triggered by the DCI if at least one of the following conditions is met:
[0064] The first uplink symbol reported by the CSI triggered by this DCI is no earlier than symbol Z, provided that timing advance (TA) is included. ref .
[0065] The first uplink symbol carrying the nth CSI report is no earlier than symbol Z' if it includes TA. ref (n).
[0066] The above Z ref Defined as at least T after the last symbol of DCI proc,CSI =(Z)(2048+144)·κ2 -μ ·T C +T switch The first rising symbol, Z' ref (n) is defined as the periodic CSI-RS resource used for the most recent channel measurement corresponding to the nth CSI report, at least T' after the last symbol. proc,CSI =(Z')(2048+144)·κ2 -μ ·T C The first upward symbol.
[0067] When the reportQuantity IE is set to 'cri-RSRP', 'ssb-Index-RSRP', 'cri-RSRP-Index', or 'ssb-Index-RSRP-Index', the value of (Z,Z') is (Z3,Z′3), where X μ KB is obtained based on the UE's reporting capability beamReportTiming. l Obtained based on the UE's reporting capability beamSwitchTiming.
[0068] In scenarios involving reporting enhanced Type II codebooks or enhanced Type II port selection codebooks, (Z,Z') takes the value (Z2,Z′2). However, in scenarios involving Rel-18 CSI prediction, i.e., when codebookType IE is set to 'typeII-Doppler-r18' or 'typeII-Doppler-PortSelection-r18', (Z,Z') takes at least one of the following values:
[0069] If N4 = 1 is configured, the Channel State Information Reference Signal (CSI-RS) resource set is aperiodic and each resource set is configured with K CSI-RS resources, then (Z, Z') takes the value of (Z2 + 14(K-1)m, Z′2), where m is the time slot interval between adjacent CSI-RS resources.
[0070] If N4 = 1 is configured, the CSI-RS resource set is periodic or semi-persistent, and each resource set is configured with 1 CSI-RS resource, then (Z, Z') takes the value (Z2 + w, Z'2), where w = 56. (K P -1) or 56.K P A symbol, K P ∈{1,2,4} represents the number of active resources, indicated by UE capabilities.
[0071] If N4>1 is configured, the CSI-RS resource set is aperiodic and each resource set is configured with K CSI-RS resources, then (Z,Z') takes the value of (Z2+14(K-1)m,Z′2) or (Z2+14(K-1)m+Z′2,2Z′2), the value of which depends on the UE capability.
[0072] If N4>1 is configured, the CSI-RS resource set is periodic or semi-persistent and each resource set is configured with 1 CSI-RS resource, then (Z,Z') takes the value of (Z2+w,Z′2) or (Z2+w+Z′2,2Z′2), the value of which depends on the UE capability.
[0073] The values of Z2, Z′2, Z3, and Z′3 are shown in Table 1A.
[0074] Table 1A: CSI calculation delay requirements.
[0075] However, in the process of promoting the inclusion of AI-related CSI reports in communication standards, at least one of the following technical issues still needs to be addressed:
[0076] Technical Issue 1: Definition and calculation of processing unit resources in AI-related CSI reports.
[0077] In AI / ML model applications, models can run on different hardware modules, such as central processing units (CPUs), graphics processing units (GPUs), neural network processors (NNPMs), or other dedicated modules. The choice of different hardware modules directly affects the allocation of computational resources for AI-related CSI reports. This means that AI-related CSI reports may utilize APUs, CPUs, or both, depending on the model type and system configuration.
[0078] Base stations can determine or instruct the computing resources used by the UE in CSI reports through various methods. First, the base station can use a real-time interactive approach, dynamically instructing the UE to use the APU or CPU via scheduling commands or signaling messages. This method ensures flexibility in resource allocation, allowing for dynamic adjustments based on network load and UE capabilities. Second, the base station can also configure AI-related CSI report processing rules through predefined protocol settings. In this method, the UE determines whether to use the APU or CPU for CSI report calculations based on the settings in the protocol or the base station's configuration file.
[0079] With AI-related CSI reports consuming APU, O APU The computation depends on several factors. First is the UE's hardware capabilities, specifically the computing power of the APUs supported by the UE, such as the total number of APUs and their parallel processing capabilities. Second is the model type; different AI / ML models may have different computational complexities. The complexity of the AI / ML model affects the consumption of computing resources. Furthermore, the computational complexity of model inference is also a factor; the base station can determine the O(n) computational complexity based on the model type and inference complexity. APU Size.
[0080] When AI-related CSI reports consume CPU resources—that is, when AI-related CSI reports and traditional CSI reports share the CPU resource pool—resource allocation needs to consider dynamism and flexibility. The base station can adopt a dynamic resource allocation strategy, dynamically adjusting the allocation of CPU resources between AI-related CSI reports and traditional CSI reports based on current CPU utilization and task priority. Alternatively, a static allocation strategy can be used, predefining the CPU allocation ratio between AI-related CSI reports and traditional CSI reports during initialization based on the UE's reporting capabilities and model type. In this case, the CPU usage of AI-related CSI reports is O. UPU The calculation will be dynamically determined based on the CPU capabilities reported by the UE and the computing requirements reported by the CSI.
[0081] With AI-related CSI reports consuming significant APU and CPU resources, resource allocation and computation become more complex. APUThe calculations can be determined based on UE capabilities and model type, similar to the case of a pure APU. In this case, CSI report processing may involve model inference on the APU and data preprocessing and post-processing on the CPU. The base station can dynamically adjust the resource allocation of the APU and CPU according to the model complexity and UE load.
[0082] For time-series calculations in CSI reports, using only the APU, O APU The occupancy time can be determined based on model inference time, data loading time, and data output time. The base station can be based on O... APU Calculate the processing timing of CSI reports to ensure they are completed within the specified time. When AI-related CSI reports utilize both the APU and CPU, the timing calculation must consider the usage time of both. If the APU and CPU work in parallel, the timing depends on the maximum usage time. If the APU and CPU work serially, the total timing will be the sum of their respective usage times.
[0083] Through flexible resource allocation and timing calculation methods, base stations can effectively manage computing resources for AI-related CSI reports, ensuring efficient CSI report processing under various hardware architectures and AI / ML models.
[0084] Technical Issue 2: CPU resource definition and time occupancy definition for AI-related CSI reports based on multiple measurement sets.
[0085] In scenarios where AI / ML technology is applied to CSI reporting, especially for applications based on beam management (BM-Case 2) and CSI prediction, UEs typically need to collect data from multiple past measurement sets to predict future beams or CSI. This multi-measurement set data collection method provides the data foundation for AI-related CSI reports. However, the traditional method of defining CPU resource usage time—that is, fixing the CPU resource usage time to N consecutive measurement windows—can lead to resource waste in AI-related scenarios. Therefore, to improve resource utilization and reduce overhead, it is crucial to design a mechanism for sharing measurement sets across multiple tasks, enabling multiple AI-related CSI reports to be generated based on the same or the same set of measurement sets.
[0086] During data collection, the UE's processing flow typically includes at least one of the following steps: Channel estimation: Calculating channel information using the received reference signal. Model input data calculation: The calculation method for model input data varies depending on the application scenario: For BM-Case 2 (AI-based beam management): Calculating the Reference Signal Received Power (RSRP). For CSI prediction: Calculating the Precoding Matrix Indicator (PMI), Channel Quality Indicator (CQI), and Rank Indicator (RI).
[0087] In traditional CSI reporting methods, CPU resource occupancy time is typically defined as the sum of N measurement windows within a measurement cycle. This means that CPU resources are constantly occupied throughout the entire N measurement windows. However, with the introduction of AI / ML technologies, the measurement process is more focused on data collection rather than real-time computation, providing opportunities for resource optimization. By introducing AI / ML models, multiple AI-related CSI reports can share the same measurement set, thereby significantly reducing CPU resource occupancy time.
[0088] In this scenario, determining the number of CSI reports to process and the CPU time required by user devices based on their own capabilities and different AI application scenarios (such as BM-Case 2 or CSI prediction) becomes a key focus of resource optimization. To address this, a dynamic allocation mechanism can be designed to flexibly adjust CPU resource allocation according to actual measurement needs. For example, when AI-related CSI reports are generated based on a shared measurement set, the CPU time can be dynamically adjusted according to the number and type of measurement sets, rather than being fixed for all measurement windows.
[0089] In addition, AI-related CSI reports may exhibit at least one of the following characteristics over time:
[0090] CSI reports follow prediction results: In this case, consider how to determine the validity of prediction results prior to CSI reporting. Specifically, a quantitative standard is needed to determine the effective duration of prediction results, thereby filtering valid predictions before CSI reporting. This approach not only helps reduce reporting overhead but also improves reporting effectiveness by informing base stations which predictions are discarded.
[0091] CSI reports precede prediction results: In this case, the effective duration of the prediction results may affect the timing of subsequent CSI configurations and reports. To ensure the accuracy of prediction results, UEs need to dynamically filter and report CSI quantities that differ from past time periods.
[0092] By rationally designing the CPU resource usage time and multi-measurement set sharing mechanism for AI-related CSI reports, resource utilization can be significantly improved, the computation and reporting overhead of CSI reports can be reduced, and the effectiveness and reliability of CSI reports can be ensured at the same time.
[0093] Based on the above background, some embodiments of this application, building upon the research presented and / or the New Radio (NR) air interface AI / ML research based on Rel-18 and the NR air interface AI / ML work item under discussion in Rel-19, further propose an AI / ML-based CSI enhancement application scenario and a unified CSI computing resource allocation and CSI processing timing design. This design is applicable to different lifecycle management stages and aims to enhance the AI features in 5G / 6G wireless communication systems, effectively improving system intelligence and performance.
[0094] Some embodiments of this application propose a unified design for CSI computational resource allocation and CSI processing timing for AI / ML-based CSI enhancement use cases, aiming to improve the processing efficiency of CSI reports in 5G / 6G wireless communication systems. To conserve CPU processing resources, some embodiments of this application introduce a multi-task shared measurement set design. In this design, multiple AI-related CSI reports can share the same measurement set, thereby reducing CPU resource consumption. Through this sharing mechanism, not only is the CPU resource consumption time defined, but a scheme for merging multi-task CSI reports is also proposed, further reducing processing overhead.
[0095] Secondly, some embodiments of this application provide a dynamic method for calculating processing unit resource usage based on AI use cases and UE capabilities. For different AI use cases (such as beam management and CSI prediction) and different hardware configurations (such as CPU and APU), some embodiments of this application can flexibly determine the processing unit resource usage. This dynamic allocation mechanism ensures that AI-related CSI reports can operate efficiently in various scenarios.
[0096] Furthermore, some embodiments of this application propose a timing design for hybrid APU and CPU usage. In this hybrid mode, AI-related CSI reports can utilize either the APU or the CPU. Some embodiments of this application define the CSI processing timing in this hybrid mode, ensuring that AI-related and non-AI-related CSI reports can be processed in parallel, maximizing resource utilization efficiency.
[0097] Some embodiments of this application define the effective duration of AI prediction results and design a time-series processing mechanism for CSI reports and prediction results. When the reporting time sequence of CSI reports is inconsistent with the time sequence of prediction results, some embodiments of this application provide flexible processing methods to ensure the reliability and accuracy of AI model prediction results and further improve the intelligence level of the system.
[0098] Some embodiments of this application achieve at least one of the following technical effects through at least one technical solution.
[0099] Some embodiments of this application clarify the method for calculating the number of processing units occupied in a hybrid APU and CPU architecture. Whether it's an AI-related CSI report or a non-AI-related CSI report, processing unit resources can be dynamically allocated based on the UE's capabilities and model type, thereby ensuring maximum resource utilization in various application scenarios.
[0100] Some embodiments of this application further define the processing unit's uptime in a hybrid APU and CPU architecture. In this architecture, the APU and CPU can handle AI-related and non-AI-related tasks separately, or jointly process AI-related CSI reports. Through this design, some embodiments of this application can flexibly determine the PU's uptime, avoid resource waste, and improve the processing efficiency of CSI reports.
[0101] Furthermore, some embodiments of this application significantly reduce the time processing unit resources are occupied in multi-measurement set scenarios through a multi-task shared measurement set design. In traditional designs, each CSI report needs to be measured and processed independently, while the multi-task sharing mechanism of some embodiments of this application allows multiple AI-related CSI reports to be generated based on the same or the same set of measurement sets, thereby reducing the overhead of repeated measurements and processing.
[0102] Some embodiments of this application address the mismatch between the effective time of prediction results and the time of CSI reporting. For AI-related CSI predictions, some embodiments of this application define the effective duration of prediction results and provide a flexible time-series processing mechanism when the CSI reporting time and the prediction result time are inconsistent. This design ensures the reliability of AI model prediction results while guaranteeing the accuracy and timeliness of CSI reports.
[0103] All steps in the air interface interaction process involved in some embodiments of this application are not mandatory, nor do they represent all processes of interaction between the network and user equipment. Rather, they are possible air interface interaction processes proposed based on some embodiments of this application. In addition, the order of different signaling is not fixed, that is, there are no strict timing constraints between different signaling.
[0104] The technical solutions of this application can be applied to various wireless communication systems, such as: Long Term Evolution (LTE) systems, LTE Frequency Division Duplex (FDD) systems, LTE Time Division Duplex (TDD) systems, 5G communication systems, 6G communication systems, or future wireless communication systems.
[0105] For example, the wireless communication system 100 used in this application embodiment is shown in FIG2. The wireless communication system 100 may include a network-side device 110, which may be a device communicating with a user equipment 120. The network-side device 110 can provide communication coverage for a specific geographical area and can communicate with user equipment located within that coverage area. Optionally, the network-side device 110 may be a base station or a Location Management Function (LMF) for providing location services. Optionally, the base station may be an evolved Node B (eNB or eNodeB) in an LTE system, or the base station may be a mobile switching center, relay station, access point, vehicle-mounted equipment, wearable device, hub, switch, bridge, router, network equipment in a 5G network, or a base station in a future communication system, etc.
[0106] The wireless communication system 100 also includes at least one user equipment 120 located within the coverage area of the network-side device 110. As used herein, "user equipment" includes, but is not limited to, devices configured to receive / transmit communication signals via wired connections, such as via Public Switched Telephone Networks (PSTN), Digital Subscriber Line (DSL), digital cable, direct cable connection; and / or another data connection / network; and / or via a wireless interface, such as for cellular networks, Wireless Local Area Networks (WLAN), digital television networks such as DVB-H networks, satellite networks, AM-FM broadcast transmitters; and / or other user equipment. User equipment configured to communicate via a wireless interface may be referred to as a "wireless communication terminal," "wireless terminal," or "mobile terminal." Examples of mobile terminals include, but are not limited to, satellite or cellular phones; personal communications system (PCS) terminals that can combine cellular radiotelephone with data processing, fax, and data communication capabilities; PDAs that may include radiotelephones, pagers, Internet / intranet access, web browsers, notebooks, calendars, and / or Global Positioning System (GPS) receivers; and conventional laptop and / or handheld receivers or other electronic devices that include radiotelephone transceivers. User equipment can refer to access terminals, user units, user stations, mobile stations, mobile stations, remote stations, remote user equipment, mobile devices, wireless communication equipment, or user agents. Access terminals can be cellular phones, cordless phones, Session Initiation Protocol (SIP) phones, Wireless Local Loop (WLL) stations, personal digital assistants (PDAs), handheld devices with wireless communication capabilities, computing devices or other processing devices connected to a wireless modem, in-vehicle equipment, wearable devices, user equipment in 5G networks, or user equipment in future PLMN evolutions, etc.
[0107] This application provides a communication method executed by a user equipment 120. The communication method includes: reporting the capability information of the user equipment 120 to a network-side device 110 (e.g., a base station); receiving CSI report configuration and / or CSI resource configuration sent by the network-side device 110, wherein the CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signals, the CSI report configuration includes a configuration for merging CSI reports corresponding to multiple tasks into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports. Through the above technical solution, this communication method can reduce reporting overhead and resource consumption, and improve the system's computing and communication efficiency by merging CSI reports of multiple tasks into one report and / or associating a set of CSI resources with multiple CSI reports. The order, combination, specific content, and execution of the above operations do not constitute a limitation on this application embodiment and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0108] This application provides a communication method executed by a user equipment 120. The communication method includes: reporting the capability information of the user equipment 120 to a network-side device 110 (e.g., a base station); receiving channel state information (CSI) report configuration and / or CSI resource configuration sent by the network-side device 110; determining the number of CSI reports that the user equipment 120 can process, and generating and sending CSI reports according to the processing results. Through the above technical solution, this communication method can flexibly determine the number of CSI reports based on the processing capability of the user equipment, optimize resource utilization, and improve the processing efficiency of CSI reports and system performance. The order, combination, specific content, and execution of the above operations do not constitute a limitation on this application embodiment and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0109] This application provides a communication method executed by a network-side device 110 (e.g., a base station). The communication method includes: receiving capability information reported by a user equipment 120; sending Channel State Information (CSI) report configuration and / or CSI resource configuration to the user equipment 120, wherein the CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signals, the CSI report configuration includes a configuration for merging CSI reports corresponding to multiple tasks into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports. Through the above technical solution, this communication method can reduce reporting overhead and resource consumption, and improve the system's computational and communication efficiency by merging CSI reports of multiple tasks into one report and / or associating a set of CSI resources with multiple CSI reports. The order, combination, specific content, and execution of the above operations do not constitute a limitation on this application embodiment and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0110] This application provides a communication method executed by a network-side device 110 (e.g., a base station). The communication method includes: receiving capability information reported by a user equipment 120; sending Channel State Information (CSI) report configuration and / or CSI resource configuration to the user equipment 120; and receiving CSI reports reported by the user equipment 120, wherein the CSI reports are based on the number of CSI reports that the user equipment 120 can process. Through this technical solution, the communication method can flexibly determine the number of CSI reports according to the processing capability of the user equipment, optimize resource utilization, and improve the processing efficiency of CSI reports and system performance. The order, combination, specific content, and execution of the above operations do not constitute a limitation on this application embodiment and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0111] Optionally, user equipment 120 can perform device-to-device (D2D) communication with each other.
[0112] The wireless communication system 100 also includes a network 130. The network 130 may be an IP mobile communication network operated by a mobile communication operator. For example, the network 130 may be the core network used by the mobile communication operator that operates and manages the wireless communication system 100, or it may be the core network used by a virtual mobile communication operator such as a Mobile Virtual Network Operator (MVNO).
[0113] Network 130 can be connected to network-side device 110 as a relay device for transmitting user data. User equipment 120 sends and receives user data via network 130. It should be noted that user data communication is not limited to IP communication, but can also be non-IP communication.
[0114] Figure 2 exemplarily illustrates a network-side device 110, two user devices 120, and a network 130. Optionally, the wireless communication system 100 may include multiple network-side devices 110, and the coverage area of each network-side device 110 may include other numbers of user devices. This application embodiment does not limit this.
[0115] Optionally, the wireless communication system 100 may also include other network entities such as a network controller, a mobility management entity, and network elements; this application embodiment does not limit this. For example, network 130 may include other network entities such as a network controller, a mobility management entity, and network elements; this application embodiment does not limit this.
[0116] It should be understood that devices with wireless communication functions in the network / system of this application embodiment can be referred to as wireless communication devices. Taking the wireless communication system 100 shown in FIG2 as an example, the wireless communication device may include a network-side device 110, a user equipment 120, and a network 130 with communication functions. The network-side device 110 and the user equipment 120 can be the specific devices described above, which will not be repeated here. The wireless communication device may also include other devices (network 130) in the wireless communication system 100. For example, the network 130 may include other network entities such as a network controller and a mobility management entity. This is not limited in this application embodiment.
[0117] It should be understood that the terms "system" and "network" are often used interchangeably in this document. The term "and / or" in this document merely describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects are in an "or" relationship. In some embodiments, the term "configuration" can refer to "pre-configuration" and "network configuration." The terms "definition" or "pre-defined" in the embodiments of this application can be implemented by pre-storing corresponding codes, tables, or other information indicative indices in the device (e.g., including UE and network devices). This application does not limit specific implementations. For example, "definition" or "pre-defined" can refer to those defined in a protocol. It should also be understood that "protocol" in this invention can refer to standard protocols in the field of communications, such as Long Term Evolution (LTE) protocols, new radio protocols, 5G NR, future 6G, 7G, WiFi, and other related protocols used in future communication systems. This application does not limit this.
[0118] In some embodiments, Figure 3A is a flowchart illustrating a communication method performed by a user equipment according to an embodiment of this application. The communication method includes: Operation 301A: reporting the capability information of the user equipment to a base station; Operation 302A: receiving Channel State Information (CSI) report configuration and / or CSI resource configuration sent by the base station, wherein the CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signal configurations, the CSI report configuration includes a configuration for merging CSI reports corresponding to multiple tasks into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports. Through the above technical solution, this communication method can reduce reporting overhead and resource consumption, and improve the computational and communication efficiency of the system by merging CSI reports of multiple tasks into one report and / or associating a set of CSI resources with multiple CSI reports. The order, combination, specific content, and execution of Operations 301A and 302A do not constitute a limitation on the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0119] In some embodiments, Figure 3B is a flowchart illustrating a communication method performed by a user equipment according to an embodiment of this application. The communication method includes: Operation 301B: reporting the capability information of the user equipment to a base station; Operation 302B: receiving channel state information (CSI) report configuration and / or CSI resource configuration sent by the base station; Operation 303B: determining the number of CSI reports that the user equipment can process, and generating and sending CSI reports according to the processing results. Through the above technical solution, this communication method can flexibly determine the number of CSI reports based on the processing capability of the user equipment, optimize resource utilization, and improve the processing efficiency of CSI reports and system performance. The order, combination, specific content, and execution of operations 301B, 302B, and 303B do not constitute a limitation on the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0120] In some embodiments, Figure 3C is a schematic flowchart of a communication method performed by a base station according to an embodiment of this application. The communication method includes: operation 301C: receiving capability information of the user equipment reported by the user equipment; operation 302C: sending Channel State Information (CSI) report configuration and / or CSI resource configuration to the user equipment, wherein the CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signal configurations, the CSI report configuration includes a configuration for merging CSI reports corresponding to multiple tasks into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports. Through the above technical solution, this communication method can reduce reporting overhead and resource consumption, and improve the system's computational and communication efficiency by merging CSI reports of multiple tasks into one report and / or associating a set of CSI resources with multiple CSI reports. The order, combination, specific content, and execution of operations 301C and 302C do not constitute a limitation on the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0121] In some embodiments, Figure 3D is a schematic flowchart of a communication method executed by a base station according to an embodiment of this application. The communication method includes: Operation 301D: receiving capability information of the user equipment reported by the user equipment; Operation 302D: sending Channel State Information (CSI) report configuration and / or CSI resource configuration to the user equipment; Operation 303D: receiving a CSI report reported by the user equipment, wherein the CSI report is based on the number of CSI reports that the user equipment can process. Through the above technical solution, this communication method can flexibly determine the number of CSI reports according to the processing capability of the user equipment, optimize resource utilization, and improve the processing efficiency of CSI reports and system performance. The order, combination, specific content, and execution of Operations 301D, 302D, and 303D do not constitute a limitation on the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0122] The operations in Figures 3A, 3B, 3C, and 3D will be further explained below with reference to Figure 3E. It should be understood that the steps shown in Figure 3E are merely examples, and their execution order can be deleted, added, or adjusted according to actual needs.
[0123] In some embodiments, FIG3E illustrates the overall flow of a communication method provided by this application, and proposes at least one solution, such as Solution 1, Solution 2, and Solution 3.
[0124] The overall solution of this application includes at least one of the following steps:
[0125] In the first step, the user equipment reports its capability information to the base station.
[0126] After receiving the user equipment capability information, the base station bases it according to the N contained therein. APU N CPU O APU and / or O CPU Information determines the configuration of CSI reports, including the number and type of CSI reports. APU For example, it refers to the number N of available CSI processing units (APUs) in the user equipment. CPU N CPU For example, this refers to the number of available CSI processing units (CPUs) in the user equipment. CPU For example, it refers to the number of CPUs required by the user equipment for CSI reports corresponding to multiple use cases, feature sets, models, or CSI report types. APUFor example, this refers to the number of APUs required by the user equipment for CSI reports corresponding to multiple use cases, function sets, models, or CSI report types. Specifically, the base station can select at least one of the following configuration modes: The first is a mode that depends on both APU and CPU, where CSI report tasks can be dynamically allocated between APU and CPU to ensure balanced computing tasks; the second is a mode that depends only on APU, which is typically suitable for AI-related CSI report tasks and is processed independently by the APU; the third is a configuration based on enhanced CPU rules, in which the base station can flexibly adjust resource allocation according to enhanced CPU rules, such as prioritizing the processing of CSI reports for priority tasks in a multi-tasking scenario. After determining the priority configuration mode, the base station will send it to the user equipment via RRC messages or other control signaling to ensure that CSI reports are processed and reported according to the predetermined mode.
[0127] In the second step, the user equipment receives CSI report configuration and / or CSI resource configuration sent by the base station. In some embodiments, the CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signals, the CSI report configuration includes a configuration for merging CSI reports corresponding to multiple tasks into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports.
