Wireless communication method of AI / ML, UE and network equipment

Abstract

The invention provides a wireless communication method of artificial intelligence (AI) / machine learning (ML), and the wireless communication method comprises the following steps: configuring a specific reference signal configuration of an AI / ML model for user equipment (UE), and transmitting the specific reference signal configuration of the AI / ML model to the UE according to the specific reference signal configuration of the AI / ML model. The AI / ML model-specific reference signal configuration includes an indication to indicate a reference signal usage of an AI / ML model-specific reference signal, the reference signal usage including at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wireless communication systems, and more particularly, to a wireless communication method, a user equipment (UE) and a network device of artificial intelligence (AI) / machine learning (ML). For example, the present application is related to AI / ML of a new radio (NR) air interface discussed in Release 18 (Rel. 18), the main idea of which is mainly focused on RAN1 topics, including uplink and downlink measurement configuration (such as reference signal configuration: channel state information-reference signal (CSI-RS), sounding reference signal (SRS), demodulation reference signal (DMRS), etc.), and AI / ML-based measurement and reporting to support AI / ML models to work properly in the network with little signaling interaction between gNB and UE. BACKGROUND

[0002] As of the 111th meeting of 3GPP RAN1, many topics related to wireless AI / ML have been discussed, including use cases (CSI compression / prediction / measurement, positioning, beam management, radio resource management RRM, etc.), system-level and link-level simulation, life cycle management, etc., and some key conclusions and consensus have been reached, especially the AI / ML life cycle management framework, which is the basis for wireless AI / ML standardization. Even though a general framework has been proposed, how to define a general process that can meet the needs of different use cases is still a great challenge. Future research is needed to support AI / ML model-specific reference signal configuration and data collection (such as CSI, positioning, beam management, inference, training, and monitoring, etc.), and there is currently no existing technology to solve this problem. SUMMARY

[0003] The present application aims to propose a wireless communication method, a UE and a network device of AI / ML to solve the problems in the prior art, reduce signaling overhead, provide scalability for a certain type of AI / ML model to meet the needs of multiple models, and unify the processing process of unilateral or bilateral models through a single AI / ML model-specific process.

[0004] In a first aspect of the application, a method of AI / ML wireless communication comprises configuring, to a UE, an AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration comprises an indication for indicating a reference signal usage, the usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario.

[0005] In a second aspect of the application, a method of AI / ML wireless communication comprises determining, by a UE, an AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration comprises an indication for indicating an AI / ML model-specific reference signal usage, the reference signal usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario.

[0006] In a third aspect of the application, a wireless communication device comprises a configurator configured to configure, to a UE, an AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration comprises an indication for indicating an AI / ML model-specific reference signal usage, the reference signal usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario.

[0007] In a fourth aspect of the application, a wireless communication device comprises a determinator configured to determine an AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration comprises an indication for indicating an AI / ML model-specific reference signal usage, the reference signal usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario.

[0008] In some embodiments, the indication for indicating reference signal usage includes at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model process; an identifier of a reference signal; a string-based indication; an ID-based indication; a bitmap-based indication; a hierarchical method-based indication; and / or a sequential-based indication. In some embodiments, with the sequential-based indication, the usage of AI / ML model specific reference signals can be indicated in different configuration sequences. In some embodiments, the AI / ML model specific reference signals include at least one of: a channel state information reference signal (CSI-RS), a demodulation reference signal (DMRS), a sounding reference signal (SRS), a phase tracking reference signal (PTRS), and / or a positioning reference signal (PRS). In some embodiments, the method further includes receiving, from the UE, an AI / ML model specific reporting information configuration, wherein the AI / ML model specific reporting information configuration includes an indication for indicating AI / ML model information. In some embodiments, the indication for indicating AI / ML model information includes at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model process; an identifier of an AI / ML scenario; and / or an identifier of a reference signal.

[0009] In some embodiments, the AI / ML model specific reference signal configuration and / or the AI / ML model specific reporting information configuration is activated and / or deactivated by trigger information. In some embodiments, the trigger information includes at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model process; an identifier of a reference signal; an identifier of a reporting quantity; an activation timing; and / or a deactivation timing. In some embodiments, the method further includes configuring a report to the UE, wherein the report is configured to report measurement information between the network device and the UE based on the AI / ML model specific reference signal indication.

[0010] In some embodiments, the method further includes configuring, to the UE, AI