[0128] After performing measurements, the user equipment (UE) determines how to process the CSI report based on the measurement results. For example, after performing the measurement, the UE processes the measurement results according to the CSI report configuration to generate the CSI reporting volume required for a traditional CSI report, or the model input volume required for an AI-related CSI report. The UE can process the CSI report through an AI / ML processing unit or a traditional processing unit (non-AI / ML related processing unit). Alternatively, the UE can ignore the CSI report and not report it. Specifically, if the CSI report is deemed important, the UE will report it normally according to the configuration; however, if the CSI report is assessed as unnecessary or redundant, it will be ignored and no longer reported to the base station. This mechanism effectively reduces unnecessary CSI report transmission, lowers network load, and ensures the accurate transmission of critical CSI reports.
[0129] In the third step, the user equipment receives CSI report triggering information sent by the base station via Downlink Control Information (DCI) and / or Medium Access Control Control Element (MAC-CE) messages. In some embodiments, the CSI report triggering information includes a message for triggering CSI report merging, and / or configuration selection information for CSI report merging. Based on this triggering information, the user equipment determines whether CSI reports for multiple tasks should be merged or reported separately. Specifically, when CSI reports for multiple tasks have similar reporting periods or configurations, the user equipment may choose to merge multiple CSI reports into one CSI report to reduce the number of reports and lower signaling overhead; while when the reporting requirements of the tasks are different or involve different types of CSI, the user equipment may choose to report separately to ensure the accuracy of each task's report. This mechanism improves the flexibility and transmission efficiency of CSI reporting.
[0130] In the fourth step, the user equipment (UE) determines the number of CSI reports it can process and generates and sends a CSI report according to the processing result. In some embodiments, the sending time of the CSI report is configured by the base station or determined by the UE based on the CSI report processing time corresponding to the AI function. In some embodiments, if the CSI report is a merged report, the CSI report includes a header indicating relevant information of the merged multiple CSI reports, including CSI report IDs and / or sorting order; and / or the UE reports the CSI reporting volume corresponding to multiple different AI tasks according to the configuration of the base station or the order indicated in the header. In some examples of this application, the sending time of the CSI report is configured by the base station or determined by the UE based on the CSI report processing time corresponding to the AI function.
[0131] Specifically, the overall solution of this application is mainly subdivided according to at least one of the following situations:
[0132] In the first scenario, AI-related CSI reports and non-AI-related CSI reports occupy separate processing units, with AI-related and non-AI-related CSI reports processed using an AI processing unit and a CPU, respectively. For ease of explanation, this application refers to the AI-related CSI report processing unit as an APU; however, it should be understood that other names may be used in specific implementations. Specifically, both the AI and non-AI portions of the AI-related CSI reports are computed using the APU, ensuring that AI-related and non-AI-related tasks are processed separately on independent computing resources, thereby avoiding mutual interference and improving processing efficiency.
[0133] In the second scenario, AI-related CSI reports and non-AI-related CSI reports share CPU resources, but are divided into different parts. AI-related CSI reports occupy part A of the CPU, while non-AI-related CSI reports occupy part B. This resource allocation can be predefined by a protocol or preconfigured by network-side equipment (e.g., a base station). The AI portion of the AI-related CSI reports further occupies part A1 of part A, and the non-AI portion occupies part A2. The specific configuration can be further refined: part A of the AI-related CSI reports and part B of the non-AI-related CSI reports can share resources (corresponding to the second embodiment); or part A2 of the AI-related CSI reports and part B of the non-AI-related CSI reports can share resources (corresponding to the second embodiment). In some cases, part A of the AI-related CSI reports and part B of the non-AI-related CSI reports are strictly separated and cannot be shared (corresponding to the third embodiment).
[0134] In the third scenario, AI-related CSI reports utilize both APU and CPU resources, while non-AI-related CSI reports utilize only CPU resources. Specifically, the AI portion of the AI-related CSI report is computed using the APU, while the non-AI portion is processed using the CPU. This hybrid computing model ensures that AI tasks can fully utilize the efficient APU, while non-AI tasks can flexibly utilize the CPU. Furthermore, the non-AI portion of the AI-related CSI report and the non-AI-related CSI report share CPU resources, and CPU usage is considered together when they are computed (corresponding to the third embodiment). This model can be further subdivided into two configurations: the non-AI portion of the AI-related CSI report occupies part A of the CPU, while the non-AI-related CSI report occupies part B of the CPU. The two parts can share CPU resources (corresponding to the second embodiment) or be strictly separated and not interchangeable (i.e., CPU sharing is not possible).
[0135] Solution 1 corresponds to the first scenario, where AI-related CSI reports are calculated and processed entirely based on the APU. This solution includes at least one of the following steps:
[0136] In the first step, the UE reports its capability information to the base station via Radio Resource Control (RRC) messages, such as N... APU (Number of available APUs) and O APU (Indication information such as the number of APUs required by the UE for CSI reports corresponding to multiple use cases, feature sets, models, or CSI report types.) APUIt will also be mapped to various AI use cases, feature sets, or models to ensure that the base station can reasonably allocate resources according to the UE's AI processing capabilities.
[0137] In the second step, the base station configures the UE's CSI report and measurement resources via RRC messages (corresponding to the eighth embodiment). The CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signal resources, including Channel State Information Reference Signal (CSI-RS) and Synchronization Signal Block (SSB). Furthermore, the CSI report configuration can also include a configuration to merge CSI reports from multiple tasks into a single CSI report, reducing reporting complexity. Correspondingly, the CSI resource configuration allows a set of CSI resources to be associated with multiple CSI reports, further improving resource utilization.
[0138] In the third step, the base station sends CSI report triggering information to the UE via DCI and / or MAC-CE messages. This triggering information may include a message that triggers CSI report merging and specific configuration options for merging CSI reports. The base station can flexibly specify merging strategies to ensure the flexibility and efficiency of report processing.
[0139] In the fourth step, the UE determines the number of CSI reports it can process based on its capabilities and the base station configuration, and generates and sends a CSI report based on the processing results (corresponding to the fourth embodiment). The sending time of the CSI report can be configured by the base station or dynamically determined by the UE based on the CSI report processing time corresponding to the AI function. If it is a merged CSI report, a header will be included in the report indicating the relevant information of each merged CSI report (such as CSI report ID and / or sorting order). The UE reports the reporting volume of different tasks sequentially according to the base station configuration or the order indicated in the header.
[0140] In this scheme, the method for determining CPU usage time includes at least one of the following:
[0141] Based on the configuration of CSI reporting volume, it is determined by the selected measurement reference signal resource location and the CSI calculation time corresponding to the current AI task;
[0142] The CSI reporting volume is determined based on the selected measurement reference signal resource location, the CSI reporting time, and the CSI calculation time corresponding to the current AI task.
[0143] The CSI reporting volume is determined based on the time of receiving the CSI report, the time of the CSI report, and the CSI calculation time corresponding to the current AI task.
[0144] The beneficial effects of this solution are reflected in the CSI report merging and the design of CSI acquisition based on a shared measurement set. For BM-Case 2 and CSI prediction application scenarios, data collection from multiple past measurement sets is typically required to predict future beams or CSI. Traditionally, each measurement requires dedicated CPU resources for N measurement windows, leading to resource waste. By introducing AI / ML technology, UE measurements become more data-collection-oriented. This solution designs a shared measurement set mechanism, enabling multiple AI-related CSI reports to be generated based on the same or the same set of measurement sets, thereby reducing measurement overhead. Furthermore, multiple task CSI reports based on the shared measurement set can be further merged to reduce overall reporting overhead. This design significantly improves resource utilization efficiency.
[0145] Solution 2 corresponds to the second scenario, where AI-related CSI reports are entirely computed and processed using CPU. This solution includes at least one of the following steps:
[0146] In the first step, the UE reports its capability information to the base station via an RRC message. Specifically, the UE's capability report includes indications related to CPU resources, including N. CPU (Number of available CPUs) and O CPU (The number of CPUs required by the UE for CSI reports corresponding to multiple use cases, feature sets, models, or CSI report types). In addition, AI-related O cPU This will be further mapped to various AI use cases, feature sets, or models, for example Corresponding to function set 1, This corresponds to function set 2, etc. The UE capability report can also indicate how CPU resources are allocated between AI-related CSI reports and non-AI-related CSI reports. The allocation method can be proportional, such as specifying a percentage for each type of report; or it can be an absolute allocation, specifying the number of CPUs used by each type of report.
[0147] In the second step, the base station configures the UE's CSI reports and measurement resources via RRC messages (corresponding to the eighth embodiment). Specifically, the CSI report configuration indicates that multiple CSI reports can be associated with the same or partially the same set of measurement reference signal resources, including CSI-RS and SSB. This configuration can also specify merging CSI reports required by multiple tasks into a single CSI report, reducing report complexity. Furthermore, in the CSI resource configuration, a set of CSI resources can be associated with multiple CSI reports to further improve resource utilization efficiency.
[0148] In the third step, the base station sends CSI report triggering information to the UE via DCI and / or MAC-CE messages. This triggering information may include a message that triggers CSI report merging and specific configuration options for merging CSI reports. The base station can flexibly specify merging strategies to ensure the flexibility and efficiency of report processing.
[0149] In the fourth step, the UE determines the number of CSI reports it can process based on its capabilities and the base station configuration, and generates and sends CSI reports based on the processing results. The UE can dynamically determine CPU resource allocation after receiving a CSI report request triggered by the base station. Specifically, CPU resources can be dynamically allocated between AI-related and non-AI-related CSI reports to maximize CPU resource utilization efficiency (corresponding to the second embodiment). Furthermore, to improve CPU utilization efficiency, the UE can further optimize based on the number of CPUs occupied by AI-related CSI reports and the CPU usage time (corresponding to the sixth embodiment). In particular, the CSI report processing time can be determined by looking up a table, ensuring efficient and reasonable CPU resource allocation for various tasks.
[0150] This solution offers at least one of the following benefits: By processing both AI-related and non-AI-related CSI reports on the CPU, it achieves flexibility in resource utilization. CPU resources can be dynamically shared between AI-related and non-AI-related CSI reports, avoiding resource waste. Furthermore, by supporting report merging in CSI reports, it further reduces reporting overhead. In practical applications, this design allows the UE to dynamically adjust the number of CSI reports that can be processed based on current CPU resource usage, ensuring efficient utilization of system resources.
[0151] Solution 3 corresponds to the third scenario, where AI-related CSI reports are processed and computed jointly by the APU and CPU. This solution aims to fully utilize the heterogeneous computing resources of the UE to achieve efficient generation and reporting of AI-related CSI reports.
[0152] In the first step, the UE reports its capability information to the base station via an RRC message. Specifically, the UE's capability report includes the following indication information: N CPU (Number of available CPUs), O CPU N APU (Number of available APUs) and O APU Among them, AI-related O CPU This is further mapped to various AI use cases, feature sets, or models. For example, Corresponding to function set 1, This corresponds to function set 2, etc. Here, it is also necessary to indicate how CPU resources are allocated between the non-AI portion of AI-related CSI reports and traditional CSI reports. This allocation can be achieved by indicating the respective proportions or absolute quantities. Optionally, for O... APU This can be determined through protocol pre-definition, with the network side automatically learning the O based on the UE type, capabilities, and function set. APU The mapping relationship.
[0153] In the second step, the base station configures the UE's CSI reports and measurement resources via RRC messages (corresponding to the eighth embodiment). Specifically, the CSI report configuration indicates that multiple CSI reports can be associated with the same or partially the same set of measurement reference signal resources, including CSI-RS and SSB. Furthermore, the CSI report configuration can also specify merging CSI reports required for multiple tasks into a single CSI report, reducing reporting complexity. In the CSI resource configuration, a set of CSI resources can be associated with multiple CSI reports, thereby improving resource utilization efficiency.
[0154] In the third step, the base station sends CSI report triggering information to the UE via DCI and / or MAC-CE messages. This triggering information contains a message that triggers CSI report merging, as well as specific configuration options for merging CSI reports. The base station can flexibly specify merging strategies to ensure the flexibility and efficiency of report processing.
[0155] In the fourth step, the UE determines the number of CSI reports it can process based on its capabilities and the base station configuration, and generates and sends CSI reports according to the processing results (corresponding to the first embodiment). The UE's processing method maximizes CPU resource utilization efficiency by combining serial and parallel processing. Specifically, the UE divides the observation window or sub-observation window of CSI measurement into multiple processing windows, and executes one group of CSI reports within each processing window. The sum of the CSI reports processed in multiple processing windows across multiple task groups is the maximum number of CSI reports that the UE can process.
[0156] In addition, the UE employs at least one of the following strategies (corresponding to the seventh embodiment) to manage the APU and CPU usage time during CSI report processing:
[0157] CPU usage time is defined as a piecewise function that varies depending on the CSI report type and task.
[0158] APU time is defined as a continuous function, making it more suitable for AI tasks that require continuous processing.
[0159] APU usage time includes data measurement (data collection) and inference, while CPU usage time includes the generation of the Physical Uplink Control Channel (PUCCH) / Physical Uplink Shared Channel (PUSCH).
[0160] When an AI task uses the APU for inference, the CPU is only responsible for data measurement and uplink channel generation; while when an AI task only uses the CPU, the CPU also needs to handle AI-related tasks.
[0161] The beneficial effects of this solution are reflected in at least one of the following aspects:
[0162] It provides a flexible APU and CPU resource sharing mechanism, allowing AI-related CSI reports to be dynamically allocated between APU and CPU.
[0163] By merging CSI reports and designing based on shared measurement sets, reporting overhead can be further reduced.
[0164] It supports parallel processing of multiple tasks, dynamically allocating CPU and APU resources within the measurement observation window / sub-observation window to maximize resource utilization.
[0165] By defining the CPU and APU usage time using piecewise and continuous functions respectively, the computational efficiency of different tasks is guaranteed.
[0166] It provides a variety of CPU and APU resource allocation strategies, enabling the UE to flexibly select computing resources based on the UE's capabilities and the current task.
[0167] To facilitate understanding of the technical solutions of the embodiments of this application, the technical solutions related to the embodiments of this application are described below. The technical solutions of the first embodiment, the second embodiment, the third embodiment, the fourth embodiment, the fifth embodiment, the sixth embodiment, the seventh embodiment, the eighth embodiment, the ninth embodiment, the tenth embodiment, and the eleventh embodiment are listed below for description, but this application is not limited thereto.
[0168] In some embodiments of this application, the solution of the first embodiment can be implemented in conjunction with the solutions of the second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth, and / or eleventh embodiments, or it can be implemented independently of the solutions of the second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth, and eleventh embodiments. In some embodiments of this application, the solutions of multiple embodiments can be implemented in combination or independently.
[0169] First embodiment: Calculation of CSI reported resources when multiple tasks share measurement resources.
[0170] In some examples of this application, the user equipment determines the number of CSI reports it can process through a combination of serial and parallel processing, including at least one of the following: dividing the observation window or sub-observation window of CSI measurement into multiple processing windows, and executing the CSI report processing of a task group in parallel or serially within each processing window; the sum of the CSI reports processed in the multiple processing windows constitutes the maximum number of CSI reports that the user equipment can process within the corresponding observation window or sub-observation window. In some examples of this application, the order, combination, specific content, and whether or not the steps are executed do not constitute a limitation on the embodiments of this application, and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0171] To further improve resource utilization efficiency and ensure efficient UE operation in scenarios with multiple tasks sharing measurement resources, this embodiment proposes a CSI report calculation method combining serial and parallel processing. In this method, the UE can divide the observation window or sub-observation window into multiple processing windows, each of which can independently execute CSI report processing for a task group. This design not only maximizes CPU resource utilization in parallel processing but also ensures efficient completion of each task under limited resource conditions through reasonable allocation and scheduling of task groups. By flexibly configuring the number of processing windows, the allocation method of task groups, and the execution order of tasks, the UE can dynamically adjust its CSI report processing method according to the current processing capacity and task requirements, ensuring both full utilization of resources and flexibility in multi-task processing. Whether in high-load or low-load scenarios, the UE can maximize the number of CSI reports by adjusting the allocation of processing windows and the configuration of task groups.
[0172] This embodiment provides a method for calculating CSI reporting resources in a scenario where multiple tasks share measurement resources. The core of this technical solution lies in improving resource utilization efficiency by rationally allocating CPU computing resources to support more CSI reports within the observation window. Specifically, this embodiment is applicable to AI / ML air interface use cases, such as BM Case 2 and CSI prediction, where the UE typically needs to perform multiple measurements over past time periods to obtain sufficient measurement samples for AI / ML model inference. In this case, the system usually maintains a relatively long observation window, where measurement reference signals are transmitted at fixed intervals.
[0173] Within this observation window, the time interval between adjacent reference signals is defined as a sub-observation window. Within each sub-observation window, the UE needs to process the received reference signals using its processing unit to generate data that meets the input requirements of the AI / ML model. Specifically, the generation of model input data mainly involves at least one step:
[0174] Channel estimation: The UE performs channel estimation using the received reference signal (such as CSI-RS or SSB) to generate a channel matrix H, which serves as the basis for subsequent data calculations.
[0175] Model input data calculation: Based on the channel matrix H, the UE further calculates the input data for AI / ML model inference, the specific task depending on the AI / ML use case used:
[0176] BM Case 2: Calculate the received signal power (e.g., RSRP).
[0177] CSI prediction: Calculate channel quality metrics (such as PMI, CQI, RI).
[0178] UCI generation and PUCCH mapping: After obtaining the above model input data, the UE further generates uplink control information (UCI) and maps it to the physical uplink control channel for reporting.
[0179] In this scenario of multiple tasks sharing measurement resources, determining the CPU usage time (O CPU The key lies in assessing the CPU computing resources consumed by each task in processing model input data. Since channel estimation and model input data computation are both completed during the measurement phase, CPU resources are mainly concentrated on the computation of model input data. By rationally arranging the processing order and CPU resource allocation of each task within the sub-observation window, more CSI reports can be achieved under limited resource conditions, effectively improving resource utilization.
[0180] Option 1: Resource consumption of processing units combining serial and parallel processing.
[0181] This paper proposes a resource allocation method for processing units that combines serial and parallel processing to efficiently handle multi-task CSI reports, especially in AI / ML-related CSI reporting scenarios. This method maximizes the utilization of processing unit resources by flexibly scheduling processing tasks within the same time period, significantly improving the UE's reporting capabilities.
[0182] Specifically, this scheme combines serial and parallel processing of the computational processes of multiple tasks within a sub-observation window. Taking Figure 4A as an example, the sub-observation window is divided into L processing windows, each of which can be considered an independent computation time period. The number M of CSI reports that can be processed within each processing window satisfies... Within the entire sub-observation window, the number of CSI reports that can be processed satisfies the following expression:
[0183] At this point, the total number of CSI reports supported within each sub-observation window satisfies:
[0184] Without dividing the processing window, the number of CSI reports that can be processed is only M1, which clearly means M >> M1. k Therefore, it is evident that by employing a combined serial and parallel processing approach, the UE's CSI report processing capability is significantly improved. Although the specific task scheduling and processing quantity (M) decisions are implemented on the UE side, if the base station can anticipate the UE's processing capacity, it can configure the number of CSI reports more precisely. This combination of anticipation and configuration can effectively avoid resource waste caused by an excessive number of CSI reports or information loss caused by an insufficient number of reports, thereby maximizing the utilization efficiency of UE processing resources.
[0185] Within these processing windows, the UE can flexibly allocate computing resources according to task requirements: some tasks can be computed in parallel within a processing window, while others can be scheduled serially between processing windows. For example, for CSI reporting that requires AI / ML, the UE can perform channel estimation in one processing window and model inference in another, with seamless integration between these tasks achieved through data caching and asynchronous computation. This design ensures that the UE can efficiently process multiple CSI reporting tasks with the same processing unit resources. Furthermore, this method supports dynamically adjusting the number of processing windows and task allocation based on task priority and computational requirements. Whether in high-load or low-load application scenarios, the UE can maximize resource utilization and ensure efficient processing of multiple tasks by adjusting the number and configuration of processing windows. This design reduces resource waste while ensuring computational accuracy and performance, representing an innovative approach to efficient multi-task processing.
[0186] In scenarios where multiple users share measurement resources, employing a combined serial and parallel CSI processing method can significantly increase the number of CSI tasks processed within the same timeframe. To ensure the effectiveness and flexibility of this efficient processing approach, it is necessary to reach a consensus between the network (NW) and the user equipment (UE), enabling the network to configure CSI reports on demand based on the user equipment's processing capabilities, thus avoiding resource waste and reporting conflicts. This consensus can be achieved through at least one of the following methods:
[0187] Method 1: UE initiates actively.
[0188] In this method, the UE proactively initiates CSI report configuration for serial processing. The core of this method lies in the UE proactively reporting its supported serial processing capabilities to the NW, enabling the network to flexibly configure CSI reports based on the UE's processing capabilities. This spontaneous reporting mechanism ensures that network configuration fully considers the UE's processing capabilities, avoiding resource waste or reporting conflicts. Method 1 includes at least one of the following steps:
[0189] Step 1: The UE reports its capability information.
[0190] The UE reports its ability to support serial processing to the NW via a capability report. This capability report may include at least one of the following information: the granularity of the sub-observation window division, i.e., the length of the processing window, ensuring the NW can flexibly configure the processing windows according to actual needs; the value of L, a positive integer indicating the number of processing windows within a sub-observation window; and optional non-uniform distribution information, indicating the length of each sub-observation window. This information allows the NW to clearly understand the UE's processing capabilities and limitations.
[0191] Step 2: NW configures CSI reports.
[0192] Based on the capability information reported by the UE, the NW can flexibly configure the UE's CSI reports. Specifically, the NW can configure the CSI report processing strategy according to the number and length of the UE report processing windows, ensuring that the number of CSI reports and the processing window division match the UE's processing capabilities. This flexible configuration can avoid resource waste and ensure that the UE can efficiently process CSI reports under existing resource conditions.
[0193] Method 2-1: Base station configuration (static configuration).
[0194] In this method, the NW uses static configuration to determine the UE's CSI report processing mode. In this method, the NW directly configures the UE's CSI report processing mode based on the UE's capability information or the base station's predefined policies, thereby achieving flexible and efficient resource utilization. Method 2-1 includes at least one of the following steps:
[0195] In optional step 0, the UE can report its ability to support serial processing to the NW via a UE Capability Report or a Scheduling Request (SR). This reporting can be implemented using a one-bit indicator to indicate whether the UE has the capability to serially process CSI reports. This step is optional because in some cases, the NW can configure this directly based on predefined policies without requiring the UE to actively report its capabilities.
[0196] In step 1, the NW sends a configuration to the UE via an RRC message to specify the UE's CSI report processing mode. This configuration includes at least one of the following:
[0197] Granularity of sub-observation windows: i.e., the length of the processing window. This parameter determines the minimum time unit within which CSI report processing can be performed in each sub-observation window.
[0198] Value of L: L is a positive integer used to specify the number of processing windows within each sub-observation window. NW can choose the value of L based on the UE's capabilities or a predefined strategy, thereby flexibly controlling the processing granularity of CSI reports.
[0199] In step 2 (optional), the NW can dynamically activate the UE's serial processing mode via DCI. Specifically, the NW can include a bit field in the DCI message to indicate whether the serial processing mode is activated. This DCI message typically also includes a CSI request to trigger the UE to perform a CSI report.
[0200] Through this dynamic activation mechanism, the network controller (NW) can flexibly control the UE's CSI report processing mode. For example, when the NW detects high network load or insufficient UE processing capacity, it can choose to activate the serial processing mode to improve resource utilization efficiency. Conversely, when the network load is low or the UE's processing capacity is strong, the NW can also choose not to activate serial processing, thereby achieving higher report throughput.
[0201] The advantage of this DCI activation method is that it can achieve dynamic adjustment based on network status and UE capabilities, without requiring the UE to re-report its capabilities or the NW to reconfigure the CSI reporting mode each time, thereby reducing signaling overhead while ensuring flexibility.
[0202] Method 2-2: Base station configuration (dynamic configuration).
[0203] In this approach, the NW provides the UE with flexible serial processing options through dynamic configuration. Unlike static configuration, this method allows the NW to dynamically select the most suitable CSI reporting configuration based on network requirements or the UE's actual capabilities, thereby achieving a balance between resource utilization and signaling efficiency.
[0204] In step 0 (optional), the UE can report its ability to support serial processing to the NW via a capability report or scheduling request. This capability reporting can be implemented using a 1-bit indicator to show whether the UE supports serial processing. The NW can then use this information to further determine its subsequent configuration strategy.
[0205] In step 1, the NW sends multiple sets of serial processing options to the UE via RRC messages. These options can include multiple different CSI reporting configurations, each defining a different processing window partitioning method. Specifically, each configuration set can include at least one of the following:
[0206] Multiple optional sub-observation window granularity options, i.e., the length of the processing window.
[0207] Multiple values of L (positive integers) represent the number of processing windows within each sub-observation window.
[0208] Through this multi-configuration approach, NW can preset multiple processing modes for UE, thereby flexibly responding to different network loads and UE computing capabilities.
[0209] In step 2, the NW activates a specific serial processing configuration via dynamic signaling. The NW can choose at least one of the following activation methods:
[0210] DCI Activation: When the number of configuration groups is small, NW can directly activate a specific configuration via DCI. This method is suitable for situations with few configuration options, and the signaling is simple and the response is fast.
[0211] MAC-CE message combined with DCI activation: When there are many configuration groups, to reduce DCI overhead, the NW can use a combination of MAC-CE messages and DCI. The NW selects a subgroup from multiple configuration groups via MAC-CE messages, and then activates the specific configuration in that subgroup via DCI. This method is more efficient when there are many configuration groups and can reduce the consumption of downlink control information.
[0212] This dynamic configuration method has at least one advantage: firstly, it ensures that the NW can flexibly select the appropriate CSI report configuration based on the UE's capabilities; secondly, by combining MAC-CE messages with DCI, it significantly reduces signaling overhead when there are many configuration options, thereby improving overall signaling efficiency. Through this design, the NW and UE can maximize the utilization efficiency of system resources while meeting the UE's processing capabilities.
[0213] Option 2: Parallel processing unit resource consumption.
[0214] This solution employs a parallel processing approach, significantly improving the utilization efficiency of processing unit resources by processing CSI reports in parallel within the UE, thereby supporting more CSI reporting tasks within the same time period. This design approach is particularly suitable for AI / ML-related CSI report processing scenarios that require completing a large number of computational tasks in a short period of time.
[0215] During the data collection phase, the reference signal data received by the UE will be used in parallel to calculate different AI / ML-related CSI reports. This means that within each sub-observation window, the UE can process multiple CSI report tasks simultaneously without sequential processing between different tasks. The efficiency of this parallel processing depends on the allocation of processing unit resources. Specifically, taking Figure 4B as an example, the number M of CSI reports that can be processed within each sub-observation window satisfies the following condition: N represents the amount of CPU resources required for the m-th CSI reporting task. CPU This indicates the total amount of available CPU resources in the UE. L represents the amount of CPU resources currently being used.
[0216] Through this design, M>>M k The number of CSI reports M that the UE can process simultaneously within each sub-observation window far exceeds M1 under the traditional method (i.e., the number processed only serially). This parallel processing method significantly improves the UE's resource utilization, making it possible to complete more tasks in the same amount of time.
[0217] Furthermore, although the specific value of M depends on the UE's implementation and resource allocation strategy, if the NW can anticipate the UE's processing capabilities (e.g., through UE capability reports), the base station can flexibly configure the number of CSI reports based on this information, ensuring that neither UE resource waste nor resource shortage occurs. This design maximizes UE resource utilization while ensuring the timeliness and reliability of CSI reports.
[0218] Second embodiment: CPU resources are dynamically allocated between AI-related CSI reports and traditional (Legacy) CSI reports (e.g., non-AI-related CSI reports).
[0219] In some examples of this application, the CSI report includes AI-related CSI reports and / or non-AI-related CSI reports. The AI-related CSI reports occupy a first portion of the CPU, and the non-AI-related CSI reports occupy a second portion of the CPU. The AI portion of the AI-related CSI reports occupies a first sub-portion of the first portion of the CPU, and the non-AI portion of the AI-related CSI reports occupies a second sub-portion of the first portion of the CPU. In some examples of this application, the first portion of the CPU occupied by the AI-related CSI reports and the second portion of the CPU occupied by the non-AI-related CSI reports are shared. In some examples of this application, the non-AI portion of the AI-related CSI reports occupies a second sub-portion of the first portion of the CPU, and the second portion of the CPU occupied by the non-AI-related CSI reports is shared. In some examples of this application, the user equipment determines the number of CSI reports it can process by allocating CPU resources between the AI-related CSI reports and the non-AI-related CSI reports. In some examples of this application, the user equipment allocates CPU resources between the AI-related CSI reports and the non-AI-related CSI reports, including at least one of the following: the AI-related CSI reports and the non-AI-related CSI reports jointly occupy the CPU resources, and the CPU resources are dynamically allocated between the AI-related CSI reports and the non-AI-related CSI reports; the AI-related CSI reports and the non-AI-related CSI reports jointly occupy the CPU resources, and the CPU resources are dynamically allocated between the non-AI-related user equipment behavior corresponding to the AI-related CSI reports and the non-AI-related CSI reports; the AI-related user equipment behavior corresponding to the AI-related CSI reports occupies the AI processing unit resources, the non-AI-related user equipment behavior corresponding to the AI-related CSI reports and the non-AI-related CSI reports jointly occupy the CPU resources, and the CPU resources are dynamically allocated between the non-AI-related user equipment behavior corresponding to the AI-related CSI reports and the non-AI-related CSI reports. In some examples of this application, the reporting priority of the CSI report depends on at least one of the following parameters: reporting configuration ID, serving cell index, reporting volume, reporting configuration type, user equipment behavior corresponding to the report, and CPU resource usage of the corresponding report.
[0220] In some examples of this application, the order, combination, specific content, and whether or not the steps are executed do not constitute a limitation on the embodiments of this application. They can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0221] In some embodiments of this application, the UE can flexibly allocate resources between AI-related CSI reports and non-AI-related CSI reports based on the allocated CPU resources. Specifically, the UE can dynamically allocate resources between AI-related CSI reports and non-AI-related CSI reports based on the CSI reporting policy configured in the NW and its own CPU resource usage. This allocation method not only adapts to the mixed processing needs of AI-related and non-AI-related tasks, but also dynamically adjusts the resource allocation strategy according to task priority and CPU resource usage. In practical applications, this CPU resource allocation mechanism can be implemented in various ways, such as resource scheduling based on task priority, resource allocation based on task category, or dynamic adjustment based on real-time CPU resource utilization. Regardless of the method used, the system can ensure that high-priority tasks are processed first under resource constraints, maximizing resource utilization and ensuring timely task processing. Through the above flexible CPU resource allocation mechanism, the UE can efficiently process CSI reports in various complex scenarios, improving system resource utilization and ensuring the real-time performance and accuracy of CSI reports. The specific resource allocation mechanism and its working principle in at least one scenario will be described in detail below.
[0222] In some examples of this application, the mechanism for dynamic allocation of CPU resources is applicable to at least one of the following scenarios:
[0223] Scenario 1: AI-related CSI reports and non-AI-related CSI reports share CPU resources, which can be dynamically allocated between them. Assume there are N reports on the same OFDM symbol. AI One AI-related CSI report and N that need to be submitted. non-AI For each non-AI-related CSI report that needs to be reported, AI-related CSI reports are allocated to the remaining available CPUs according to their reporting priority, from highest to lowest, until the maximum number of AI-related CSI reports (M) that can be allocated to AI-related CSI reports is determined. AI,1 ,satisfy in N represents the CPU usage required for the nth AI-related CSI report. CPU,AI To allocate the total CPU to AI-related CSI reports, L CPU,AI The used CPUs are allocated from the total CPUs allocated to AI-related CSI reports. Non-AI-related CSI reports are then allocated to the remaining available CPUs allocated to them, ranked from highest to lowest priority according to their second CSI reporting priority, until the maximum number of non-AI-related CSI reports (M) allocated to them is determined.non-AI,1 ,satisfy in N represents the CPU usage required for the nth non-AI related CSI report. CPU,non-AI To allocate the total CPU to non-AI related CSI reports, L CPU,non-AI Used CPUs are allocated to the total CPU count in non-AI-related CSI reports. In some examples of this application, at least one of the following situations exists:
[0224] Case 1: N AI The CPU required for each AI-related CSI report exceeds the remaining available CPU allocated to AI-related CSI reports, i.e., M AI,1 <N AI Then the remaining N AI -M AI,1 AI-related CSI reports can be allocated to the remaining available CPU for non-AI-related CSI reports according to their third CSI reporting priority, from high to low, until the maximum number of AI-related CSI reports (M) that can be allocated to non-AI-related CSI reports is determined. AI,2 ,satisfy UE does not require reporting the remaining N AI -M AI,1 -M AI,2 A CSI report related to AI.
[0225] Case 2: N non-AI The CPU required for a single non-AI-related CSI report exceeds the remaining available CPU allocated to non-AI-related CSI reports, i.e., M non - AI,1 <N non - AI Then the remaining N non - AI -M non - AI,1 Each non-AI-related CSI report can be allocated to the remaining available CPU for AI-related CSI reports in descending order of fourth CSI reporting priority, until the maximum number of non-AI-related CSI reports (M) allocated to AI-related CSI reports is determined. non-AI,2 ,satisfy UE does not require reporting the remaining N non-AI -M non-AI,1 -M non-AI,2 A non-AI related CSI report.
[0226] In Scenario 1, by dynamically allocating CPU resources, the system ensures flexible allocation and priority management of AI-related and non-AI-related CSI reports within limited CPU resources, maximizing CPU resource utilization efficiency. Specifically, when the total number of AI-related or non-AI-related CSI reports exceeds their respective allocated CPU resources, the system will transfer the excess CSI reports to the other party's resource pool according to predefined CSI reporting priorities, utilizing the remaining available CPU. This design ensures that, with limited CPU resources, the system can prioritize reporting high-priority CSI reports while avoiding CPU resource waste, thus improving overall system performance and resource utilization.
[0227] Scenario 2: AI-related CSI reports and non-AI-related CSI reports share CPU resources. These CPU resources can be dynamically allocated among non-AI-related UE behaviors corresponding to AI-related CSI reports and non-AI-related CSI reports. Assume there are N [unclear - possibly referring to a specific number of CSI reports] on the same OFDM symbol. AI One AI-related CSI report and N that need to be submitted. non-AI For each non-AI-related CSI report that needs to be reported, AI-related CSI reports are allocated to the remaining available CPU of the non-AI-related UE behavior corresponding to the AI-related CSI report, in descending order of their first CSI reporting priority, until the maximum number M of AI-related CSI reports (M) corresponding to the non-AI-related UE behavior corresponding to the AI-related CSI report is determined. AI,1 ,satisfy in N represents the CPU required for the non-AI-related UE behavior corresponding to the nth AI-related CSI report. CPU,AI,2 To allocate the total CPU to non-AI-related UE behaviors corresponding to AI-related CSI reports, L CPU,AI,2 This is the total CPU allocated to non-AI-related UE behaviors corresponding to AI-related CSI reports. Non-AI-related CSI reports are then ranked according to their second CSI reporting priority from highest to lowest, utilizing the remaining available CPU allocated to them, until the maximum number of non-AI-related CSI reports (M) allocated to them is determined. non-AI,1 ,satisfy in N represents the CPU usage required for the nth non-AI related CSI report. CPU,non-AI To allocate the total CPU to non-AI related CSI reports, L CPU,non-AI Used CPUs are allocated to the total CPU count in non-AI-related CSI reports. In some examples of this application, at least one of the following situations exists:
[0228] Case 1: N AI The CPU required for non-AI-related UE behavior corresponding to an AI-related CSI report exceeds the remaining available CPU allocated to non-AI-related UE behavior corresponding to an AI-related CSI report, i.e., M AI,1 <N AI Then the remaining N AI -M AI,1 AI-related CSI reports can be allocated to the remaining available CPU for non-AI-related CSI reports according to their third CSI reporting priority, from high to low, until the maximum number of AI-related CSI reports (M) that can be allocated to non-AI-related CSI reports is determined. AI2 ,satisfy UE does not require reporting the remaining N AI -M AI,1 -M AI,2 A CSI report related to AI.
[0229] Case 2: N non-AI The CPU required for a single non-AI-related CSI report exceeds the remaining available CPU allocated to non-AI-related CSI reports, i.e., M non-AI,1 <N non-AI Then the remaining N non-AI -M non-AI,1 Each non-AI-related CSI report can be allocated to the remaining available CPU for the non-AI-related UE behavior corresponding to the AI-related CSI report, in descending order of the fourth CSI reporting priority, until the maximum number of non-AI-related CSI reports (M) that can be allocated to the non-AI-related UE behavior corresponding to the AI-related CSI report are determined. non-AI2 ,satisfy UE does not require reporting the remaining N non-AI -M non-AI,1 -M non-AI,2 A non-AI related CSI report.
[0230] The aforementioned non-AI-related UE behaviors include, but are not limited to, the measurement of reference signals (such as RSRP, channel matrix, precoding matrix, channel measurement, UE location), calculation of model monitoring indicators, calculation of model monitoring results, and generation and reporting of CSI reports.
[0231] In Scenario 2, by dynamically allocating CPU resources, the system ensures that non-AI-related UE behaviors and CSI reports corresponding to AI-related CSI reports are flexibly allocated and prioritized within limited CPU resources, thereby maximizing CPU resource utilization efficiency. Specifically, when the total number of non-AI-related CSI reports and CSI reports exceeds their respective allocated CPU resources, the system will allocate resources according to predefined CSI reporting priorities. Excess CSI reports can be transferred to the resource pool of the other party based on priority, occupying the remaining CPU. This design not only ensures that the system can prioritize reporting high-priority CSI reports but also avoids wasting CPU resources and can dynamically adapt to the actual load and processing capacity of the UE, improving the overall system performance and resource utilization.
[0232] Scenario 3: AI-related CSI reports correspond to AI-related UE behaviors that consume APU resources, while non-AI-related UE behaviors and non-AI-related CSI reports share CPU resources. These CPU resources can be dynamically allocated among the non-AI-related UE behaviors and non-AI-related CSI reports. Assume there are N [unclear - possibly related to a specific type of OFDM symbol]. AI One AI-related CSI report and N that need to be submitted. non-AI For each non-AI-related CSI report that needs to be reported, AI-related CSI reports are allocated to the remaining available CPU of the non-AI-related UE behavior corresponding to the AI-related CSI report, in descending order of their first CSI reporting priority, until the maximum number M of AI-related CSI reports (M) corresponding to the non-AI-related UE behavior corresponding to the AI-related CSI report is determined. AI,1 ,satisfy in N represents the CPU required for the non-AI-related UE behavior corresponding to the nth AI-related CSI report. CPU,AI,3 To allocate the total CPU to non-AI-related UE behaviors corresponding to AI-related CSI reports, L CPU,AI,3 This is the total CPU allocated to non-AI-related UE behaviors corresponding to AI-related CSI reports. Non-AI-related CSI reports are then ranked according to their second CSI reporting priority from highest to lowest, utilizing the remaining available CPU allocated to them, until the maximum number of non-AI-related CSI reports (M) allocated to them is determined. non-AI,1 ,satisfy in N represents the CPU usage required for the nth non-AI related CSI report. CPU,non-AITo allocate the total CPU to non-AI related CSI reports, L CPU,non-AI Used CPUs are allocated to the total CPU count in non-AI-related CSI reports. In some examples of this application, at least one of the following situations exists:
[0233] Case 1: N AI The CPU required for non-AI-related UE behavior corresponding to an AI-related CSI report exceeds the remaining available CPU allocated to non-AI-related UE behavior corresponding to an AI-related CSI report, i.e., M AI,1 <N AI Then the remaining N AI -M AI,1 AI-related CSI reports can be allocated to the remaining available CPU for non-AI-related CSI reports according to their third CSI reporting priority, from high to low, until the maximum number of AI-related CSI reports (M) that can be allocated to non-AI-related CSI reports is determined. AI,2 ,satisfy UE does not require reporting the remaining N AI -M AI,1 -M AI,2 A CSI report related to AI.
[0234] Case 2: N non-AI The CPU required for a single non-AI-related CSI report exceeds the remaining available CPU allocated to non-AI-related CSI reports, i.e., M non-AI,1 <N non-AI Then the remaining N non-AI -M non-AI,1 Each non-AI-related CSI report can be allocated to the remaining available CPU for the non-AI-related UE behavior corresponding to the AI-related CSI report, in descending order of the fourth CSI reporting priority, until the maximum number of non-AI-related CSI reports (M) that can be allocated to the non-AI-related UE behavior corresponding to the AI-related CSI report are determined. non-AI,2 ,satisfy UE does not require reporting the remaining N non-AI -M non-AI,1 -M non-AI,2 A non-AI related CSI report.
[0235] The aforementioned AI-related UE behaviors include, but are not limited to, model training, model updating, and model inference. Non-AI-related UE behaviors include, but are not limited to, the measurement of reference signals (such as RSRP, channel matrix, precoding matrix, channel measurement, and UE location), calculation of model monitoring indicators, calculation of model monitoring results, and generation and reporting of CSI reports.
[0236] In Scenario 3, by dynamically allocating CPU and APU resources, the system effectively achieves flexible resource management for both AI-related and non-AI-related CSI reports. Specifically, the AI portion of AI-related CSI reports (such as model training and inference) exclusively uses APU resources, while the non-AI portion (such as reference signal measurement and model monitoring) shares CPU resources with the non-AI-related CSI reports. This resource allocation method ensures that AI-related computing tasks always have dedicated APU resources for efficient processing, while non-AI-related computing tasks are dynamically allocated CPU resources. Under this mechanism, when the non-AI portion of an AI-related CSI report cannot fully utilize its allocated CPU resources, the remaining resources can be shared by non-AI-related CSI reports. Conversely, when a non-AI-related CSI report cannot fully utilize its allocated CPU resources, the remaining resources can also be shared by the non-AI portion of an AI-related CSI report. This solution improves the system's resource utilization, ensures the processing priority of high-priority tasks, and avoids idle CPU or APU resources. This design can effectively and dynamically adapt to the computational needs of AI / ML-related CSI reports and traditional CSI reports, ensuring the system's efficiency and flexibility in scenarios with multiple tasks and multiple types of CSI reports.
[0237] For scenario 1, the CPU resources allocated to AI-related CSI reports and non-AI-related CSI reports are determined based on at least one of the following parameters:
[0238] Parameters: The proportion of total CPU usage corresponding to AI-related CSI reports to the total CPU usage on the UE side, and the proportion of total CPU usage corresponding to non-AI-related CSI reports to the total CPU usage on the UE side. The total CPU usage on the UE side is reported to the base station by the UE as part of the UE's capabilities. Assume the proportion of total CPU usage corresponding to AI-related CSI reports to the total CPU usage on the UE side is P. CPU,AI The proportion of total CPU usage corresponding to non-AI-related CSI reports to total CPU usage on the UE side is P. CPU,non-AI Then P satisfies CPU,AI +P CPU,non-AI ≤1.
[0239] Parameters: Total CPU usage for AI-related CSI reports and total CPU usage for non-AI-related CSI reports. Assume the total CPU usage on the UE side is N. CPU The total CPU usage for AI-related CSI reports and non-AI-related CSI reports are N respectively. CPU,AI and N CPU,non-AI Then N satisfies CPU,AI +N CPU,non-AI ≤N CPU .
[0240] The above parameters are determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined.
[0241] For scenario 2, the CPU resources allocated to AI-related UE behaviors corresponding to AI-related CSI reports, non-AI-related UE behaviors corresponding to AI-related CSI reports, and non-AI-related CSI reports are determined based on at least one of the following parameters:
[0242] Parameters: The proportion of total CPU usage corresponding to AI-related UE behaviors in AI-related CSI reports to the total CPU usage on the UE side; the proportion of total CPU usage corresponding to non-AI-related UE behaviors in AI-related CSI reports to the total CPU usage on the UE side; and the proportion of total CPU usage corresponding to non-AI-related CSI reports to the total CPU usage on the UE side. The total CPU usage on the UE side is reported to the base station by the UE as part of the UE's capabilities. Assume the proportion of total CPU usage corresponding to AI-related UE behaviors in AI-related CSI reports to the total CPU usage on the UE side is P. CPU,AI,1 The proportion of CPU usage for non-AI-related UE behaviors corresponding to AI-related CSI reports to the total CPU usage on the UE side is P. CPU,AI,2 The proportion of total CPU usage corresponding to non-AI-related CSI reports to total CPU usage on the UE side is P. CPU,non-AI Then P satisfies CPU,AI,1 +P CPU,AI,2 +P CPU,non-AI ≤1.
[0243] Parameters: Total CPU usage for AI-related UE behaviors corresponding to AI-related CSI reports, total CPU usage for non-AI-related UE behaviors corresponding to AI-related CSI reports, and total CPU usage for non-AI-related CSI reports. Assume the total CPU usage on the UE side is N. CPU The AI-related UE behaviors corresponding to AI-related CSI reports, the non-AI-related UE behaviors corresponding to AI-related CSI reports, and the total CPU usage corresponding to non-AI-related CSI reports are respectively N. CPU,AI,1 N CPU,AI,2 N CPU,non-AI Then N satisfies CPU,AI,1 +N CPU,AI,2 +N CPU,non-AI ≤N CPU .
[0244] The above parameters are determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined.
[0245] For scenario 3, the CPU resources allocated to non-AI-related UE behaviors and non-AI-related CSI reports corresponding to AI-related CSI reports are determined based on at least one of the following parameters:
[0246] Parameters: The proportion of total CPU usage corresponding to non-AI-related UE behaviors in AI-related CSI reports to the total CPU usage on the UE side, and the proportion of total CPU usage corresponding to non-AI-related UE behaviors in AI-related CSI reports to the total CPU usage on the UE side. The total CPU usage on the UE side is reported to the base station by the UE as part of its UE capabilities. Assume that the proportion of total CPU usage corresponding to non-AI-related UE behaviors in AI-related CSI reports to the total CPU usage on the UE side is P. CPU,AI,3 The proportion of total CPU usage corresponding to non-AI-related CSI reports to total CPU usage on the UE side is P. CPU,non-AI Then P satisfies CPU,AI,3 +P CPU,non-AI ≤1.
[0247] Parameters: Total CPU usage corresponding to non-AI-related UE behaviors in AI-related CSI reports and total CPU usage corresponding to non-AI-related CSI reports. Assume the total CPU usage on the UE side is N. CPU The total CPU usage for non-AI-related UE behaviors and non-AI-related CSI reports corresponding to AI-related CSI reports are respectively N CPU,AI,3 and N CPU,non-AI Then N satisfies CPU,AI,3 +N CPU,non - AI ≤N CPU .
[0248] The above parameters are determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined.
[0249] In some embodiments of this application, regardless of scenario 1, scenario 2, or scenario 3, the allocation of CPU resources is always based on the proportion or absolute number of the total CPU on the UE side, and these resource allocation methods can be flexibly determined through UE capability reporting, base station configuration, or protocol predefinition. This flexible resource allocation mechanism can effectively ensure that the UE can reasonably allocate CPU resources and maximize resource utilization when various complex AI-related and non-AI-related tasks coexist, while ensuring the processing performance of various CSI reporting tasks and the overall efficiency of the system.
[0250] In some embodiments of this application, the reporting priority of CSI reports (including first CSI reporting priority, second CSI reporting priority, third CSI reporting priority, and fourth CSI reporting priority) can be determined based on at least one of the following parameters:
[0251] The reporting configuration ID determines the priority of CSI reports; different configuration IDs correspond to different priority settings. Secondly, the serving cell index also affects reporting priority, especially in multi-cell configurations, where the serving cell may have a higher priority than non-serving cells. Furthermore, the amount of data reported is also a crucial parameter influencing priority, particularly when whether the reported data carries L1-RSRP (Reference Signal Received Power) or L1-SINR (Signal-to-Noise Ratio) data becomes a key differentiator. Generally, CSI reports carrying L1-RSRP or L1-SINR data have a higher priority than those without these metrics.
[0252] Furthermore, the reporting configuration type also determines the priority of CSI reports, specifically including at least one of the following configuration types: PUSCH-based non-periodic CSI reporting; PUSCH-based semi-persistent CSI reporting; PUCCH-based semi-persistent CSI reporting; and PUCCH-based periodic CSI reporting.
[0253] In determining the reporting priority, the UE behavior corresponding to the report is also a factor. Specifically, the reported behavior can include at least one of the following: non-AI related CSI reports; AI-related UE behavior corresponding to AI-related CSI reports; and non-AI related UE behavior corresponding to AI-related CSI reports.
[0254] CPU resource usage also affects the priority of CSI reports. This CPU usage can be indicated in at least one of the following ways: directly indicating the CPU resources used, usually represented by a non-negative real number, indicating the specific amount of CPU usage; or using an index, where each index corresponds to a predefined range of CPU usage, thereby simplifying the complexity of CPU resource allocation and priority calculation.
[0255] CPU usage is indicated in at least one of the following ways: directly indicating the CPU usage, where the value is a non-negative real number; and / or using an index, where each index corresponds to a range of CPU usage, and the index value ranges from {0, 1, ..., M}. d -1}, where M d This represents the number of index values.
[0256] The four CSI reporting priority rules mentioned above can be the same or different, without limitation. Through flexible combinations of the above parameters, some embodiments of this application can dynamically adjust the CSI reporting priority according to the needs of the UE and the network in scenarios with multiple tasks and multiple types of CSI reports, thereby improving the efficiency of CSI reporting and resource utilization.
[0257] The first example regarding CSI reporting priority is that the CSI reporting priority depends on the CSI reporting configuration ID, serving cell index, reporting volume, reporting configuration type, and the corresponding UE behavior. In this case, the CSI reporting priority value can be expressed as Pri. CSI (f,y,k,c,s)=8·N cells ·M s ·f+2·N sells ·M s ·y+N cells ·M s ·k+M s ·c+s.
[0258] In the above formula, f represents the UE behavior corresponding to CSI reporting, y=0 represents aperiodic CSI reporting based on PUSCH, y=1 represents semi-persistent CSI reporting based on PUSCH, y=2 represents semi-persistent CSI reporting based on PUCCH, y=3 represents periodic CSI reporting based on PUCCH, k=0 represents CSI reporting carrying L1-RSRP or L1-SINR, k=1 represents CSI reporting without carrying L1-RSRP or L1-SINR, c represents the serving cell index, s represents the reporting configuration ID, and N... cells M represents the maximum number of serving cells. s This indicates the maximum number of CSI reporting configuration IDs. The lower the priority value for a CSI report, the higher its priority.
[0259] The first example based on the above formula is that AI-related CSI reports have a higher priority than non-AI-related CSI reports. At the same time, in AI-related UE behaviors corresponding to AI-related CSI reports, model inference has a higher priority than model update, and model update has a higher priority than model training. In this case, f=0 in the above formula represents model inference, f=1 represents model update, f=2 represents model training, and f=3 represents non-AI-related CSI reports.
[0260] The second example based on the above formula is that AI-related CSI reports have a higher priority than non-AI-related CSI reports. At the same time, among the non-AI-related UE behaviors corresponding to AI-related CSI reports, UE behaviors with model monitoring index calculations or model monitoring result calculations have a higher priority than UE behaviors without model monitoring index calculations or model monitoring result calculations. In this case, f=0 in the above formula indicates UE behaviors with model monitoring index calculations or model monitoring result calculations, f=1 indicates UE behaviors without model monitoring index calculations or model monitoring result calculations, and f=2 indicates non-AI-related CSI reports.
[0261] The third example based on the above formula is that AI-related CSI reports have a higher priority than non-AI-related CSI reports. Furthermore, among the UE behaviors corresponding to AI-related CSI reports, non-AI-related UE behaviors involving model monitoring metric calculation or model monitoring result calculation have a higher priority than non-AI-related UE behaviors that do not involve model monitoring metric calculation or model monitoring result calculation, but whose AI-related UE behaviors are model inference. Similarly, UE behaviors that do not involve model monitoring metric calculation or model monitoring result calculation, but whose AI-related UE behaviors are model inference, have a higher priority than AI-related UE behaviors that are model updates. Finally, AI-related UE behaviors that are model updates have a higher priority than AI-related UE behaviors that are model training. In this case, f=0 in the above formula indicates non-AI-related UE behaviors involving model monitoring metric calculation or model monitoring result calculation; f=1 indicates non-AI-related UE behaviors that do not involve model monitoring metric calculation or model monitoring result calculation, but whose AI-related UE behaviors are model inference; f=2 indicates AI-related UE behaviors are model updates; f=3 indicates AI-related UE behaviors are model training; and f=4 indicates non-AI-related CSI reports.
[0262] The second example regarding CSI reporting priority is that the CSI reporting priority depends on the CSI reporting configuration ID, serving cell index, and CPU usage. This example has at least one of the following conditions:
[0263] Scenario 1: The lower the CPU usage of the CSI report, the higher the priority of the CSI report. With the same CPU usage, the lower the serving cell index, the higher the priority of the CSI report. With the same serving cell index, the lower the CSI reporting configuration ID, the higher the priority of the CSI report. When CPU usage is indicated by an index, the priority value of the CSI report can be expressed as Pri. CSI (d,c,s)=N cells ·M s ·d+M s •c+s. In the above formula, d represents the CPU usage index. Here, it is assumed that the smaller the index value, the less CPU is used, and vice versa. c represents the serving cell index, s represents the CSI reporting configuration ID, and N cells M represents the maximum number of serving cells. s This indicates the maximum number of CSI reporting configuration IDs. The lower the priority value of a CSI report, the higher its priority. This applies to the third and fourth CSI reporting priorities, where the remaining available CPU may be limited; reporting CSIs with lower CPU usage can improve CPU resource utilization. Or
[0264] Scenario 2: The higher the CPU usage of the CSI report, the higher the priority of the CSI report. With the same CPU usage, a smaller serving cell index results in a higher CSI report priority. Similarly, with the same serving cell index, a smaller CSI report configuration ID results in a higher CSI report priority. When CPU usage is indicated by an index, the priority value of the CSI report can be expressed as Pri. CSI (d,c,s)=N cells ·M s ·(M d -1-d)+M s •c+s. In the above formula, d represents the CPU usage index. Here, it is assumed that the smaller the index value, the less CPU is used, and vice versa. c represents the serving cell index, s represents the CSI reporting configuration ID, and N cells M represents the maximum number of serving cells. s M represents the maximum number of CSI reported configuration IDs. d This indicates the number of CPU-intensive index values. The lower the priority value for a CSI report, the higher its priority. In this case, CSIs with higher CPU usage can be reported first to reduce the time required to complete the reporting task for CSIs with higher CPU usage within the limited available CPU time.
[0265] The aforementioned CSI reporting priority design achieves efficient resource allocation in various complex scenarios through flexible priority calculation formulas and parameter configurations. The core of this design lies in ensuring that CSI reports are reported in an orderly manner under resource constraints by considering multiple factors such as the type of CSI report (AI-related or non-AI-related), the corresponding UE behavior, CPU usage, serving cell index, and CSI reporting configuration ID, thereby improving system resource utilization efficiency. Based on the priority design of CSI reporting configuration ID, serving cell index, reporting volume, reporting configuration type, and corresponding UE behavior, different characteristics of CSI reports are prioritized, giving priority to AI-related CSI reports over non-AI-related CSI reports, and further refining the priority within AI-related reports. For example, in AI-related CSI reports, model inference takes precedence over model updates, and model updates take precedence over model training, ensuring that resources are prioritized for more critical AI tasks. Secondly, the CPU-based priority design provides two different strategies: in the third and fourth CSI reporting priorities, prioritizing CSI reports with lower CPU usage enables efficient resource utilization under limited resources; while in specific scenarios, prioritizing CSI reports with higher CPU usage allows for the completion of computationally intensive tasks with limited remaining CPU. This flexible priority strategy adapts to resource allocation needs in different scenarios. Overall, this priority design can dynamically adjust the CSI reporting order based on the characteristics of CSI reports and CPU resource usage, effectively improving system resource utilization and reporting efficiency. Especially in complex scenarios where AI-related and non-AI-related CSI reports coexist, it ensures the priority execution of critical tasks (such as AI model inference and monitoring tasks).
[0266] Third embodiment: Calculation of processing resources for high AI-related CSI reports in non-shared scenarios.
[0267] In some examples of this application, the CSI report includes an AI-related CSI report and / or a non-AI-related CSI report. The AI-related CSI report occupies a first portion of the CPU, and the non-AI-related CSI report occupies a second portion of the CPU. The AI portion of the AI-related CSI report occupies a first sub-portion of the first portion of the CPU, and the non-AI portion of the AI-related CSI report occupies a second sub-portion of the first portion of the CPU. In some examples of this application, the first portion of the CPU occupied by the AI-related CSI report and the second portion of the CPU occupied by the non-AI-related CSI report are not shared. In some examples of this application, the non-AI portion of the AI-related CSI report and the non-AI-related CSI report jointly occupy the CPU. In some examples of this application, the order, combination, specific content, and execution of the steps do not limit the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0268] In some examples of this application, the User Equipment (UE) determines the number of CSI reports it can process by allocating CPU resources between AI-related CSI reports and non-AI-related CSI reports. Specifically, AI-related CSI reports and non-AI-related CSI reports can each occupy different portions of the CPU. This allocation can be fixed or dynamically adjusted, depending on the system configuration and current resource utilization. In this embodiment, the UE can determine its CSI report processing capacity in one of two ways: The first way is shared resource allocation, where AI-related CSI reports and non-AI-related CSI reports share CPU resources, and these resources are dynamically allocated between the two types of CSI reports. The UE flexibly allocates resources between the two types of reports based on their respective CSI report priorities, CPU usage, and system configuration to ensure maximum resource utilization. The second way is non-shared resource allocation, where AI-related CSI reports and non-AI-related CSI reports each occupy a separate CPU resource pool. In this mode, the CPU resource allocation for AI-related CSI reports and non-AI-related CSI reports is independent of each other, and the UE can calculate and process the two types of CSI reports separately based on CPU and APU availability. This approach avoids resource contention, ensuring the computational reliability and efficiency of AI-related CSI reports under high load. Regardless of the method used, the UE can dynamically determine the number of CSI reports it can process based on the current CPU resource usage. Through this flexible resource allocation strategy, the system can meet the needs of AI tasks while ensuring the stability and reliability of traditional CSI reports.
[0269] The technical solution of this embodiment is applicable to scenarios where AI-related CSI reports simultaneously consider CPU and APU computation. Specifically, in this scenario, when the UE selects and processes AI-related CSI reports, it needs to ensure that the selected CSI report can obtain resource support from both APU and CPU during the computation process.
[0270] This means that when determining the number of AI-related CSI reports that can be processed, the UE needs to consider not only CPU availability and allocation, but also sufficient APU resources, and that the computing resources of both are coordinated. This solution can effectively ensure that AI-related CSI reports can still be processed efficiently under resource constraints or multi-tasking conditions, thereby improving system resource utilization and the reliability of CSI reports.
[0271] The technical solution of this embodiment has at least one of the following situations:
[0272] Scenario 1: Use APU and CPU to calculate CSI reporting resources.
[0273] Example 1: APU and CPU are determined independently.
[0274] Assume the total number of CPUs and APUs are N respectively. CPU and N APU The CPU or APU usage of the m-th CSI report is recorded as O. CPU and O APU The currently occupied or reserved CPUs and APUs are L. CPU and L APU .
[0275] Considering APU limitations independently, Where M A The number of AI-related CSI reports selected (rules).
[0276] Considering CPU-limited scenarios independently, we can divide them into the following two cases:
[0277] - When traditional CSI reports and the non-AI portion of AI-related CSI reports share the CPU resource pool Where M C M′ represents the number of AI-related CSI reports selected. C The number of traditional CSI reports selected.
[0278] -In AI-related CSI reports, when the CPU resource pool is used independently. Where M C The number of AI-related CSI reports selected.
[0279] The actual number of AI-related CSI reports that can be supported is: M ≤ min(M C M A Assume M A The set of CSI report IDs corresponding to the selected AI-related CSI reports is: M C The set of CSI report IDs corresponding to the selected AI-related CSI reports is: Then, M = |R A ∩R C |, where |R A ∩R C | represents set R A and set R C The number of elements in the intersection of and . Where is the number of elements in the intersection of and . k Report K on behalf of CSI.
[0280] Example 2: APU priority judgment.
[0281] For example, prioritize CSI reports based on APU-limited conditions. First, based on the APU-limited conditions, i.e. Select M A A CSI report, recorded as Then in M A M was selected from the CSI reports. C Each report meets or Record Where M c ≤M A ,and That is, R C It is R A A subset of.
[0282] Example 3: CPU priority determination.
[0283] For example, prioritize CSI reports based on CPU-limited conditions. First, based on the CPU-limited conditions, i.e. or Select M C A CSI report, recorded as Then in M C M was selected from the CSI reports. A Each report meets Record Where M A ≤M C ,and That is, R A It is R C A subset of.
[0284] In Scenario 1 of this embodiment, by selecting and scheduling CPU and APU resources, CSI reports can be filtered based on resource constraints, ensuring that the number of AI-related CSI reports and non-AI-related CSI reports is supported even with limited CPU and APU resources. Scenario 1 of this embodiment achieves efficient selection of AI-related and non-AI-related CSI reports under resource-constrained conditions, ensuring efficient resource allocation and improving the system's computing and communication performance.
[0285] Scenario 2: Using the CPU to calculate CSI reporting resources.
[0286] In this scenario, AI-related and traditional CSI reports occupy separate CPU resource pools. Assume the total CPU resources used for AI-related and traditional reports are N respectively. CPU,AI and N CPU,R18 The CPU or APU usage of the m-th CSI report is recorded as O. CPU,AI and O CPU,R18 The currently occupied or reserved CPUs and APUs are L. CPU,AI and L CPU,R18 .
[0287] The number of AI-related CSI reports is, Where M A The number of AI-related CSI reports selected. The number of traditional CSI reports (e.g., the number of non-AI-related CSI reports) is... Where M C The number of traditional CSI reports selected.
[0288] This embodiment proposes a resource calculation method for processing AI-related CSI reports in a non-shared scenario, ensuring that AI-related CSI reports are independently guaranteed in terms of computing resources and avoiding resource contention. By independently allocating resources between the APU and CPU, this scheme can achieve at least one of the following technical effects: Efficient resource allocation: By allocating independent computing resource pools (such as APU and CPU) for AI-related CSI reports and non-AI-related CSI reports respectively, resource conflicts and contention are avoided, improving resource utilization efficiency. Flexibility and scalability: Based on the actual capabilities of the UE and the current resource occupancy, computing tasks are dynamically allocated between the APU and CPU, ensuring reasonable allocation even when resources are insufficient. Enhanced resource control: Based on the allocation rules of APU and CPU, the processing of AI-related CSI reports becomes more deterministic, especially in the case of multi-task parallelism or resource scarcity, allowing for flexible adjustment of resource allocation strategies. Support for multiple computing modes: This scheme supports two computing modes: one is a mode where AI-related CSI reports and non-AI-related CSI reports share CPU resources, and the other is a mode where each occupies an independent CPU resource pool. Prioritizing AI Tasks: In situations where APU and CPU resources are limited, prioritizing resource allocation for AI-related CSI reports ensures that AI tasks always receive priority computing resources. In summary, this embodiment enables efficient and flexible CSI report computation in non-shared scenarios through independent CPU and APU resource allocation, providing a stable and reliable computing environment for AI-related CSI reports.
[0289] Fourth embodiment: APU occupancy time for all CSI reporting using APU.
[0290] In some examples of this application, the CSI report includes AI-related CSI reports and / or non-AI-related CSI reports. The AI-related CSI reports are measured using an AI processing unit, and the non-AI-related CSI reports are measured using a CPU. Both the AI and non-AI portions of the AI-related CSI reports are calculated using the AI processing unit. In some examples of this application, the CPU usage time is determined by at least one of the following methods: based on the configuration of the CSI reporting volume, combined with the location of the selected measurement reference signal resource and the CSI calculation time corresponding to the current AI task; based on the configuration of the CSI reporting volume, combined with the location of the selected measurement reference signal resource, the transmission time of the CSI report, and the CSI calculation time corresponding to the current AI task; based on the configuration of the CSI reporting volume, combined with the time of receiving the CSI report trigger information, the transmission time of the CSI report, and the CSI calculation time corresponding to the current AI task. In some examples of this application, when the user equipment performs CSI report calculation, it determines the start and end time of the AI processing unit's occupation based on pre-configuration, capability reporting, or base station configuration information, depending on whether CSI reporting is required and the number of tasks. This includes continuous or intermittent time periods within the symbol interval after the reference signal resource or physical downlink control channel (PDCCH) is triggered. In some examples of this application, the order, combination, specific content, and execution of the steps do not limit the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0291] In some examples of this application, the allocation of computing resources differs between AI-related CSI reports and non-AI-related CSI reports. Specifically, AI-related CSI reports typically rely on the APU for computation, while non-AI-related CSI reports primarily use the CPU. This resource allocation ensures that AI-related tasks can fully utilize the efficient computing power of the APU, while non-AI-related tasks utilize the CPU for routine computational operations. In this case, the CPU occupancy time can be determined based on the CSI reporting volume configuration, using at least one of the following methods: (1) combining the location of the selected measurement reference signal resource and the CSI calculation time corresponding to the current AI task; (2) combining the location of the selected measurement reference signal resource, the CSI report transmission time, and the CSI calculation time corresponding to the current AI task; (3) based on the time of receiving the CSI report trigger information, the CSI report transmission time, and the CSI calculation time corresponding to the current AI task. In some examples of this application, when the user equipment performs CSI report calculations, it can determine the start and end times of the AI processing unit's occupancy time based on the number of tasks and whether CSI reporting is required, using at least one of the following methods: pre-configured task timing, UE capability reporting, or dynamic configuration of the base station. For example, within the symbol range of the CSI measurement reference signal resources, continuous or intermittent APU occupancy periods can be allocated to each task. Furthermore, if CSI reporting is triggered by the Physical Downlink Control Channel (PDCCH), the APU occupancy period can begin from a specific symbol after the PDCCH trigger and continue until the CSI calculation for the current task is completed. This flexible timing configuration mechanism ensures that resource utilization efficiency is maximized while meeting task processing requirements.
[0292] Option 1: APU usage time in a single-task scenario.
[0293] If the CSI reporting quantity is none (none), meaning no CSI report is required, then the APU occupancy time in a single-task scenario can take at least one of the following forms:
[0294] Format 1: The CSI report begins occupying the APU from the first symbol of the first reference signal resource within the Kth latest reference signal transmission occasion no later than the CSI reference resource, until it ends at a symbol Z′ symbols away from the last symbol of the last reference signal resource within the last reference signal transmission occasion no later than the CSI reference resource, where K is the number of reference signal transmission occasions required to complete the current task, and Z′ is the CSI calculation time corresponding to the current task, which is determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined.
[0295] Form 2: The CSI report begins occupying the APU from the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity no later than the CSI reference resource, and the APU occupation time corresponding to each reference signal transmission opportunity is X symbols, until the APU occupation ends at a symbol Z′ symbols away from the last symbol of the last reference signal resource within the last reference signal transmission opportunity no later than the CSI reference resource. Here, K is the number of reference signal transmission opportunities required to complete the current task, X is the APU occupation time corresponding to each reference signal transmission opportunity under the current task, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined, and Z′ is the CSI calculation time corresponding to the current task, which is determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the start time of APU occupation is symbol 0, the APU occupation time under this task can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, and M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource within the reference signal transmission opportunity.
[0296] Form 3: The CSI report occupies the APU from the first symbol after the PDCCH that triggers the CSI report until the latest symbol among the symbols that are Z symbols away from the last symbol of the PDCCH that triggers the CSI report and the symbols that are Z′ symbols away from the last symbol of the last reference signal resource within the last reference signal transmission time no later than the last reference signal resource of the CSI reference resource. Here, Z is the CSI calculation time relative to the PDCCH that triggers the CSI report for the current task, and Z′ is the CSI calculation time relative to the reference signal resource for the current task. Both Z and Z′ can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined.
[0297] Form 4: The CSI report begins occupying the APU from the first symbol after the PDCCH that triggers the CSI report, and the APU occupation time corresponding to each reference signal transmission opportunity is X symbols, until the APU occupation ends at the symbol Z′ symbols away from the last symbol of the last reference signal resource within the last reference signal transmission opportunity no later than the CSI reference resource, where X is the APU occupation time corresponding to each reference signal transmission opportunity under the current task, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined, and Z′ is the CSI calculation time corresponding to the current task, which is determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the start time of APU occupation is symbol 0, the APU occupation time under this task can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, D is the symbol difference between the first symbol of the first reference signal resource in the Kth latest reference signal transmission opportunity no later than the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource in the reference signal transmission opportunity.
[0298] Form 5: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupancy time for processing the PDCCH is Y symbols. Simultaneously, the APU occupancy time for each reference signal transmission opportunity is X symbols, until the APU occupancy ends at the symbol Z′ of the last symbol of the last reference signal resource within the last reference signal transmission opportunity no later than the CSI reference resource. Here, X is the APU occupancy time for each reference signal transmission opportunity under the current task, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined; Y is the APU occupancy time for processing the PDCCH that triggers the CSI report, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined; and Z′ is the CSI calculation time for the current task, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the APU occupancy starts at symbol 0, the APU occupancy time under this task can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, D is the symbol difference between the first symbol of the first reference signal resource in the Kth latest reference signal transmission opportunity no later than the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource in the reference signal transmission opportunity.
[0299] If the CSI reporting quantity is not zero, meaning CSI reporting is required, then the APU occupancy time in a single-task scenario can take at least one of the following forms:
[0300] Form 1: The CSI report occupies the APU from the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity no later than the CSI reference resource, until the APU occupies the last symbol of the PUCCH or PUSCH carrying the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task.
[0301] Form 2: The CSI report occupies the APU starting from the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity no later than the CSI reference resource, and the APU occupancy duration corresponding to each reference signal transmission opportunity is X symbols, until the APU occupancy ends at the last symbol of the PUCCH or PUSCH carrying the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and X is the APU occupancy duration corresponding to each reference signal transmission opportunity under the current task, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the start time of APU occupancy is symbol 0, the APU occupancy time under this task can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, and Q is the number of symbols occupied from the first symbol of the first reference signal resource within the last reference signal transmission opportunity of the CSI reference resource to the last symbol of the PUCCH or PUSCH carrying the CSI report.
[0302] Form 3: The CSI report occupies the APU from the first symbol after the PDCCH that triggered the CSI report until the last symbol of the PUCCH or PUSCH that carries the CSI report.
[0303] Form 4: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupation time corresponding to each reference signal transmission opportunity is X symbols, until the APU occupation ends at the last symbol of the PUCCH or PUSCH carrying the CSI report, where X is the APU occupation time corresponding to each reference signal transmission opportunity under the current task, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the start time of APU occupation is symbol 0, the APU occupation time under this task can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, D is the symbol difference between the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity of the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and Q is the number of symbols occupied from the first symbol of the first reference signal resource within the last reference signal transmission opportunity of the CSI reference resource to the last symbol of the PUCCH or PUSCH carrying the CSI report.
[0304] Form 5: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupancy time for processing the PDCCH corresponding to this report is Y symbols. Simultaneously, the APU occupancy time for each reference signal transmission opportunity is X symbols, until the APU occupancy ends at the last symbol of the PUCCH or PUSCH carrying the CSI report. Here, X is the APU occupancy time for each reference signal transmission opportunity under the current task, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined; Y is the APU occupancy time for processing the PDCCH that triggers the CSI report, determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the APU occupancy starts at symbol 0, the APU occupancy time under this task can be expressed as:
[0305] In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, D is the symbol difference between the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity of the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and Q is the number of symbols occupied from the first symbol of the first reference signal resource within the last reference signal transmission opportunity of the CSI reference resource to the last symbol of the PUCCH or PUSCH carrying the CSI report. In this embodiment, the calculation of APU occupancy time is based on the reporting requirements of the CSI report and the configuration of the CSI reference resource, ensuring flexible allocation of APU resources in a single-task scenario. Specifically: In the case of no CSI report reporting: APU occupancy time only involves the duration of the reference signal transmission opportunity, which can be determined according to the number of reference signals in the task and the CSI calculation time (Z'). Even if no CSI report needs to be generated, the APU will still occupy resources to complete data processing or model inference. Different APU occupancy time designs (such as forms 1 to 5) can flexibly adapt to the processing requirements of different tasks. In cases with CSI report reporting: APU occupancy time begins after the reference signal or PDCCH is triggered and continues until the CSI report reporting ends. In this case, the APU needs to consider the symbol duration from the triggering of the CSI report to the PUCCH or PUSCH, while ensuring that the necessary computational tasks have been completed before the report. Through various APU occupancy time designs (such as forms 1 to 5), efficient processing and flexible configuration can be balanced, ensuring that tasks can be completed on time even under limited computing resources. Flexible configuration mechanism: The determination of the above APU occupancy time can be flexibly configured in the following ways: first, through UE capability reporting, where the UE actively reports its supported APU occupancy timing; second, through base station configuration, where the base station dynamically issues APU occupancy timing based on task requirements; and third, through predefined rules, applicable to task processing modes explicitly specified in the protocol. This flexible configuration method ensures that the system can efficiently allocate resources according to actual task requirements and resource status. In summary, this embodiment can flexibly support the allocation of computing time for different task requirements by using various forms of APU time occupation, ensuring efficient utilization of APU resources whether or not a CSI report is submitted.
[0306] Option 2: APU usage time in multi-tasking scenarios.
[0307] If the reporting volume for all tasks is none, meaning there is no need to report CSI, then the APU occupancy time in a multi-task scenario can take at least one of the following forms:
[0308] Format 1: The CSI report occupies the APU starting from the first symbol of the first reference signal resource within the Kth latest reference signal transmission time no later than the CSI reference resource, until the last symbol of the last reference signal resource within the last reference signal transmission time no later than the CSI reference resource is Z′. m The symbol-on-end of each symbol occupies the APU, where K is the number of reference signal transmission opportunities required to complete the current task, and Z′ m The CSI calculation time for multiple tasks depends on the UE's processing method for multiple tasks. The processing method includes, but is not limited to, serial processing, parallel processing, and serial-parallel processing of multiple tasks. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined.
[0309] Format 2: The CSI report occupies the APU starting from the first symbol of the first reference signal resource within the Kth latest reference signal transmission time no later than the CSI reference resource, and the APU occupancy time corresponding to the kth reference signal transmission time (k=0 corresponding to the Kth latest reference signal transmission time no later than the CSI reference resource) is X. m,k The last symbol is Z, up to the last symbol of the last reference signal resource within a time frame no later than the last reference signal transmission time of the CSI reference resource. m The symbol-on-end of ' symbols occupies the APU, where K is the number of reference signal transmission opportunities required to complete the current task, and X m,k Z′ represents the APU occupancy time corresponding to the k-th reference signal transmission opportunity. This time depends on the UE's processing method for reference signal measurements corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. m The CSI calculation time for multiple tasks depends on the UE's processing method for these tasks. This method includes, but is not limited to, serial processing, parallel processing, and serial-parallel processing. It can be determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined methods. Assuming the APU's initial time is symbol 0, the APU's time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, and M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource within the reference signal transmission opportunity.
[0310] Form 3: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggered the CSI report, until the last symbol of the PDCCH that triggered the CSI report is Z. m The symbol and distance of each symbol are no later than the last symbol of the last reference signal resource within the last reference signal transmission time of the CSI reference resource, which is Z′. m The latest symbol among the symbols ends the APU, where Z m Z′ represents the CSI calculation time of the PDCCH for relative triggering CSI reporting corresponding to multiple tasks. m Z is the CSI calculation time for the relative reference signal resources corresponding to multiple tasks. m and Z′ m All depend on the UE's handling method for multiple tasks. The handling method includes, but is not limited to, serial multitasking, parallel multitasking, and serial-parallel multitasking. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined.
[0311] Form 4: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupancy time corresponding to the k-th reference signal transmission timing (not later than the k-th latest reference signal transmission timing corresponding to k=0 of the CSI reference resource) is X. m,k The last symbol is Z′, up to the last symbol of the last reference signal resource within a time frame no later than the last reference signal transmission time of the CSI reference resource. m The sign-on end of each symbol occupies the APU, where X m,k Z′ represents the APU occupancy time corresponding to the k-th reference signal transmission opportunity. This time depends on the UE's processing method for reference signal measurements corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. m The CSI calculation time for multiple tasks depends on the UE's processing method for these tasks. This method includes, but is not limited to, serial processing, parallel processing, and serial-parallel processing. It can be determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined methods. Assuming the APU's initial time is symbol 0, the APU's time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, D is the symbol difference between the first symbol of the first reference signal resource in the Kth latest reference signal transmission opportunity no later than the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource in the reference signal transmission opportunity.
[0312] Format 5: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupancy time for processing the PDCCH is Y symbols. Simultaneously, the APU occupancy time for the k-th reference signal transmission timing (not later than the k-th latest reference signal transmission timing corresponding to k=0 of the CSI reference resource) is X symbols. m,k The last symbol is Z′, up to the last symbol of the last reference signal resource within a time frame no later than the last reference signal transmission time of the CSI reference resource. m The sign-on end of each symbol occupies the APU, where X m,k Z′ represents the APU occupancy time corresponding to the k-th reference signal transmission opportunity. This time depends on the UE's processing method for reference signal measurements corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Y represents the APU occupancy time corresponding to the PDCCH that triggers CSI reporting. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. m The CSI calculation time for multiple tasks depends on the UE's processing method for these tasks. This method includes, but is not limited to, serial processing, parallel processing, and serial-parallel processing. It can be determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined methods. Assuming the APU's initial time is symbol 0, the APU's time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, D is the symbol difference between the first symbol of the first reference signal resource in the Kth latest reference signal transmission opportunity no later than the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource in the reference signal transmission opportunity.
[0313] If the reported amount for all tasks is not 'none', meaning CSI reporting is required, then the APU occupancy time in a multi-task scenario can take at least one of the following forms:
[0314] Form 1: The CSI report occupies the APU from the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity no later than the CSI reference resource, until the APU occupies the last symbol of the PUCCH or PUSCH carrying the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task.
[0315] Format 2: The CSI report occupies the APU starting from the first symbol of the first reference signal resource within the Kth latest reference signal transmission time no later than the CSI reference resource, and the APU occupancy time corresponding to the kth reference signal transmission time (k=0 corresponding to the Kth latest reference signal transmission time no later than the CSI reference resource) is X. m,k The APU is occupied until the last symbol of the PUCCH or PUSCH carrying the CSI report ends, where K is the number of reference signal transmission opportunities required to complete the current task, and X is the number of symbols. m,k Let be the APU occupancy time corresponding to the k-th reference signal transmission timing. This time depends on the UE's processing method for the reference signal measurement corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement, and can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the start time of APU occupancy is symbol 0, the APU occupancy time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, and Q is the number of symbols occupied from the first symbol of the first reference signal resource within the last reference signal transmission opportunity of the CSI reference resource to the last symbol of the PUCCH or PUSCH carrying the CSI report.
[0316] Form 3: The CSI report occupies the APU from the first symbol after the PDCCH that triggered the CSI report until the last symbol of the PUCCH or PUSCH that carries the CSI report.
[0317] Form 4: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupancy time corresponding to the k-th reference signal transmission timing (not later than the k-th latest reference signal transmission timing corresponding to k=0 of the CSI reference resource) is X. m,kThe APU is occupied by a symbol until the last symbol of the PUCCH or PUSCH carrying the CSI report ends, where X m,k Let be the APU occupancy time corresponding to the k-th reference signal transmission timing. This time depends on the UE's processing method for the reference signal measurement corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement, and can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the start time of APU occupancy is symbol 0, the APU occupancy time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, D is the symbol difference between the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity of the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and Q is the number of symbols occupied from the first symbol of the first reference signal resource within the last reference signal transmission opportunity of the CSI reference resource to the last symbol of the PUCCH or PUSCH carrying the CSI report.
[0318] Format 5: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupancy time for processing the PDCCH is Y symbols. Simultaneously, the APU occupancy time for the k-th reference signal transmission timing (not later than the k-th latest reference signal transmission timing corresponding to k=0 of the CSI reference resource) is X symbols. m,k The APU is occupied by a symbol until the last symbol of the PUCCH or PUSCH carrying the CSI report ends, where X m,k Let Y be the APU occupancy time corresponding to the k-th reference signal transmission timing. This time depends on the UE's processing method for the reference signal measurement corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Y is the APU occupancy time corresponding to the PDCCH that triggers CSI reporting. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Assuming the start time of APU occupancy is symbol 0, the APU occupancy time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbols occupied by the APU, P is the period of the reference signal transmission opportunity, D is the symbol difference between the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity of the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task, and Q is the number of symbols occupied from the first symbol of the first reference signal resource within the last reference signal transmission opportunity of the CSI reference resource to the last symbol of the PUCCH or PUSCH carrying the CSI report.
[0319] If the reporting volume for some tasks is none, while the reporting volume for other tasks is not none, the APU occupancy time in a multi-task scenario can take at least one of the following forms:
[0320] Format 1: The CSI report occupies the APU starting from the first symbol of the first reference signal resource within the Kth latest reference signal transmission time no later than the CSI reference resource, until the first symbol of the first reference signal resource within the last reference signal transmission time no later than the CSI reference resource is Z′. max The symbol-on-end of each symbol occupies the APU, where K is the number of reference signal transmission opportunities required to complete the current task, and Z′ max It can be represented as Z′ max =max(Q-1,M+Z′) m,none -1), where Q is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the PUCCH or PUSCH carrying the CSI report within the last reference signal transmission time no later than the last reference signal transmission time of the CSI reference resource, M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource within the reference signal transmission time, and Z′ m,none The CSI calculation time for multi-tasks with a reported amount of 'none' depends on the UE's processing method for multi-tasks. The processing method includes, but is not limited to, serial multi-tasks, parallel multi-tasks, and serial-parallel multi-tasks. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined.
[0321] Format 2: The CSI report occupies the APU starting from the first symbol of the first reference signal resource within the Kth latest reference signal transmission time no later than the CSI reference resource, and the APU occupancy time corresponding to the kth reference signal transmission time (k=0 corresponding to the Kth latest reference signal transmission time no later than the CSI reference resource) is X. m,kThe first symbol of the first reference signal resource is Z′, up to the first symbol of the first reference signal resource within a time frame no later than the last reference signal transmission of the CSI reference resource. max The symbol-on-end of a symbol occupies the APU, where K is the number of reference signal transmission opportunities required to complete the current task, and X... m,k Z′ represents the APU occupancy time corresponding to the k-th reference signal transmission opportunity. This time depends on the UE's processing method for reference signal measurements corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. max It can be represented as Z′ max =max(Q-1,M+Z′) m,none -1), where Q is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the PUCCH or PUSCH carrying the CSI report within the last reference signal transmission time no later than the last reference signal transmission time of the CSI reference resource, M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource within the reference signal transmission time, and Z′ m,none The CSI calculation time for multiple tasks with a reported quantity of 'none' depends on the UE's processing method for multiple tasks. This method includes, but is not limited to, serial processing, parallel processing, and serial-parallel processing. It can be determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined methods. Assuming the APU's initial time is symbol 0, the APU's time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbol occupied by the APU, and P is the period of the reference signal transmission opportunity.
[0322] Form 3: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggered the CSI report, until the last symbol of the PDCCH that triggered the CSI report is Z″. m " ax The sign-on end of each symbol occupies the APU, where Z″ m " ax It can be represented as Z″ max =max(S+Q,Z) m,none ,S+M+Z′ m,noneIn this expression, S is the symbol difference between the first symbol of the first reference signal resource within the last reference signal transmission time of the CSI reference resource and the first symbol after the last symbol of the PDCCH that triggered the CSI report; Q is the number of symbols occupied from the first symbol of the first reference signal resource within the last reference signal transmission time of the CSI reference resource to the last symbol of the PUCCH or PUSCH carrying the CSI report; and Z is the symbol difference between the first symbol of the first reference signal resource within the last reference signal transmission time of the CSI reference resource and the last symbol of the PUCCH or PUSCH carrying the CSI report. m,none Z′ is the CSI calculation time of the PDCCH that triggers CSI reporting for multiple tasks with a reporting quantity of 'none'. m,none Z is the CSI calculation time for the relative reference signal resources corresponding to multiple tasks with a reporting quantity of 'none'. m,none and Z′ m,none All depend on the UE's multi-task processing method, which includes but is not limited to multi-task serial, multi-task parallel, and multi-task serial parallel. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. M is the number of symbols occupied by the first symbol of the first reference signal resource to the last symbol of the last reference signal resource within the reference signal transmission time.
[0323] Form 4: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupancy time corresponding to the k-th reference signal transmission timing (not later than the k-th latest reference signal transmission timing corresponding to k=0 of the CSI reference resource) is X. m,k The first symbol of the first reference signal resource is Z′, up to the first symbol of the first reference signal resource within a time frame no later than the last reference signal transmission of the CSI reference resource. max The sign-on end of each symbol occupies the APU, where X m,k Z′ represents the APU occupancy time corresponding to the k-th reference signal transmission opportunity. This time depends on the UE's processing method for reference signal measurements corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. max It can be represented as Z′ max =max(Q-1,M+Z′) m,none -1), where Q is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the PUCCH or PUSCH carrying the CSI report within the last reference signal transmission time no later than the last reference signal transmission time of the CSI reference resource, M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource within the reference signal transmission time, and Z′ m,noneThe CSI calculation time for multiple tasks with a reported quantity of 'none' depends on the UE's processing method for multiple tasks. This method includes, but is not limited to, serial processing, parallel processing, and serial-parallel processing. It can be determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined methods. Assuming the APU's initial time is symbol 0, the APU's time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbol occupied by the APU, P is the period of the reference signal transmission opportunity, and D is the symbol difference between the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity of the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task.
[0324] Format 5: The CSI report occupies the APU starting from the first symbol after the PDCCH that triggers the CSI report, and the APU occupancy time for processing the PDCCH is Y symbols. Simultaneously, the APU occupancy time for the k-th reference signal transmission timing (not later than the k-th latest reference signal transmission timing corresponding to k=0 of the CSI reference resource) is X symbols. m,k The first symbol of the first reference signal resource is Z′, up to the first symbol of the first reference signal resource within a time frame no later than the last reference signal transmission of the CSI reference resource. max The sign-on end of each symbol occupies the APU, where X m,k Z′ represents the APU occupancy time corresponding to the k-th reference signal transmission opportunity. This time depends on the UE's processing method for reference signal measurements corresponding to multiple tasks. The processing method includes, but is not limited to, serial measurement, parallel measurement, and serial-parallel measurement. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. Y represents the APU occupancy time corresponding to the PDCCH that triggers CSI reporting. It can be determined by at least one of the following methods: UE capability reporting, base station configuration, and / or predefined. max It can be represented as Z′ max =max(Q-1,M+Z′) m,none -1), where Q is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the PUCCH or PUSCH carrying the CSI report within the last reference signal transmission time no later than the last reference signal transmission time of the CSI reference resource, M is the number of symbols occupied from the first symbol of the first reference signal resource to the last symbol of the last reference signal resource within the reference signal transmission time, and Z′ m,noneThe CSI calculation time for multiple tasks with a reported quantity of 'none' depends on the UE's processing method for multiple tasks. This method includes, but is not limited to, serial processing, parallel processing, and serial-parallel processing. It can be determined through at least one of the following methods: UE capability reporting, base station configuration, and / or predefined methods. Assuming the APU's initial time is symbol 0, the APU's time under multiple tasks can be expressed as: In the above formula, n is the ordinal number of the symbol occupied by the APU, P is the period of the reference signal transmission opportunity, and D is the symbol difference between the first symbol of the first reference signal resource within the Kth latest reference signal transmission opportunity of the CSI reference resource and the last symbol of the PDCCH that triggers the CSI report, where K is the number of reference signal transmission opportunities required to complete the current task.
[0325] This embodiment achieves efficient resource allocation in single-task scenarios through a flexible APU occupancy time design. Regardless of whether CSI reports need to be generated and reported, the APU can flexibly allocate computing time according to task requirements, ensuring efficient computation even under resource constraints or parallel task conditions. This solution adapts to various task requirements and network environments through three configuration methods: UE capability reporting, dynamic base station configuration, and predefined rules, providing strong guarantees for the reliability and flexibility of AI-related CSI reports in 5G / 6G systems. This embodiment also provides a flexible and efficient APU resource allocation and management mechanism through its APU occupancy time design in multi-task scenarios. When no CSI reports are required for any task (reporting volume is zero), APU resources are used only for data processing or model inference, ensuring efficient task computation. When a task involves CSI report reporting, APU resources dynamically switch between data processing and report generation, allowing high-priority tasks to obtain APU resources first. This design flexibly adapts to different task processing modes (serial, multi-task parallel, multi-task serial parallel) through various forms (such as forms 1 to 5) and supports dynamic adjustments based on UE capabilities, base station configuration, or predefined rules. This mechanism significantly improves the utilization efficiency of APU resources and ensures efficient processing of AI-related CSI reports in multi-tasking scenarios.
[0326] Fifth embodiment: Determine O based on UE capabilities and other parameters CPU The method.
[0327] In some examples of this application, the capability information of the user equipment includes the number N of available AI processing units of the user equipment, the number O of AI processing units required by the user equipment for CSI reports corresponding to multiple use cases, feature sets, models, or CSI report types, and the number N of available CPUs of the user equipment. CPUAnd / or the number of CPUs required by the user equipment for the CSI reports corresponding to the multiple use cases, feature sets, models, or CSI report types. CPU The instructions include: AI-related O CPU This maps to each use case, feature set, model, or CSI report type, and / or indicates the allocation of CPU resources between AI-related and non-AI-related CSI reports. In some examples of this application, the O is determined by a predefined mapping relationship based on the user equipment type, the capability information, and / or the configured use cases, feature sets, or models. In some examples of this application, when the user equipment reports its capability information to the base station, it reports the O of multiple CSI reports corresponding to the feature groups, models, or use cases supported by the user equipment. CPU Value. In some examples of this application, when the user equipment reports its capability information to the base station or in an uplink message, it reports the functional groups, models, or use case sets supported by the user equipment; through a predefined mapping relationship, it determines the O corresponding to the AI-related CSI report. CPU Value. N is, for example, N in some examples of this application. APU O, for example, is O in some examples of this application. APU In some examples of this application, the order, combination, specific content, and whether or not the steps are executed do not constitute a limitation on the embodiments of this application. They can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0328] In this embodiment, when reporting capability information, the UE can indicate its CPU resource usage (O) under different task types and configuration conditions in various ways. CPU These methods aim to ensure that the base station can accurately and rationally allocate computing resources according to the UE's capabilities and task requirements. Specifically, the UE can include at least one of the following in its capability information: UE computing capability information: including the number of available AI processing units (N, such as N... APU ), number of CPUs (N) CPU ), and the AI processing units (O, such as O) required for each task type (functional group, model, or use case). APU ) and CPU resources (O CPU This information helps the base station determine the UE's computing resource limits and task processing capabilities. CPU Method of determination: In this embodiment, O CPU This can be determined through a predefined mapping relationship. This mapping relationship is automatically calculated based on the UE type, reported capability information, and configured task type (such as function group, model, or use case). CPUThe predefined mapping relationships not only enhance configuration flexibility but also ensure efficient resource utilization across various task configurations.
[0329] Flexible choice between explicit or implicit reporting: Explicit indication: The UE can directly report the corresponding O for each task in the capability report (such as RRC message). CPU Numerical values. Upon receiving the capability report, the base station can clearly understand the resource usage of the UE under different task configurations, avoiding the complexity of subsequent resource configuration. Implicit indication: The UE can report only the function groups, models, and use cases it supports. The base station automatically determines the corresponding O(n) values for these tasks based on predefined mapping rules. CPU Numerical values. This approach simplifies the UE reporting process while ensuring that the base station can flexibly adjust its configuration. Predefined mapping table (Table 1B): In this embodiment, Table 1B shows the combination relationships between UE capabilities and functional groups, models, and use cases, and assigns a corresponding O to each combination. CPU Numerical values. Table 1B further details the O values under different combinations of functional groups, models, and use cases. CPU Allocation. Through these tables, the base station can quickly determine the CPU resource usage of each task on the UE side, avoiding problems of resource redundancy or insufficiency.
[0330] Through this design, this embodiment effectively ensures the rational allocation of CPU resources in multi-tasking and multi-configuration scenarios, avoiding resource waste or insufficient configuration, and ensuring system stability and efficiency. The base station can flexibly adjust task configurations based on the information reported by the UE, thereby maximizing the UE's computing power.
[0331] This embodiment demonstrates the determination of O using multiple known parameters. CPU Multiple methods for determining O. When the protocol predefines O CPU When determining the method, NW can do so if the UE has not reported O. CPU It is also possible to have an expectation of the number of reports that UEs can make, thereby optimizing the configuration of CSI reports and avoiding the additional resource and signaling overhead of redundant configurations.
[0332] In this embodiment, O CPU Configuration can be performed based on UE capabilities and task type (functional group / model / use case), and the specific O can be determined through the combination mapping in Table 1B. CPU Numerical values. The design of Table 1B ensures that the UE can flexibly report CPU resource usage under different task types and configuration conditions.
[0333] Table 1B: UE Capabilities and Functional Groups / Models / Use Cases to O CPU Combination mapping.
[0334] OCPU O can be implemented using explicit or implicit methods. CPU The report was submitted.
[0335] Table 1B shows the combination relationships between UE capabilities and functional groups, models, and use cases, and assigns a corresponding O to each combination. CPU Numerical values. In this table, UE capabilities represent the computing capabilities reported by the UE, including supported function groups and model types. Function groups indicate the different task categories the UE can handle, such as group A, group B, and group C, each corresponding to different types of computing tasks. Models refer to the models used for different task types, such as AI models, statistical models, or hybrid models. Use cases further describe specific task application scenarios, such as CSI prediction, beam selection, and channel estimation.
[0336] In Table 1B, each combination (functional group, model, and use case) corresponds to a specific O. CPU Numerical values, such as O1, O2, O2, etc. Each O CPU The numerical values are further indicated by indices A, B, and C, specifying their exact CPU usage under different configurations. Furthermore, types A, B, and C represent different task processing modes, such as high-priority tasks, regular tasks, or low-priority tasks. This classification allows for precise control over CPU resource allocation, ensuring a clear distinction in CPU resource usage across different task types.
[0337] O CPU The reporting can be implemented in two ways: explicit and implicit. In explicit reporting, the UE explicitly indicates its capabilities under different task types through capability reporting messages (such as RRC messages). CPU CPU usage. For example, a UE can report that a task in group A uses O1 CPU resources, while a task in group B uses O4 CPU resources. In implicit reporting, O... CPU The allocation is automatically determined through predefined rules or base station configuration. The base station can automatically calculate O based on the UE's function group and model type. CPU The numerical value. For example, for group A and index A, O CPU It can be automatically mapped to O1. This O CPU The mapping table design not only ensures the UE's flexibility in multi-tasking and multi-configuration scenarios but also simplifies the configuration process. The base station can quickly determine the CPU resource allocation for each task based on the UE's reported information or predefined rules, thereby achieving efficient resource management and task scheduling. This design is particularly suitable for scenarios involving multi-task parallel processing and AI-related tasks, significantly improving system resource utilization efficiency and task processing reliability.
[0338] Method 1: Explicit Instruction: In this method, the UE directly reports the CPU resource usage (O) of the CSI reports corresponding to each functional group, model, and use case through capability reports (such as RRC messages). CPU Specifically, the UE clearly lists the indexes of the functional groups, models, and use cases it supports in its capability report, and reports the CPU resources it uses in the CSI report corresponding to each configuration.
[0339] For example, the UE can explicitly indicate this in the capability report using the following format as shown in Table 2:
[0340] Table 2: UE Capability Report Format.
[0341] By using this explicit indication method, the base station can directly understand the CPU resource usage of the UE under different tasks and configurations, thus making CSI report configuration and resource allocation more accurate and efficient. This method ensures that the UE can flexibly adapt to various task types while reducing the complexity of resource configuration for the base station.
[0342] Method 2: Implicit Indication: In this method, the UE indirectly reports its supported functional groups, models, and use cases through capability reports, without directly reporting the CPU resource usage (O) corresponding to each configuration. CPU The operating procedure is as follows:
[0343] Step 1: The UE reports the functional groups, models and use cases it supports in the capability report.
[0344] Optionally, the UE can further report the currently available function groups, models, and use cases in capability reports or other uplink messages (such as RRC or UCI messages). This optional reporting can dynamically reflect the task types and models supported by the UE at a specific time.
[0345] Step 2: Based on the information reported by the UE, the NW selects and activates the corresponding function group, model, or use case.
[0346] Once the NW determines the AI-related CSI report types to be configured, the NW will automatically determine the CPU resource consumption (O) required for these CSI reports based on predefined mapping rules. CPU ).
[0347] The advantage of this implicit indication method is that it simplifies the UE's reporting process and avoids directly reporting the CPU usage corresponding to each task. At the same time, network-side devices can flexibly select task configurations based on the UE's capabilities and ensure efficient resource allocation.
[0348] This embodiment employs multiple methods to ensure that the UE can effectively report and configure its CPU resource usage (O) under different task types.CPU This improves the efficiency of system resource utilization and the reliability of task processing. Specifically, the UE can report the AI processing units (N) it supports in its capability information. APU N_APU), CPU(N) CPU ) and the AI processing units (O) required for each task type (such as functional groups, models, or use cases). APU ) and CPU resources (O CPU This information is used to automatically calculate O through predefined mapping relationships. CPU Numerical values ensure flexible adaptation in multi-task and multi-configuration scenarios. The UE can choose to explicitly report (directly reporting O values for each task type). CPU () or implicit reporting (only the task type is reported; the base station determines O based on the mapping relationship) CPU Table 1B illustrates the combination relationship between UE capabilities and task types, clarifying the O under different task combinations. CPU Allocation. This design ensures that the base station can efficiently determine and allocate the UE's CPU resources, regardless of whether the UE explicitly reports O. CPU Both methods enable flexible resource management and task scheduling, avoiding resource waste or unstable task processing. This embodiment flexibly determines and reports the UE's CPU resource usage (O) through various methods such as predefined mapping, explicit and implicit reporting. CPU This ensures that base stations can efficiently allocate resources and achieve efficient resource management and task scheduling in multi-task and multi-configuration scenarios.
[0349] Sixth embodiment: Enhanced CPU mechanism for CSI reporting using CPU.
[0350] This embodiment defines the CPU usage time when using the CPU for AI-related CSI reporting. Since AI model inference requires more computing resources than traditional algorithms, we need to allocate more computing resources when performing AI-related CSI reporting.
[0351] When using a traditional CPU for computation, the total computational resources R can be represented as follows: in t represents the number of CPUs used. i The product of CPU usage time and CPU time usage determines the total CPU usage. Therefore, when we expand the total computing resources R used in AI-related CSI reports, we can do so from two dimensions, corresponding to the two options below.
[0352] Option 1: O CPU The quantity has increased.
[0353] O for a specific CSI reporting task within a unit of time CPUTo expand in quantity, O CPU The time taken remains unchanged.
[0354] The CPU usage of AI-related CSI report processing on an OFDM symbol segment is shown in at least one of the following:
[0355] O CPU =M, M∈{1,2,3,4}. If the high-level parameter reportQuantity is set to 'P-CRI', 'P-CRI-RSRP', 'P-SSBRI', 'P-SSBRI-RSRP', 'P-L1-RSRP', 'P-cri-SINR', 'P-SSBRI-SINR', 'P-SINR', 'P-cri-RSRP-Index', 'P-ssbRI-RSRP-Index', 'P-cri-SINR-Index', 'P-ssb-Index-SINR-Index', or 'none' and CSI-RS-ResourceSet is not configured as 'trs-Info'.
[0356] O CPU =X*N TRP ,X∈{1,2,4,8}. If the high-level parameter reportQuantity is set to 'P-CRI-PMI', 'P-CRI-RI-PMI', 'P-CRI-PMI-CQI', or 'P-cri-RI-PMI-CQI', and codebookType is set to 'typeII-CJT-r18' or 'typeII-CJT-PortSelection-r18'.
[0357] If the high-level parameter reportQuantity is set to 'P-CRI-PMI', 'P-CRI-RI-PMI', 'P-CRI-PMI-CQI', or 'P-cri-RI-PMI-CQI', and codebookType is set to 'typeII-Doppler-r18' or 'typeII-Doppler-PortSelection-r18'.
[0358] If the CSI-RS resource set used for channel measurement is aperiodic and K CSI-RS resources are configured, where Y1 is indicated by the UE capability:
[0359] If the CSI-RS resource set used for channel measurement is periodic or semi-persistent and a CSI-RS resource is configured, where N4 is configured by the higher-layer parameters of the RRC and Y2 is indicated by the UE capability.
[0360] In other cases, O CPU =K s Where K_s is the number of CSI-RS resources in the CSI-RS resource set used for channel measurements.
[0361] Option 2: O CPU Time enhancement.
[0362] O for a specific CSI reporting task per unit of time CPU Unchanged, for O CPU The time occupied is extended.
[0363] Method 1: Expand by addition.
[0364] As shown in Table 3, the CSI computation latency requirement considered when processing AI-related CSI reports has an additional time difference compared to the CSI computation latency requirement considered when processing traditional CSI reports. Of the original CSI computation time, X... μ No extension is needed because it is configured by NW via RRC.
[0365] Table 3: AI / ML CSI computation latency requirements.
[0366] In some examples of this application, the latency requirements for AI / ML-related CSI computations are shown in Table 3. This table lists the latency requirements for different types of AI / ML computation tasks, with latency represented by symbols, categorized into three types: Z... AL 1. Z AI 2 and Z AI 3. Z AL 1 indicates the latency requirement for AI / ML computation type 1. Z AI 2 indicates the latency requirement for AI / ML computation type 2. Z AI The value 3 indicates the latency requirement for AI / ML computation type 3, using a different calculation formula. For example, some latency is expressed as a minimum value, meaning that the AI / ML computation latency in a task should be the minimum of these two values. Each row in the table represents a task or computation scenario. This latency requirement design ensures that AI / ML computation can be flexibly configured in multi-task parallel or multi-configuration scenarios, maximizing latency optimization while meeting computational needs. By explicitly listing the latency and computation method of each task, this table provides the system with an efficient basis for CSI computation scheduling.
[0367] Method 2: Expand by multiplication.
[0368] As shown in Table 4, the CSI computation latency requirement considered when processing AI-related CSI reports is multiplied by a time factor compared to the CSI computation latency requirement considered in traditional CSI reports. In the original CSI computation time, X... μ No extension is needed because it is configured by NW via RRC.
[0369] Table 4: AI / ML CSI computation latency requirements.
[0370] This embodiment improves the computational efficiency of AI-related CSI reporting tasks by increasing CPU resource usage and time. Specifically, this embodiment proposes at least one enhancement method: one is to increase the number of CPUs (O CPU This ensures that more computing resources are allocated per unit of time, especially for AI-related CSI reporting tasks, whose computational complexity and resource requirements are significantly higher than traditional tasks. Depending on the task type and configuration conditions, O CPU The system can dynamically adjust according to predefined rules to ensure sufficient CPU support for computationally intensive tasks. Secondly, it further enhances computing power by extending CPU usage time. Latency requirements can be extended using addition or multiplication to ensure flexible configuration and efficient execution of AI / ML computations in multi-task parallel or complex task scenarios. Addition extension adds a fixed time difference to the traditional CSI computation latency, while multiplication extension amplifies latency requirements by introducing a time coefficient. Tables 3 and 4 show the latency calculation formulas for these two extension methods, which can flexibly adapt to different task types and computational needs. In summary, this embodiment, through flexible CPU usage and time extension design, ensures that AI-related CSI reporting tasks can be completed efficiently even under resource constraints, thereby improving the system's computational resource utilization efficiency and task processing reliability.
[0371] Seventh embodiment: CPU usage time for CSI reporting using APU and CPU.
[0372] In some examples of this application, the CSI report includes AI-related CSI reports and / or non-AI-related CSI reports. The AI-related CSI reports occupy a portion of the AI processing unit and CPU, while the non-AI-related CSI reports occupy the CPU. The AI portion of the AI-related CSI reports is calculated using the AI processing unit, and the non-AI portion is calculated using the CPU. In some examples of this application, the AI processing unit occupancy time and / or CPU occupancy time in the CSI reports are determined by at least one of the following methods: the CPU occupancy time is a piecewise function, and the AI processing unit occupancy time is a continuous function; the AI processing unit occupancy time includes at least one of measurement data collection and inference, and the CPU occupancy time includes at least one of Physical Uplink Control Channel (PUCCH) or Physical Uplink Shared Channel (PUSCH) generation; the AI processing unit occupancy time includes inference processing, and the CPU occupancy time includes at least one of measurement data collection and PUCCH or PUSCH generation. In some examples of this application, the order, combination, specific content, and execution of the steps do not limit the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0373] In some examples of this application, the time spent by the AI processing unit and the CPU can be dynamically adjusted based on task type, task stage, and resource usage. Specifically, when the CSI report involves AI-related computational tasks (such as model inference), the AI processing unit is primarily responsible for model inference computation, while the CPU is responsible for preprocessing measurement data and generating reporting results. In this scenario, the time spent by the AI processing unit and the CPU can be coordinated in the following way: on the one hand, the AI processing unit immediately enters the inference stage after completing the collection of measurement data to calculate the AI-related CSI report; on the other hand, the CPU prepares the relevant measurement data during the AI processing unit's inference process and immediately generates and reports the CSI report after the inference is completed. This task allocation ensures that the AI processing unit and the CPU can cooperate efficiently, avoiding resource idleness or task delays, especially in multi-task or multi-configuration scenarios, enabling optimal resource allocation.
[0374] This embodiment provides a method for determining CPU occupancy time, specifically for scenarios involving model inference on the UE side. In this case, the number of CSI reports uploaded is not "none," meaning a CSI report is uploaded. In this scenario, CPU occupancy time begins when CSI calculation and model inference start and continues until the final generation and transmission of the CSI report are completed. Specifically, CPU occupancy will continue until the last symbol of the PUCCH or PUSCH carrying the CSI report. This design ensures that CPU resources are always effectively allocated during CSI report generation and uploading, avoiding resource waste and ensuring the efficiency of CSI report calculation and uploading.
[0375] In the embodiments of this application, Method 1 provides a method for determining the occupancy time of an artificial intelligence processing unit (e.g., an APU) and CPU, particularly suitable for AI-related CSI reports. Specifically, Figure 5A is a schematic diagram of an APU occupancy time in an embodiment of this application that only includes model inference. As shown in Figure 5A, the APU occupancy time includes two stages: data collection and model inference, while the CPU occupancy time is mainly used for the generation and reporting of CSI reports. In this method, the APU starts occupancy from the first or last symbol of the first reference signal resource within the observation window, continuously performing data collection and model inference until the first symbol after model inference is completed. This design ensures that the APU is always in a highly efficient operating state during the calculation of CSI reports, with both data collection and model inference completed on the APU. Meanwhile, the CPU occupancy time starts from the first symbol after model inference is completed and continues until the last symbol of the PUCCH or PUSCH carrying the CSI report is completed. This design ensures that the CPU is primarily responsible for the generation and reporting of CSI reports, avoiding idle CPU resources during model inference. Furthermore, by separating and seamlessly integrating the time spent by the APU and CPU, this method can effectively improve the efficiency of CSI report generation and submission, ensuring efficient resource utilization in AI-related task scenarios.
[0376] In the embodiments of this application, Method 2 provides a method for determining APU and CPU occupancy time, particularly suitable for AI-related CSI reports, where the division of labor between the APU and CPU is clearly defined. Specifically, Figure 5B is a schematic diagram of an embodiment of this application where APU occupancy time only includes model inference. As shown in Figure 5B, the APU is only responsible for model inference computation, while the CPU is responsible for data measurement (data collection) and the generation and reporting of CSI reports. In this method, CPU occupancy time begins from the first or last symbol of the first reference signal resource within the CSI report observation window and ends after the first symbol of the last reference signal resource within the observation window or after the last symbol plus Y symbols. This design ensures that the CPU efficiently completes the data collection task during the observation phase of the CSI report, providing the necessary data support for subsequent CSI report generation. Subsequently, APU occupancy time begins from the last symbol of the last reference signal resource within the observation window plus Y or plus (Y+Z) symbols and ends after the first symbol after model inference is completed. Through this design, the APU only occupies resources during the model inference phase, avoiding resource waste during data collection and ensuring efficient utilization of computing resources. Finally, the CPU resumes its use from the first symbol after model inference ends to generate and report CSI reports, continuing until the last symbol of the PUCCH or PUSCH carrying the CSI report. This design rationally allocates the task time slots of the APU and CPU, ensuring that both operate efficiently on their respective focused tasks, thereby improving the overall processing efficiency of CSI reports. The value of Y is configured by the network side via RRC messages, MAC-CE messages, or DCI. The value of Z is configured by the network side via RRC messages, MAC-CE messages, or DCI, or obtained by the UE side and reported to the base station via UCI, RRC messages, or Sounding Reference Signal (SRS). The unit of Y can be a slot, symbol, or millisecond. Specifically, the value of Y can be related to the time-domain location of the CSI reference resource (CSI RS). For example, the value of Y can be equal to the time interval between the first symbol of the last reference signal resource within the observation window or the last symbol and the time-domain location of the CSI reference resource. Alternatively, the value of Y can be equal to the time interval between the first symbol of the last reference signal resource within the observation window or the last symbol and the time-domain location of the CSI reference resource plus a time offset value. This scheme, by configuring the values of Y and / or Z, combined with various reporting methods such as RRC messages, MAC-CE messages, DCI, and / or UE-side reporting, achieves time slot, symbol, or millisecond-level time control, thereby enhancing the flexibility and accuracy of time-domain scheduling of CSI reference resources.
[0377] In the embodiments of this application, Method 3 provides a method for determining APU and CPU occupancy time, particularly applicable to AI-related CSI reports. The APU is solely responsible for model inference computation, while the CPU is responsible for data measurement (data collection) and the generation and reporting of CSI reports. Figure 5C is a schematic diagram illustrating an embodiment of this application where APU occupancy time only includes model inference. As shown in Figure 5C, in this method, CPU occupancy time begins from the first or last symbol of the first reference signal resource within the CSI report observation window. At this time, the CPU occupancy time corresponding to each reference signal transmission timing is fixed at X symbols, and the CPU continuously occupies these symbols for data collection. This occupancy time continues until the first symbol of the last reference signal resource within the observation window plus X symbols. This design ensures that the CPU efficiently completes the data collection task during the data measurement phase, providing a reliable data foundation for subsequent CSI report generation. After the data measurement phase ends, APU occupancy time begins from the last symbol of the last reference signal resource within the observation window plus Y or (Y+Z) symbols, and ends with the first symbol after model inference is completed. This design ensures that the APU only occupies resources during the model inference phase, avoiding resource waste during data collection and improving the efficiency of computing resource utilization. The CPU resumes operation on the first symbol after model inference ends, used for generating and reporting CSI reports, until the last symbol of the PUCCH or PUSCH carrying the CSI report ends. This design ensures efficient allocation of CPU resources during CSI report generation and reporting, thereby guaranteeing the efficiency of CSI report computation and reporting. The value of X is configured by the network side via RRC messages, MAC-CE messages, or DCI.
[0378] In the embodiments of this application, Method 4 provides a method for determining APU and CPU occupancy time, particularly applicable to AI-related CSI reports. The APU is solely responsible for model inference computation, while the CPU is responsible for data measurement (data collection) and the generation and reporting of CSI reports. Figure 5D is a schematic diagram illustrating an embodiment of this application where APU occupancy time only includes model inference. As shown in Figure 5D, in this method, CPU occupancy time begins from the first or last symbol of the Kth or last reference signal resource within the CSI report observation window. The CPU continuously occupies this time, performing data measurement and collection tasks, and ends after the first or last symbol plus Y symbols of the last reference signal resource within the observation window. This design ensures that the CPU can fully utilize the reference signal resources within the observation window to complete all measurement data collection required for the CSI report. Subsequently, APU occupancy time begins after the last symbol plus Y symbols or (Y+Z) symbols of the last reference signal resource within the observation window and continues until the first symbol after model inference ends. This design guarantees that the APU only occupies resources during the model inference phase, avoiding resource waste during data collection and thus improving system resource utilization efficiency. The CPU is repurposed for the first symbol after model inference ends, used to generate and report CSI reports, until the last symbol of the PUCCH or PUSCH carrying the CSI report ends. This design ensures efficient allocation of CPU resources during CSI report generation and reporting, guaranteeing the efficiency of CSI report computation and reporting. This method achieves efficient resource allocation during the measurement, inference, and reporting processes of CSI reports through flexible CPU and APU time allocation.
[0379] This embodiment provides a method for effectively configuring APU and CPU usage time in AI-related CSI reports, ensuring efficient resource allocation during the measurement, inference, and reporting processes of CSI reports. Specifically, CSI report processing is divided into two main parts: data measurement and report generation are handled by the CPU, while model inference is handled by the APU. By rationally dividing tasks between the APU and CPU, the system can flexibly adjust resource allocation according to task type and task stage. In Method 1, the APU undertakes both data collection and model inference tasks simultaneously from the beginning of the observation window, while the CPU only generates and reports the report after inference is completed. Methods 2 and 3 further optimize task allocation: the CPU focuses on data collection, while the APU only occupies resources during the model inference stage, avoiding resource waste during data collection. Method 4 adopts a more flexible CPU and APU usage time design, dynamically adjusting resource allocation based on task priority and task requirements. Overall, this embodiment ensures efficient processing of AI-related CSI reports in multi-task, multi-configuration scenarios through clear task division and flexible resource scheduling, avoiding resource idleness and task delays, and significantly improving the system's resource utilization efficiency and task processing reliability.
[0380] Eighth Example: Merged transmission of AI-related CSI reports.
[0381] In some examples of this application, the configuration for merging the CSI reports corresponding to the multiple tasks into the single CSI report is based on default merging, uplink control information (UCI) format merging, activation merging, and / or configuration merging. In some examples of this application, in the absence of signaling configuration, if multiple AI-related CSI reports share the same CSI Resource Configuration Identifier ID, Synchronization Signal Block (SSB) group index, or CSI Resource Configuration ID, then the default merging is adopted. The UCI-based merging means that when the configured UCI format is X, and X≥4, the merged transmission of K AI-related CSI reports is triggered, where K is a predefined value or configured via Radio Resource Control (RRC) messages, Media Access Control (MAC-CE) messages, or Downlink Control Information (DCI) messages. The activation-based merging means that the merging of K AI-related CSI reports is activated through a 1-bit field in the DCI or RRC message, and the merged report indicates the merged report through a 1-bit field or information containing the CSI report configuration ID. The configuration-based merging means that the CSI report group used for merging is configured through the RRC message, and the K AI-related CSI reports to be merged are indicated by the CSI report configuration ID. In some examples of this application, in a single CSI report, the report header indicates the order of merging CSI reports, or the corresponding CSI report ID is carried in each report segment. In some examples of this application, the configuration of the merged CSI report is determined by the base station through preset, priority selection, or reconfiguration. In some examples of this application, the order, combination, specific content, and execution of the steps do not limit the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0382] This embodiment provides a method for transmitting CSI reports when multiple AI-related tasks share a measurement set, aiming to solve the problems of merging and independently transmitting AI-related CSI reports. When AI-related tasks share the same measurement set or measurement resources, the system can employ various methods to merge or independently transmit these CSI reports to improve transmission efficiency and reduce signaling overhead.
[0383] When multiple AI-related tasks share a measurement set, AI-related CSI reports can be merged or transmitted independently through at least one of the following methods:
[0384] Method 1: Basic method of merged transmission.
[0385] Method 1: Default Report Merging. In the absence of other signaling configurations, if multiple AI-related tasks share the same measurement set (e.g., configured with the same CSI-resourceSetConfigID, the same SSB group Index, or the same CSI-resourceConfigID in CSI-resourceSetConfig), their AI-related CSI reports will be automatically merged into a single report. This merging reduces signaling overhead and ensures that all tasks sharing the same measurement set can efficiently transmit their CSI reports.
[0386] Method 2: Report Configuration Association Merging. If multiple AI-related CSI reports have configurations (i.e., CSI-reportConfig) that are associated with the same measurement set, such as the same CSI-resourceSetConfigID or the same SSB group Index, these reports will be automatically merged. This method further ensures that even if the report configurations are different, they can still be merged into a single report as long as their measurement resources are the same.
[0387] Method 3: Merging based on UCI format. When the UCI format is configured as a new special format (e.g., UCI Format X, where X≥4), multiple (K) AI-related CSI reports can be merged. The UCI format configuration is sent to the UE from the network (NW) side. K can be determined using at least one of the following options: Option 1: K is a fixed value, pre-configured by the protocol. Option 2: K is a fixed value, configured by RRC messages, MAC-CE messages, or DCI.
[0388] Method 4: Dynamic Activation of Merged Reports. K AI-related CSI reports can be merged under specific conditions via a 1-bit dynamic activation instruction in a DCI or RRC message. Specific implementations include: K can be configured using at least one of the following options: Option 1: K is a fixed value, pre-configured by the protocol. Option 2: K is a fixed value, configured by an RRC message, MAC-CE message, or DCI.
[0389] Identify the K AI-related CSI reports that need to be merged: Merge the K most recent AI-related CSI reports and indicate them as merged reports in the merged report.
[0390] Mark it as a merged report in the merged report: Option 1: Indicate it as a merged report using 1 bit. Option 2: Include information about the original AI-related CSI report configuration, such as CSI-reportConfigID, in the merged report.
[0391] Method 5: Configure merged reports directly in RRC.
[0392] Specify the K AI-related CSI reports to be merged directly in the RRC message. This method ensures flexibility and accuracy in the merging process by explicitly indicating the CSI-reportConfigID of the merged reports.
[0393] Using the above method, this embodiment can flexibly merge and independently transmit CSI reports based on the configuration and measurement resources of multiple AI-related tasks. This design not only effectively reduces signaling burden but also ensures that CSI reports for AI-related tasks can be transmitted efficiently and reliably, making it suitable for complex scenarios with multiple tasks and configurations.
[0394] Method 2: Configure CSI reports for merged transmission.
[0395] In each of the above methods, the configuration of the merged CSI report can be determined by at least one of the following methods:
[0396] When K AI-related tasks are associated with the same CSI report configuration, the merged CSI report will be directly configured with resources under that configuration. This method ensures configuration consistency without requiring additional configuration modifications. Secondly, when K AI-related tasks are associated with K different CSI report configurations, at least one of the following options can be used for configuration:
[0397] Option 1: Specify Configuration via DCI or RRC: NW can directly specify the configuration of one of the K AI-related CSI reports via DCI or RRC messages for use in the merged CSI report. This method is flexible and efficient, enabling rapid determination of the merged report configuration.
[0398] Option 2: Default selection of the k-th report configuration: Unless otherwise specified, the configuration of the k-th AI-related CSI report is used by default, where 1≤k≤K. This approach ensures the determinism of the configuration and avoids configuration conflicts.
[0399] Option 3: Predefined Protocol Principle: Select the configuration for a specific AI-related CSI report based on predefined rules in the protocol. For example, use the configuration of the first or last (Kth) AI-related CSI report. Select the configuration that includes AI-related CSI reports in UCI Format X. This approach ensures a balance between compatibility and flexibility in report merging.
[0400] Option 4: Direct Reconfiguration: Reconfigure the merged CSI report via RRC to ensure that the merged report meets the task requirements. This method is suitable for scenarios where resource configuration needs to be redefined during the merging process.
[0401] Through the above-mentioned configuration methods, this embodiment can flexibly determine the configuration of the merged report in the scenario of merging CSI reports for multiple AI-related tasks, ensuring the accuracy and reliability of the merged report.
[0402] Method 3: Merge the new UCI format for transmission.
[0403] In this embodiment, the merged CSI report adopts a new long format UCI, wherein the mapping order of the K AI-related CSI reports in the merged report can be determined by at least one of the following methods, as shown in Table 5, where the report header indicates the order:
[0404] Method 1: Report header indicates order. In the merged UCI report, the order of the K AI-related CSI reports is first explicitly indicated in the report header. This order can be determined in two ways:
[0405] Directly indicating the order: The report header explicitly lists the order of the K AI-related CSI reports, for example... This indicates that the K reports are merged and transmitted in a specified order.
[0406] Indirect Order Indication: The report header indicates the change in the order of K AI-related CSI reports in the form of a numerical sequence. For example, assuming the original configuration order is {rid1,rid2,rid3,rid4}, its default order label is "1234", which means the order has not changed. If the order label is "3214", it means that the order of the CSI reports has been adjusted to {rid3,rid2,rid1,rid4}.
[0407] Table 5: Order of Report Headers
[0408] In this approach, each report section (Report Quantity#k of CSI Report#k) in the merged UCI report retains its original report content, ensuring the integrity and accuracy of the report data.
[0409] Method 2: The report header does not indicate the order.
[0410] In this approach, the UCI report header does not explicitly indicate the order of CSI reports; instead, it directly lists each AI-related CSI report and its corresponding CSI report ID sequentially. The format of each CSI report is shown in Table 6.
[0411] Table 6: CSI report format.
[0412] This design simplifies the configuration of the report order. By default, the system lists each CSI report in the configured order, while ensuring that the report content is consistent with its original configuration through the CSI report ID in each report section.
[0413] Regardless of the method used, the merged UCI report can flexibly support the merged transmission of various AI-related CSI reports, ensuring efficient report transmission even under resource-constrained or high-load conditions.
[0414] Method 4: Non-merging transmission.
[0415] In this embodiment, when multiple AI-related tasks share the same measurement set, CSI reports are transmitted in a non-merged manner by default. This non-merged transmission mode ensures that each AI-related CSI report is generated and transmitted independently, avoiding report confusion or resource conflicts that may occur during the merging process.
[0416] Furthermore, the system can dynamically control the activation state of merged transmission via a 1-bit indication in the DCI or RRC message. When the 1-bit indication is active, multiple AI-related CSI reports can be merged and transmitted; when the indication is inactive, the reports are transmitted independently in a non-merged manner. This dynamic activation mechanism allows the system to flexibly switch between merged and non-merged modes to adapt to different task requirements and network conditions.
[0417] Additionally, when the UCI format is configured as format 3 or format 4, the system will also force a non-merged transmission mode. In this case, each AI-related CSI report is generated and sent independently and will not be merged into a single report. This non-merged transmission mode ensures the independence and accuracy of the reports, and is particularly suitable for scenarios where tasks are independent of each other.
[0418] This embodiment implements the merging and independent transmission of AI-related CSI reports through multiple methods, ensuring flexible response to different task requirements in multi-task and multi-configuration scenarios. Merging transmission offers several options, including default merging, UCI-based merging, dynamically activated merging, and RRC-configured merging, allowing the system to flexibly adjust the CSI report transmission mode according to task configuration and signaling requirements. Regarding the configuration of merged reports, the system can determine the merged CSI report format through preset, priority selection, or reconfiguration, thereby ensuring report accuracy and compatibility. Simultaneously, the non-merging transmission mode is also retained, particularly suitable for scenarios where tasks are independent and there is no shared measurement set. This flexible CSI report transmission mechanism not only significantly reduces signaling overhead but also improves system resource utilization and report transmission efficiency, providing strong support for the efficient processing of AI-related tasks in 5G / 6G systems.
[0419] Ninth Example: Merged transmission of AI-related CSI reports.
[0420] The technical solution of this embodiment defines the effective time of CSI report prediction results and designs a cost-effective filtering and reporting mechanism based on this. In this mechanism, the time position of the CSI report relative to the time position of the prediction result can be divided into at least one case, as shown in Figures 6A and 6B. In some examples of this application, the order, combination, specific content, and whether or not the steps are executed do not constitute a limitation on the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0421] In the first scenario, as shown in Figure 6A, the CSI report is generated after the prediction results. In this case, the system first generates the AI model prediction results and then performs CSI calculations and reports based on these results. By delaying the generation and submission of the CSI report until after the prediction results, it ensures that the report content fully reflects the current network condition and the accuracy of the prediction results.
[0422] In the second scenario, as shown in Figure 6B, the CSI report precedes the prediction results. In this case, the CSI report is completed before the prediction results are generated. This approach is suitable for tasks planned in advance or scenarios requiring rapid reporting before the prediction results are generated. Through this design, the system can ensure the timeliness of the CSI report even before the prediction results are generated.
[0423] By designing at least one of the above CSI reporting time locations, this embodiment not only improves the flexibility and accuracy of reporting, but also effectively reduces the system's signaling overhead through a filtering reporting mechanism.
[0424] Scenario 1: The CSI report is located after the forecast results.
[0425] When a CSI report appears after all or part of the forecast results, the forecast results received by the other end before the CSI report are considered "past" or "expired." In this case, we need to define the effective duration T of the forecast results in the CSI report. efct This serves as the basis for determining whether this portion of the CSI report is usable. It includes at least one of the following steps:
[0426] Step 1: The base station needs to be configured with T in RRC. efct For example, configure T in CSI-reportConfig. efct Configure T in DCI (CSI request). efct Indicate T in MAC-CE efct .
[0427] Step 2: The UE, according to the configured T efct Filter CSI reports. Assume the CSI report time is T. rpt For any k-th prediction result, assume its prediction time is T. k Then when T k +T efct ≤T rpt If the prediction result is within a certain time frame, it is considered invalid (expired). The UE filters the CSI reports of the prediction results and does not report prediction results within the invalid time frame. The UE indicates the unreported prediction results in the following way: The UE uses an M-bit bitmap to indicate M unreported prediction results; the UE indicates the index of all unreported prediction results; and the UE indicates the index of the invalid prediction result closest to the CSI report time.
[0428] Scenario 2: The CSI report precedes the forecast results.
[0429] When the CSI report precedes all or part of the forecast results, all forecast results are considered usable. In this case, we can still use CSI report filtering to reduce reporting overhead. This includes at least one of the following steps:
[0430] Step 1: The base station needs to be configured with T in RRC. Gefct T Gefct This indicates the retention period for a specific CSI forecast report. For example, in CSI-reportConfig, T is configured... Gefct Configure T in DCI (CSI Request). Gefct Indicate T in MAC-CE Gefct .
[0431] Step 2: The UE, according to the configured T Gefct Filter CSI reports. Assume the CSI report time is T. rpt For any k-th prediction result, assume its prediction time is T. k Then when T k +T efct ≤T rpt If the prediction result is not valid (expired), then the UE considers it invalid. The UE filters the CSI reports of the most recent prediction results. Specifically, assuming the most recent CSI prediction result is... T Gefct Other prediction results within RQ t-1 ,RQ t-2 ,…,RQ t-n Then RQ t The prediction results in RQ t-1 ,RQ t-2 ,…,RQt-n The UE compares the prediction results in the previous CSI report with the previous one. If the similarity is less than a certain threshold, the portion of the most recent prediction results that can find similar values in the historical prediction results will not be reported, and the corresponding historical prediction results will be indicated. The unreported prediction results are indicated in the following ways: The UE uses an M-bit bitmap to indicate M unreported prediction results. The UE indicates the corresponding historical prediction results for all unreported prediction results; it also indicates the ID of the historical CSI report containing the historical prediction results. The UE indicates the index of the historical prediction result in the historical CSI report. The UE indicates the time sequence of the historical prediction result in the most recent CSI report, or the corresponding prediction time.
[0432] This embodiment provides a technical solution for merging, transmitting, and filtering reports based on the effective time of AI-related CSI report prediction results, aiming to reduce signaling overhead and improve the accuracy and timeliness of reports. This is achieved by defining the effective duration of the prediction results (T). efct By configuring T and filtering conditions, this scheme ensures that only CSI prediction results within the valid time frame are reported, thereby avoiding the transmission of invalid or redundant data. For cases where the CSI report follows the prediction results, as shown in Figure 6A, the system configures T... efct The system filters and reports only prediction results that are within their validity period. For cases where the CSI report precedes the prediction results, as shown in Figure 6B, the system uses T... Geftt The system filters and compares the similarity between the latest and historical forecasts, reporting only those with significant differences. This mechanism not only reduces signaling overhead but also ensures the accuracy and timeliness of CSI reports, adapting to complex scenarios with multiple tasks and configurations.
[0433] 10th Implementation: Priority is given to reporting performance monitoring results.
[0434] In some embodiments, FIG7A is a flowchart illustrating a communication method performed by a user equipment according to an embodiment of this application. The communication method includes: operation 701A: determining the reporting priority of multiple Channel State Information (CSI) reports; operation 702A: reporting the CSI reports to a base station according to the reporting priority of the CSI reports; wherein, the CSI reports include AI-related CSI reports and non-AI-related CSI reports, and the user equipment determines the reporting priority of the CSI reports based on whether the CSI reports carry physical layer reference signal received power L1-RSRP or physical layer signal interference noise ratio L1-SINR, the CSI reports being performance monitoring results or intermediate quantities of whether the AI-related CSI reports carry the L1-RSRP or the L1-SINR, and / or based on the reporting time-domain behavior and bearer type of the CSI reports. Through the above technical solution, the communication method of this embodiment flexibly adjusts the transmission order of CSI reports based on the reporting priority, reporting time-domain behavior, and bearer type of CSI reports, ensuring that AI-related and non-AI-related CSI reports can be efficiently reported according to task importance, thereby improving system resource utilization and communication performance. The order, combination, specific content, and execution of operations 701A and 702A described above do not constitute a limitation on the embodiments of this application and can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0435] In some embodiments, Figure 7B is a flowchart illustrating a communication method executed by a base station according to an embodiment of this application. The communication method includes: Operation 701B: receiving multiple Channel State Information (CSI) reports reported by a user equipment, wherein the CSI reports are based on the reporting priority of the CSI reports. The CSI reports include AI-related CSI reports and non-AI-related CSI reports. The user equipment determines the reporting priority of the CSI reports based on whether the CSI reports carry Physical Layer Reference Signal Received Power (L1-RSRP) or Physical Layer Signal-to-Interference-Noise Ratio (L1-SINR). The CSI reports are performance monitoring results or intermediate values regarding whether the AI-related CSI reports carry the L1-RSRP or L1-SINR, and / or based on the reporting time-domain behavior and bearer type of the CSI reports. Through the above technical solution, the communication method of this embodiment dynamically adjusts the transmission order of CSI reports based on the reporting priority, ensuring that AI-related and non-AI-related CSI reports can be efficiently reported according to task importance, thereby improving system resource utilization and communication performance. The order, combination, specific content, and execution of the above operations 701B do not limit the embodiments of this application. They can be flexibly adjusted according to actual needs to meet different scenarios and technical requirements.
[0436] In some embodiments, the reporting priority of the CSI report satisfies at least one of the following rules: when a performance monitoring result or intermediate quantity report of a CSI report carrying the L1-RSRP or the L1-SINR conflicts with a CSI report carrying the L1-RSRP or the L1-SINR, the performance monitoring result or the intermediate quantity report has higher priority; when a performance monitoring result or intermediate quantity report of a CSI report not carrying the L1-RSRP or the L1-SINR conflicts with a CSI report not carrying the L1-RSRP or the L1-SINR, the performance monitoring result or the intermediate quantity report has higher priority; when an AI-related CSI report conflicts with a non-AI-related CSI report, the AI-related CSI report has higher priority; when an AI-related CSI report carrying the L1-RSRP or the L1-SINR conflicts with a non-AI-related CSI report carrying the L1-RSRP or the L1-SINR, the AI-related CSI report has higher priority.
[0437] In some embodiments of this application, to ensure efficient transmission and accurate reporting of both AI-related and non-AI-related CSI reports, the UE transmits CSI reports through a defined reporting priority mechanism. Specifically, the UE determines the priority of each CSI report based on whether it carries L1-RSRP or L1-SINR and whether it is an AI-related report type. This priority determination considers not only whether the report carries L1-RSRP or L1-SINR, but also the report's reporting time-domain behavior (e.g., aperiodic, semi-persistent, or periodic) and bearer type (e.g., via PUCCH or PUSCH). This design ensures that, under conditions of limited system resources or complex channel conditions, CSI reports with a significant impact on system performance, especially those involving AI-related tasks such as AI beam management, CSI prediction, or CSI compression, are transmitted first. Based on this, the base station can also flexibly configure CSI resources and adjust report reception strategies according to the CSI report priority information to ensure system resource utilization efficiency and communication performance.
[0438] Specifically, in the existing standard, for two overlapping PUSCHs, if the UE has not configured sTx-2Panel, or if the UE has configured two different coresetPoolIndex values for different ControlResourceSets in the active downlink BWP through the higher-layer parameter PDCCH-Config, and the UE has configured sTx-2Panel and the two overlapping PUSCHs are associated with the same coresetPoolIndex value, then according to the provisions of the existing standard, the priority rule applies to physical channels with the same priority index.
[0439] The priority values for CSI reporting associations are as follows: Pri iCSI (y,k,c,s)=2·N cells ·M s ·y+N cells ·M s ·k+M s ·c+s. Where: when y=0, it indicates that aperiodic CSI reports are carried via PUSCH; when y=1, it indicates that semi-persistent CSI reports are carried via PUSCH; when y=2, it indicates that semi-persistent CSI reports are carried via PUCCH; when y=3, it indicates that periodic CSI reports are carried via PUCCH; when k=0, it indicates that the CSI report carries L1-RSRP or L1-SINR; when k=1, it indicates that the CSI report does not carry L1-RSRP or L1-SINR; c represents the serving cell index, N cells This represents the value of the higher-level parameter `maxNrofServingCells`; for CSI reports configured with `LTM-CSI-ReportConfig`, `c` represents the serving cell index value configured for that report; `s` represents `reportConfigID`, and `M` represents the value of the higher-level parameter `maxNrofServingCells`. s This represents the value of the high-level parameter `maxNrofCSI-ReportConfigurations`; for CSI reports configured with `LTM-CSI-ReportConfig`, `s` represents `LTM-CSI-ReportConfigID`, and `M` represents the value of the parameter `maxNrofCSI-ReportConfigurations`. s This represents the value of the high-level parameter maxNrofLTM-CSI-ReportConfigurations; when associated with the first CSI report, the Pri... iCSI The value of (y,k,c,s) is less than the Pri associated with the second CSI report. iCSI When the value is (y,k,c,s), the first CSI report is considered to have a higher priority than the second CSI report.
[0440] However, for some AI-based features or functions, such as AI-based beam management, AI-based CSI prediction, AI-based CSI compression, and AI-based CSI prediction and compression, performance monitoring is required to ensure the reliability of AI performance. For performance monitoring at the base station or UE-assisted performance monitoring, the UE needs to report performance monitoring results, intermediate performance data, or monitored performance indicators. As a new type of UCI, the reporting priority of these data needs to be considered. This embodiment provides at least one solution for this:
[0441] Option 1: CSI reporting priority is differentiated based on whether the reported data carries L1-RSRP or L1-SINR. For example, AI-based beam management CSI reports carry L1-RSRP or L1-SINR, while AI-based CSI prediction and AI-based CSI compression do not. Reporting performance monitoring results, intermediate performance data, or monitored performance metrics can also be considered a type of CSI. This CSI is further differentiated based on whether it refers to performance monitoring results, intermediate performance data, or monitored performance metrics for CSI that carries L1-RSRP or L1-SINR, or performance monitoring results, intermediate performance data, or monitored performance metrics for CSI that does not carry L1-RSRP or L1-SINR. Therefore, the following priority values are defined for CSI reporting: Pri iCSI (y,k,c,s)=2·N cells ·M s ·y+N cells ·M s ·k+M s ·c+s.
[0442] Wherein: when y = 0, it indicates that aperiodic CSI reports are carried via PUSCH, and / or AI-based aperiodic CSI reports are carried via PUSCH; when y = 1, it indicates that semi-persistent CSI reports are carried via PUSCH, and / or AI-based semi-persistent CSI reports are carried via PUSCH; when y = 2, it indicates that semi-persistent CSI reports are carried via PUCCH, and / or AI-based semi-persistent CSI reports are carried via PUCCH; when y = 3, it indicates that periodic CSI reports are carried via PUCCH, and / or AI-based periodic CSI reports are carried via PUCCH. Bearing; and / or when k=0, it indicates that the CSI report carries L1-RSRP or L1-SINR, and / or the performance monitoring results, intermediate performance monitoring values, or monitored performance indicator reports are for models, characteristics, or functions for which the CSI report carries L1-RSRP or L1-SINR; when k=1, it indicates that the CSI report does not carry L1-RSRP or L1-SINR, and / or the performance monitoring results, intermediate performance monitoring values, or monitored performance indicator reports are for CSI reports that do not carry L1-RSRP or L1-SINR; the meanings of other parameters are the same as those of the traditional Pri iCSI The meanings of (y,k,c,s) are the same in the calculation formula.
[0443] Option 2: CSI reporting priority is differentiated based on whether the reported data carries L1-RSRP or L1-SINR. For example, AI-based beam management CSI reports carry L1-RSRP or L1-SINR, while AI-based CSI prediction and AI-based CSI compression do not. Reporting performance monitoring results, intermediate performance monitoring data, or monitored performance indicators can also be considered a type of CSI. This CSI is differentiated based on whether it refers to performance monitoring results, intermediate performance monitoring data, or monitored performance indicators for CSI that carries L1-RSRP or L1-SINR, or performance monitoring results, intermediate performance monitoring data, or monitored performance indicators for CSI that does not carry L1-RSRP or L1-SINR. Therefore, the following priority values are defined for CSI reporting: Pri iCSI (y,k,c,s)=2·N cells ·M s ·y+N cells ·M s ·k+M s •c+s. Where: when y=0, it indicates that non-periodic CSI reports are carried via PUSCH; when y=1, it indicates that semi-persistent CSI reports are carried via PUSCH; when y=2, it indicates that semi-persistent CSI reports are carried via PUCCH; when y=3, it indicates that periodic CSI reports are carried via PUCCH; and / or when k=0, it indicates that CSI reports carry L1-RSRP or L1-SINR, AI-based CSI reports carry L1-RSRP or L1-SINR, and / or performance monitoring results, performance monitoring... Intermediate quantities or monitored performance indicator reports are for models, features, or functions that carry L1-RSRP or L1-SINR in CSI reports; when k=1, it indicates that CSI reports do not carry L1-RSRP or L1-SINR, AI-based CSI reports do not carry L1-RSRP or L1-SINR, and / or performance monitoring results, intermediate quantities of performance monitoring, or monitored performance indicator reports are for CSI reports that do not carry L1-RSRP or L1-SINR; the meanings of other parameters are the same as those of traditional Pri iCSI The meanings of (y,k,c,s) are the same in the calculation formula.
[0444] Option 3: CSI reporting priority is differentiated based on whether the reported data carries L1-RSRP or L1-SINR. For example, AI-based beam management CSI reports carry L1-RSRP or L1-SINR, while AI-based CSI prediction and AI-based CSI compression do not. Reporting performance monitoring results, intermediate performance monitoring data, or monitored performance indicators can also be considered a type of CSI. This CSI is differentiated based on whether it refers to performance monitoring results, intermediate performance monitoring data, or monitored performance indicators for CSI that carries L1-RSRP or L1-SINR, or performance monitoring results, intermediate performance monitoring data, or monitored performance indicators for CSI that does not carry L1-RSRP or L1-SINR. Therefore, the following priority values are defined for CSI reporting: Pri iCSI (y,k,c,s)=2·N cells ·M s ·y+N cells ·M s ·k+M s ·c+s. Where: when y=0, it means that aperiodic CSI reports are carried through PUSCH, aperiodic performance monitoring results, intermediate performance monitoring data, or monitored performance metrics reports are carried through PUSCH, and / or AI-based aperiodic CSI reports are carried through PUSCH; when y=1, it means that semi-persistent CSI reports are carried through PUSCH, semi-persistent performance monitoring results, intermediate performance monitoring data, or monitored performance metrics reports are carried through PUSCH, and / or AI-based semi-persistent CSI reports are carried through PUSCH; when y=2, it means that semi-persistent CSI reports are carried through PUCCH, semi-persistent performance monitoring results, intermediate performance monitoring data, or monitored performance metrics reports are carried through PUSCH, and / or AI-based semi-persistent CSI reports are carried through PUSCH. Intermediate quantities that can be monitored, or performance metrics reports that can be monitored, are carried via PUCCH, and / or AI-based semi-persistent CSI reports are carried via PUCCH; when y=3, it means that periodic CSI reports are carried via PUCCH, periodic performance monitoring results, intermediate quantities of performance monitoring, or performance metrics reports that can be monitored are carried via PUCCH, and / or AI-based periodic CSI reports are carried via PUCCH; and / or when k=0, it means that the CSI report carries L1-RSRP or L1-SINR; when k=1, it means that the CSI report does not carry L1-RSRP or L1-SINR; the meanings of other parameters are the same as those of traditional Pri iCSI The meanings of (y,k,c,s) are the same in the calculation formula.
[0445] Regarding the above-mentioned solutions, considering that CSI reports based on different methods or with different reporting content may have the same reporting priority value (Pri), iCSI(y,k,c,s), in order to further distinguish the priority of CSI reports from different methods or with different reporting content, this embodiment predefines or stipulates the following: the above schemes must satisfy at least one of the following rules:
[0446] Under the same time-domain reporting behavior and bearing method, such as non-periodic CSI reports being carried through PUSCH, when performance monitoring results, intermediate performance monitoring quantities, or monitored performance indicator reports for CSI reports carrying L1-RSRP or L1-SINR conflict with CSI reports carrying L1-RSRP or L1-SINR, regardless of Pri... iCSI Regardless of the value of (y,k,c,s), performance monitoring results, intermediate performance monitoring values, or performance indicator reports have higher priority. This ensures continuous monitoring of AI models or AI-based features or functions, guaranteeing the stability of their performance. Similarly, CSI reports carrying L1-RSRP or L1-SINR have higher priority, ensuring continuous CSI availability and thus better performance. These CSI reports carrying L1-RSRP or L1-SINR can be based on AI methods or non-AI methods.
[0447] Under the same time-domain reporting behavior and bearing method, such as non-periodic CSI reports being carried through PUSCH, when performance monitoring results, intermediate performance monitoring quantities, or monitored performance indicator reports for CSI reports without L1-RSRP or L1-SINR conflict with CSI reports without L1-RSRP or L1-SINR, regardless of Pri... iCSI Regardless of the value of (y,k,c,s), performance monitoring results, intermediate performance monitoring values, or performance indicator reports have higher priority. This ensures continuous monitoring of AI models or AI-based features or functions, guaranteeing the stability of their performance. Alternatively, CSI reports without L1-RSRP or L1-SINR have higher priority, ensuring continuous CSI availability and thus better performance. These CSI reports without L1-RSRP or L1-SINR can be based on AI methods or non-AI methods.
[0448] Under the same time-domain reporting behavior and carrying method, such as both being aperiodic CSI reports carried via PUSCH, when an AI-based CSI report carrying L1-RSRP or L1-SINR conflicts with another CSI report carrying L1-RSRP or L1-SINR, regardless of Pri... iCSIRegardless of the value of (y,k,c,s), AI-based CSI reports carrying L1-RSRP or L1-SINR have higher priority, which helps the system achieve better performance because the same function usually yields more benefits from AI-based implementations than traditional implementations; or CSI reports carrying L1-RSRP or L1-SINR have higher priority, which helps stabilize system performance.
[0449] Under the same time-domain reporting behavior and carrying method, such as both being aperiodic CSI reports carried via PUSCH, when an AI-based CSI report without L1-RSRP or L1-SINR conflicts with another CSI report without L1-RSRP or L1-SINR, regardless of Pri... iCSI Regardless of the value of (y,k,c,s), AI-based CSI reports without L1-RSRP or L1-SINR have higher priority, which helps with system stability; or CSI reports without L1-RSRP or L1-SINR have higher priority, which helps with system performance stability.
[0450] Under the same time-domain reporting behavior and bearing method, such as non-periodic CSI reports being carried through PUSCH, when at least two of the following conflict: performance monitoring results, intermediate performance monitoring data, or monitored performance indicator reports (represented by letter A) for CSI reports carrying L1-RSRP or L1-SINR (represented by letter B), and AI-based CSI reports carrying L1-RSRP or L1-SINR (represented by letter C), regardless of Pri... iCSIRegardless of the values of (y, k, c, s), the reporting priorities of the above-mentioned reported quantities are A≥B≥C, A≥C≥B, B≥A≥C, B≥C≥A, C≥B≥A, or C≥A≥B. If only two of A, B, and C conflict, the above priority rules still apply, and one of them can be disregarded. For example, if A and B conflict, and the priority satisfies A≥B≥C, then the priority of A is not lower than that of B. The different priorities of A, B, and C mentioned above all have certain beneficial effects. For example, A≥B≥C helps ensure continuous monitoring of AI characteristics or functions, ensuring the stability of the performance of AI models or AI-based characteristics or functions. Meanwhile, having B's priority no lower than C helps ensure system performance stability. Similarly, A≥C≥B also helps ensure continuous monitoring of AI characteristics or functions, ensuring the stability of the performance of AI models or AI-based characteristics or functions. Having C's priority no lower than B helps the system achieve better performance, because the same function usually yields more benefits from AI-based implementations than traditional implementations. Finally, B≥A≥C helps ensure system performance stability, while having A's priority no lower than B helps ensure continuous monitoring of AI characteristics or functions. Continuous monitoring of capabilities ensures the stability of the performance of AI models or AI-based features or functions. For B≥C≥A, this also helps ensure the stability of system performance, and the priority of C is no less than that of A. For C≥B≥A, this helps the system achieve better performance, and the priority of B is no less than that of A. For C≥A≥B, this helps the system achieve better performance, and the priority of A is no less than that of B. This helps ensure the continuous monitoring of AI features or functions, ensuring the stability of the performance of AI models or AI-based features or functions. Note that there may only be two of the above three reporting quantities (A, B, C) for different schemes. Even if there are only two, the above priority rules still apply, and one of them can be disregarded.
[0451] Under the same time-domain reporting behavior and bearing method, such as non-periodic CSI reports being carried through PUSCH, when performance monitoring results, intermediate performance monitoring quantities, or monitored performance indicator reports (represented by the letter D) for CSI reports without L1-RSRP or L1-SINR conflict with CSI reports without L1-RSRP or L1-SINR (represented by the letter E) and AI-based CSI reports without L1-RSRP or L1-SINR (represented by the letter F), regardless of Pri iCSIRegardless of the values of (y, k, c, s), the reporting priorities of the above-mentioned reported quantities are D≥E≥F, D≥F≥E, E≥D≥F, E≥F≥D, F≥E≥D, or F≥D≥E. If only two of D, E, and F conflict, the above priority rules still apply, and one of them can be disregarded. For example, if D and E conflict, and the priority satisfies D≥E≥F, then the priority of D is not lower than that of E. The aforementioned priority classifications of D, E, and F all have certain beneficial effects. For example, D≥E≥F helps ensure continuous monitoring of AI characteristics or functions, ensuring the stability of the performance of AI models or AI-based characteristics or functions. Furthermore, having E's priority no lower than F helps ensure system performance stability. Similarly, D≥F≥E also helps ensure continuous monitoring of AI characteristics or functions, ensuring the stability of the performance of AI models or AI-based characteristics or functions. Having F's priority no lower than E helps the system achieve better performance, as the same function typically yields more benefits from AI-based implementations than traditional implementations. Finally, E≥D≥F helps ensure system performance stability, while having D's priority no lower than E helps ensure continuous monitoring of AI characteristics or functions. Continuous monitoring of capabilities ensures the stability of the performance of AI models or AI-based features or functions. For E≥F≥D, this also helps ensure the stability of system performance, and the priority of F is no less than that of D. For F≥E≥D, this helps the system achieve better performance, and the priority of E is no less than that of D. For F≥D≥E, this helps the system achieve better performance, and the priority of D is no less than that of E. This helps ensure the continuous monitoring of AI features or functions, ensuring the stability of the performance of AI models or AI-based features or functions. Note that there may only be two of the three reporting quantities A, B, and C for different schemes. Even if there are only two, the above priority rules still apply, and one of them can be disregarded.
[0452] In the above scheme, the CSI report associated with the performance test results, intermediate quantities of performance monitoring, or monitored performance indicator reports refers to the CSI report to which the performance test results, intermediate quantities of performance monitoring, or monitored performance indicator reports are targeted. For example, the CSI report associated with the performance test results, intermediate quantities of performance monitoring, or monitored performance indicator reports based on AI-based CSI predictions is a CSI report based on AI-based CSI predictions. In the above scheme, the performance monitoring results, intermediate quantities of performance monitoring, or monitored performance indicators can be the result of comparing the actual or real measured quantities with the AI-based inference output quantities according to certain indicators. For example, the comparison indicators can be Normalized Mean Squared Error (NMSE) or Squared Cosine Similarity (SGCS).
[0453] This embodiment introduces an AI-based CSI reporting mechanism, along with a priority mechanism for reporting performance monitoring results, intermediate performance metrics, or monitored performance indicators, on top of the existing CSI reporting priority mechanism. This ensures that in the communication system, AI-based and non-AI-based CSI reports, as well as performance monitoring results, intermediate performance metrics, or monitored performance indicators, can be reported according to a predetermined priority. Specifically, the UE flexibly determines the reporting priority based on whether the CSI report carries L1-RSRP or L1-SINR, and whether it belongs to AI-related reports. The base station receives the UE's CSI reports, thereby ensuring that important non-AI-related reports, AI-related reports, and / or performance monitoring results, intermediate performance metrics, or monitored performance indicators are transmitted with priority. This mechanism not only improves the system's resource utilization efficiency but also ensures the stability of AI models and the reliability of communication services, making it particularly suitable for complex communication scenarios with multiple tasks and configurations.
[0454] Eleventh Example: Differentiation of content reported by the terminal.
[0455] In some embodiments of this application, to distinguish the purpose of CSI reporting configuration and ensure that the UE and base station have a correct and consistent understanding of the CSI reporting configuration information, specific fields are configured to align the understanding between the UE and the base station. Specifically, the UE distinguishes how the CSI reporting configuration is used based on specific fields in the CSI reporting configuration, and then reports the CSI accordingly. Based on this, the base station receives the CSI, thereby ensuring the normal operation of the system and ensuring system stability and communication performance.
[0456] Specifically, based on current standard discussions, when using terminal-side AI models for CSI prediction, a traditional codebook-type "typeII-Doppler-r18" feedback mechanism can be used as a starting point. However, from the perspective of the reported content alone, the network cannot distinguish whether the terminal's CSI feedback is obtained based on traditional prediction methods, such as "typeII-Doppler-r18" codebook, or based on AI model prediction. Furthermore, when using terminal-side AI models for different CSI predictions, the results output by the AI model during model training may not need to be reported to the network side. Therefore, the terminal side needs to align with the network side to achieve the desired model output. Additionally, model training, inference, and monitoring may use the same CSI-RS resource configuration; therefore, the network side needs to differentiate what specific content is reported at different stages of the model. This embodiment provides at least one solution for this:
[0457] Option 1: Reuse the existing standard CSI-ReportConfig field, but add a new field under this field to distinguish whether the configuration is used for AI-based CSI prediction, and to distinguish which stage of AI-based CSI prediction the configuration is used for, such as model training, model inference, and / or model monitoring; the specific implementation can take at least one of the following forms:
[0458] Method 1: Add a new indicator field. The length of this field is 1 bit. If this bit is 1 or 0, it means that the reporting configuration is used for model training; if this bit is 0 or 1, it means that the reporting configuration is used for model inference.
[0459] Method 2: Add a new indicator field with a length of 2 bits. If the field is X∈{00,01,10,11}, it indicates that the configuration is used for model training. If the field is Y, and Y is any value other than X, it indicates that the configuration is used for model inference reporting. For example, if X=00, then Y∈{01,10,11}. If the field is Z, and Z is any value other than X and Y, it indicates that the configuration is used for model or function monitoring reporting.
[0460] Method 3: Add a new indicator field. This field is of type enumeration and includes two enumeration types: {inference, training}. If this field is configured as inference, it means that the reporting configuration is used for model inference; if this field is configured as training, it means that the reporting configuration is used for model training.
[0461] Method 4: Add a new indicator field. This field is of type enumeration and includes three enumeration types: {inference, training, monitoring}. If the field is configured as inference, it means that the reporting configuration is used for model inference; if the field is configured as training, it means that the reporting configuration is used for model training; if the field is configured as monitoring, it means that the reporting configuration is used for model or function monitoring and reporting.
[0462] This embodiment distinguishes the purpose of different CSI reporting configurations by introducing specific fields in the CSI reporting configuration, thereby aligning the understanding between the base station and the UE. This ensures the normal operation of the system and guarantees system stability and communication performance.
[0463] Figure 8 is a schematic structural diagram of a wireless communication device 700 provided in an embodiment of this application. This wireless communication device can be a user equipment, a base station, or a network element. The wireless communication device 700 shown in Figure 8 includes a processor 710, which can call and run computer programs from a memory to implement the methods in the embodiments of this application.
[0464] Optionally, as shown in FIG8, the wireless communication device 700 may further include a memory 720. The processor 710 can call and run computer programs from the memory 720 to implement the methods in the embodiments of this application. The memory 720 may be a separate device independent of the processor 710, or it may be integrated into the processor 710.
[0465] Optionally, as shown in Figure 8, the wireless communication device 700 may further include a transceiver 730. The processor 710 can control the transceiver 730 to communicate with other devices. Specifically, it can send information or data to other devices or receive information or data sent by other devices. The transceiver 730 may include a transmitter and a receiver. The transceiver 730 may further include an antenna, and the number of antennas may be one or more.
[0466] Optionally, the wireless communication device 700 may specifically be the network-side device 110 in the embodiments of this application, and the wireless communication device 700 may implement the corresponding processes implemented by the network-side device 110 in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.
[0467] Optionally, the wireless communication device 700 may specifically be a user equipment in the embodiments of this application, and the wireless communication device 700 may implement the corresponding processes implemented by the user equipment in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.
[0468] Optionally, the wireless communication device 700 may specifically be a network element in the embodiments of this application, and the wireless communication device 700 may implement the corresponding processes implemented by the network element in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.
[0469] Figure 9 is a schematic structural diagram of a chip according to an embodiment of this application. The chip 800 shown in Figure 9 includes a processor 810, which can call and run computer programs from memory to implement the methods in the embodiments of this application.
[0470] Optionally, as shown in FIG9, chip 800 may further include memory 820. Processor 810 can call and run computer programs from memory 820 to implement the methods in the embodiments of this application. Memory 820 may be a separate device independent of processor 810, or it may be integrated into processor 810.
[0471] Optionally, the chip 800 may also include an input interface 830. The processor 910 can control the input interface 830 to communicate with other devices or chips; specifically, it can acquire information or data sent by other devices or chips.
[0472] Optionally, the chip 800 may also include an output interface 840. The processor 810 can control the output interface 840 to communicate with other devices or chips, specifically, to output information or data to other devices or chips.
[0473] Optionally, the chip can be applied to the network-side device 110 in the embodiments of this application, and the chip can implement the corresponding processes implemented by the network-side device 110 in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.
[0474] Optionally, the chip can be applied to the user equipment in the embodiments of this application, and the chip can implement the corresponding processes implemented by the user equipment in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.
[0475] Optionally, the chip can be applied to the network element in the embodiments of this application, and the chip can implement the corresponding processes implemented by the mobile network element in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.
[0476] Figure 10 is a schematic block diagram of a wireless communication system 900 provided in an embodiment of this application. As shown in Figure 10, the communication system 900 includes a user equipment 920 and a network-side device 910. The user equipment 920 can be used to implement the corresponding functions implemented by the user equipment 920 in the above method, and the network-side device 910 can be used to implement the corresponding functions implemented by the network-side device 910 in the above method. For simplicity, these will not be described in detail here.
[0477] It should be understood that the processor in the embodiments of this application may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method embodiments can be completed by integrated logic circuits in the processor's hardware or by software instructions.
[0478] It is understood that the memory in the embodiments of this application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory. Embodiments of this application also provide a computer-readable storage medium for storing a computer program.
[0479] Optionally, the computer-readable storage medium can be applied to the network device in the embodiments of this application, and the computer program causes the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of this application. For simplicity, further details are omitted here. Optionally, the computer-readable storage medium can be applied to the user equipment in the embodiments of this application, and the computer program causes the computer to execute the corresponding processes implemented by the user equipment in the various methods of the embodiments of this application. For simplicity, further details are omitted here.
[0480] This application also provides a computer program product, including computer program instructions.
[0481] Optionally, the computer program product can be applied to the network device in the embodiments of this application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of this application. For simplicity, further details are omitted here. Optionally, the computer program product can be applied to the user equipment in the embodiments of this application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the user equipment in the various methods of the embodiments of this application. For simplicity, further details are omitted here.
[0482] This application also provides a computer program. Optionally, the computer program can be applied to the network device in this application embodiment. When the computer program runs on a computer, it causes the computer to execute the corresponding processes implemented by the network device in the various methods of this application embodiment. For simplicity, these will not be described in detail here. Optionally, the computer program can be applied to the user device in this application embodiment. When the computer program runs on a computer, it causes the computer to execute the corresponding processes implemented by the user device in the various methods of this application embodiment. For simplicity, these will not be described in detail here. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0483] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A communication method, executed by a user equipment, comprising: Report the capability information of the user equipment to the base station; The system receives Channel State Information (CSI) report configuration and / or CSI resource configuration sent by the base station. The CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signals. The CSI report configuration includes a configuration for merging CSI reports corresponding to multiple tasks into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports.
2. The communication method according to claim 1, wherein, The configuration for merging the CSI reports corresponding to the multiple tasks into a single CSI report can be achieved through default merging, merging based on uplink control information (UCI) format, merging based on activation, and / or merging based on configuration.
3. The communication method according to claim 2, wherein, In the absence of signaling configuration, if multiple AI-related CSI reports share the same CSI resource configuration identifier ID, synchronization signal block SSB group index, or CSI resource configuration ID, then the default merging is adopted; the merging based on UCI format means that when the configured UCI format is X and X≥4, the merged transmission of K AI-related CSI reports is triggered, where K is a predefined value or configured through Radio Resource Control (RRC) messages, Media Access Control (MAC-CE) messages, or Downlink Control Information (DCI) messages; The activation-based merging means activating the merging of K AI-related CSI reports through a 1-bit field in the DCI or RRC message, and indicating the merged report in the merged report through a 1-bit field or information containing the CSI report configuration ID; The configuration-based merging means configuring the CSI report group for merging through the RRC message, and indicating the K AI-related CSI reports to be merged through the CSI report configuration ID.
4. The communication method according to claim 2, wherein In a single CSI report, a report header indicates the order in which CSI reports are merged, or each report segment carries the corresponding CSI report ID.
5. The communication method according to claim 2, wherein, The configuration of the merged CSI report is determined by the base station through preset, priority selection, or reconfiguration.
6. The communication method according to claim 1, further comprising: The system receives CSI report triggering information sent by the base station via DCI and / or MAC-CE messages, wherein the CSI report triggering information includes a message for triggering CSI report merging and / or configuration selection information for CSI report merging.
7. The communication method according to claim 6 further includes: Determine the number of CSI reports that the user equipment can process, and generate and send CSI reports according to the processing results.
8. The communication method according to claim 7, wherein The CSI report transmission time is configured by the base station or determined by the user equipment based on the CSI report processing time corresponding to the AI function.
9. The communication method according to claim 7, wherein If the CSI report is a combined report, the CSI report includes a packet header indicating related information of the combined multiple CSI reports, the related information including a CSI report ID and / or an arrangement order; and / or the user equipment reports CSI report quantities corresponding to multiple different AI tasks in an order configured by the base station or indicated in the packet header.
10. A communication method, performed by a user equipment, comprising: reporting, to a base station, capability information of the user equipment; receiving, from the base station, a channel state information (CSI) report configuration and / or a CSI resource configuration; determining a number of CSI reports that the user equipment can process, and generating and sending a CSI report according to a processing result.
11. The communication method according to claim 10, wherein The capability information of the user equipment includes an indication of a number N of available AI processing units of the user equipment and / or a number O of AI processing units required by the user equipment for CSI reports corresponding to multiple use cases, function sets, models, or CSI report types, the O being mapped to each use case, function set, model, or CSI report type.
12. The communication method according to claim 10, wherein, The capability information of the user equipment includes the number N of available CSI processing units (CPUs) of the user equipment. CPU And / or the number of CPUs required by the user equipment for CSI reports corresponding to multiple use cases, feature sets, models, or CSI report types. CPU The instructions include: AI-related O CPU Map to each use case, feature set, model, or CSI report type, and / or indicate the allocation of CPU resources between AI-related CSI reports and non-AI-related CSI reports.
13. The communication method according to claim 10, wherein, The capability information of the user equipment includes the number N of available AI processing units of the user equipment, the number O of AI processing units required by the user equipment for CSI reports corresponding to multiple use cases, function sets, models, or CSI report types, and the number N of available CPUs of the user equipment. CPU And / or the number of CPUs required by the user equipment for the CSI reports corresponding to the multiple use cases, feature sets, models, or CSI report types. CPU The instructions include: AI-related O CPU Map to each use case, feature set, model, or CSI report type, and / or indicate the allocation of CPU resources between AI-related CSI reports and non-AI-related CSI reports.
14. The communication method according to claim 13, wherein, The O is determined by a predefined mapping relationship, the predefined mapping relationship being determined based on a type of the user equipment, the capability information, and / or a configured use case, function set, or model.
15. The communication method according to claim 13, wherein, The user equipment reports the number of CSI reports corresponding to the function group, model or use case supported by the user equipment to the base station when reporting the capability information of the user equipment to the base station CPU value.
16. The communication method according to claim 13, wherein The user equipment reports a set of supported function groups, models, or use cases of the user equipment when reporting the capability information of the user equipment to the base station or in an uplink message. By a predefined mapping relationship, the AI-related CSI report corresponds to the O CPU value.
17. The communication method according to claim 10, wherein, A sending time of the CSI report is configured by the base station or determined by the user equipment according to a CSI report processing time corresponding to an AI function.
18. The communication method according to claim 10, wherein, The CSI report includes an AI-related CSI report and / or a non-AI-related CSI report, the AI-related CSI report being calculated using an AI processing unit, the non-AI-related CSI report being calculated using a CPU, and an AI part and a non-AI part of the AI-related CSI report both being calculated using the AI processing unit.
19. The communication method according to claim 10, wherein, The occupation time of the CPU is determined in at least one of the following ways: based on a configuration of a CSI report quantity, in combination with a position of a selected measurement reference signal resource and a CSI calculation time corresponding to a current AI task; based on the configuration of the CSI report quantity, in combination with the position of the selected measurement reference signal resource, a sending time of the CSI report, and the CSI calculation time corresponding to the current AI task; based on the configuration of the CSI report quantity, in combination with a time of receiving the CSI report triggering information, the sending time of the CSI report, and the CSI calculation time corresponding to the current AI task.
20. The communication method according to claim 10, wherein, When performing CSI report calculation, the user equipment determines an occupation start and end time of the AI processing unit based on pre-configuration, capability reporting, or information configured by the base station, according to whether CSI needs to be reported and a number of tasks, including a continuous or intermittent time period in a symbol interval after a reference signal resource or a physical downlink control channel (PDCCH) trigger.
21. The communication method according to claim 10, wherein The CSI report includes AI-related CSI reports and / or non-AI-related CSI reports. The AI-related CSI reports occupy a first part of the CPU, and the non-AI-related CSI reports occupy a second part of the CPU. The AI portion of the AI-related CSI reports occupies a first sub-part of the first part of the CPU, and the non-AI portion of the AI-related CSI reports occupies a second sub-part of the first part of the CPU.
22. The communication method according to claim 21, wherein The AI-related CSI report occupies the first part of the CPU, and the non-AI-related CSI report occupies the second part of the CPU.
23. The communication method according to claim 21, wherein The non-AI portion of the AI-related CSI report occupies the second sub-part of the first portion of the CPU, and the second portion of the CPU occupied by the non-AI-related CSI report is shared.
24. The communication method according to claim 21, wherein The user equipment determines the number of CSI reports it can process by allocating CPU resources between the AI-related CSI reports and the non-AI-related CSI reports.
25. The communication method according to claim 24, wherein, The user equipment allocates CPU resources between the AI-related CSI reports and the non-AI-related CSI reports by at least one of the following: The AI-related CSI reports and the non-AI-related CSI reports share the CPU resources, which are dynamically allocated between the AI-related CSI reports and the non-AI-related CSI reports. The AI-related CSI reports and the non-AI-related CSI reports share the CPU resources, and the CPU resources are dynamically allocated between the non-AI-related user device behaviors corresponding to the AI-related CSI reports and the non-AI-related CSI reports. The AI-related user device behavior corresponding to the AI-related CSI report occupies the AI processing unit resources. The non-AI-related user device behavior corresponding to the AI-related CSI report and the non-AI-related CSI report jointly occupy the CPU resources. The CPU resources are dynamically allocated between the non-AI-related user device behavior corresponding to the AI-related CSI report and the non-AI-related CSI report.
26. The communication method of claim 21, wherein, The reporting priority of the CSI report depends on at least one of the following parameters: reporting configuration ID, serving cell index, reporting volume, reporting configuration type, user equipment behavior corresponding to the report, and CPU resource usage of the corresponding report.
27. The communication method of claim 21, wherein, The AI-related CSI report occupies the first part of the CPU and the non-AI-related CSI report occupies the second part of the CPU, which are used in a non-shared manner.
28. The communication method of claim 21, wherein, When the user equipment selects an AI-related CSI report, based on the available resources of the CPU and AI processing unit, at least one of the following conditions must be met: The total AI processing unit usage of the AI-related CSI report shall not exceed the remaining resources of the AI processing unit of the user equipment. The total CPU usage of the AI-related CSI reports and the non-AI-related CSI reports does not exceed the remaining CPU resources of the user equipment; When the AI-related CSI report and the non-AI-related CSI report use independent CPU resource pools, their total usage shall not exceed the remaining resources of the corresponding CPU resource pool.
29. The communication method of claim 21, wherein, The CPU usage corresponding to the AI-related CSI report is defined by predefined parameters, the base station configuration, and / or parameters in the capability information of the user equipment.
30. The communication method of claim 21, wherein, The CPU usage time of the AI-related CSI reports corresponds to the CSI report processing time, which is determined by a lookup table.
31. The communication method according to claim 10, wherein The CSI report includes AI-related CSI reports and / or non-AI-related CSI reports. The AI-related CSI reports occupy a portion of the AI processing unit and the CPU, while the non-AI-related CSI reports occupy the CPU. The AI portion of the AI-related CSI reports is calculated using the AI processing unit, and the non-AI portion of the AI-related CSI reports is calculated using the CPU.
32. The communication method according to claim 31, wherein The non-AI portion of the AI-related CSI report and the non-AI-related CSI report together occupy the CPU.
33. The communication method according to claim 31, wherein The AI-related CSI report occupies a first portion of the CPU, and the non-AI-related CSI report occupies a second portion of the CPU.
34. The communication method of claim 31, wherein, The user equipment determines the number of CSI reports it can process through a combination of serial and parallel processing methods, including at least one of the following: Within the observation window or sub-observation window of CSI measurement, multiple processing windows are divided, and within each processing window, the CSI report processing of a task group is executed in parallel or serially. The sum of CSI reports from multiple task groups processed within the multiple processing windows constitutes the maximum number of CSI reports that the user equipment can process within the corresponding observation window or sub-observation window.
35. The communication method of claim 31, wherein, The AI processing unit usage time and / or CPU usage time in the CSI report are determined by at least one of the following methods: The CPU usage time is a piecewise function, and the AI processing unit usage time is a continuous function. The AI processing unit occupancy time includes at least one of measurement data collection and inference, and the CPU occupancy time includes at least one of physical uplink control channel PUCCH or physical uplink shared channel PUSCH generation; The time occupied by the AI processing unit includes inference processing, and the CPU time occupied includes at least one of the measurement data collection and PUCCH or PUSCH generation.
36. The communication method according to claim 10, further comprising: Receive CSI report trigger information sent by the base station through downlink control information (DCI) and / or media access control unit (MAC-CE) messages.
37. A communication method, executed by a user equipment, comprising: Determine the reporting priority of multiple Channel Status Information (CSI) reports; The CSI report is reported to the base station according to the reporting priority of the CSI report; The CSI report includes AI-related CSI reports and non-AI-related CSI reports. The user equipment determines the reporting priority of the CSI report based on whether the CSI report carries the physical layer reference signal received power L1-RSRP or physical layer signal interference-noise ratio L1-SINR. The CSI report is a performance monitoring result or intermediate quantity of whether the AI-related CSI report carries the L1-RSRP or L1-SINR, and / or based on the reporting time domain behavior and bearer type of the CSI report.
38. The communication method according to claim 37, wherein the reporting priority of the CSI report satisfies at least one of the following rules: when a performance monitoring result or intermediate quantity report of a CSI report carrying the L1-RSRP or the L1-SINR conflicts with a CSI report carrying the L1-RSRP or the L1-SINR, the performance monitoring result or the intermediate quantity report has a higher priority; when a performance monitoring result or intermediate quantity report of a CSI report not carrying the L1-RSRP or the L1-SINR conflicts with a CSI report not carrying the L1-RSRP or the L1-SINR, the performance monitoring result or the intermediate quantity report has a higher priority. When the AI-related CSI report conflicts with the non-AI-related CSI report, the AI-related CSI report has higher priority; when the AI-related CSI report carrying the L1-RSRP or the L1-SINR conflicts with the non-AI-related CSI report carrying the L1-RSRP or the L1-SINR, the AI-related CSI report has higher priority.
39. A communication method, executed by a base station, comprising: Receive the capability information of the user equipment reported by the user equipment; Send Channel State Information (CSI) report configuration and / or CSI resource configuration to the user equipment, wherein the CSI report configuration indicates that multiple CSI reports are associated with the same or partially the same set of measurement reference signals, the CSI report configuration includes a configuration for merging CSI reports corresponding to multiple tasks into one CSI report, and / or the CSI resource configuration includes a configuration for associating a set of CSI resources with multiple CSI reports.
40. The communication method of claim 39, wherein, The configuration for merging the CSI reports corresponding to the multiple tasks into a single CSI report can be achieved through default merging, merging based on uplink control information (UCI) format, merging based on activation, and / or merging based on configuration.
41. The communication method of claim 39, wherein, In a single CSI report, a report header indicates the order in which CSI reports are merged, or each report segment carries the corresponding CSI report ID.
42. The communication method according to claim 39, further comprising: The system receives CSI report triggering information sent by the base station via DCI and / or MAC-CE messages, wherein the CSI report triggering information includes a message for triggering CSI report merging and / or configuration selection information for CSI report merging.
43. The communication method according to claim 42, further comprising: Determine the number of CSI reports that the user equipment can process, and generate and send CSI reports according to the processing results.
44. A communication method, executed by a base station, comprising: Receive the capability information of the user equipment reported by the user equipment; Send Channel State Information (CSI) report configuration and / or CSI resource configuration to the user equipment; Receive CSI reports reported by the user equipment, wherein the CSI reports are based on the number of CSI reports that the user equipment can process.
45. The communication method of claim 44, wherein, The capability information of the user equipment includes the number N of available AI processing units of the user equipment, the number O of AI processing units required by the user equipment for CSI reports corresponding to multiple use cases, function sets, models, or CSI report types, and the number N of available CPUs of the user equipment. CPU And / or the number of CPUs required by the user equipment for the CSI reports corresponding to the multiple use cases, feature sets, models, or CSI report types. CPU The instructions include: AI-related O CPU Map to each use case, feature set, model, or CSI report type, and / or indicate the allocation of CPU resources between AI-related CSI reports and non-AI-related CSI reports.
46. The communication method of claim 44, wherein, The CPU usage time is determined by at least one of the following methods: The configuration is determined based on the CSI reporting volume, combined with the location of the selected measurement reference signal resources and the CSI calculation time corresponding to the current AI task; The determination is based on the configuration of the CSI reporting volume, combined with the location of the selected measurement reference signal resource, the sending time of the CSI report, and the CSI calculation time corresponding to the current AI task; The determination is based on the configuration of the CSI reporting volume, combined with the time of receiving the CSI report trigger information, the time of sending the CSI report, and the CSI calculation time corresponding to the current AI task.
47. The communication method of claim 44, wherein, The CSI reports include AI-related CSI reports and / or non-AI-related CSI reports. The user equipment determines the number of CSI reports it can process by allocating CPU resources between the AI-related CSI reports and the non-AI-related CSI reports. The allocation of CPU resources between the AI-related CSI reports and the non-AI-related CSI reports by the user equipment includes at least one of the following: The AI-related CSI reports and the non-AI-related CSI reports share the CPU resources, which are dynamically allocated between the AI-related CSI reports and the non-AI-related CSI reports. The AI-related CSI reports and the non-AI-related CSI reports share the CPU resources, and the CPU resources are dynamically allocated between the non-AI-related user device behaviors corresponding to the AI-related CSI reports and the non-AI-related CSI reports. The AI-related user device behavior corresponding to the AI-related CSI report occupies the AI processing unit resources. The non-AI-related user device behavior corresponding to the AI-related CSI report and the non-AI-related CSI report jointly occupy the CPU resources. The CPU resources are dynamically allocated between the non-AI-related user device behavior corresponding to the AI-related CSI report and the non-AI-related CSI report.
48. The communication method of claim 44, wherein, The CSI report includes AI-related CSI reports and / or non-AI-related CSI reports. The AI-related CSI reports occupy a portion of the AI processing unit and the CPU, while the non-AI-related CSI reports occupy the CPU. The AI portion of the AI-related CSI reports is calculated using the AI processing unit, and the non-AI portion of the AI-related CSI reports is calculated using the CPU.
49. The communication method according to claim 44, further comprising: The CSI report trigger information is sent to the user equipment via Downlink Control Information (DCI) and / or Media Access Control Unit (MAC-CE) messages.
50. A communication device comprising: A processor and a memory, the memory for storing a computer program, the processor for calling and running the computer program stored in the memory to perform the communication method as described in any one of claims 1 to 49.
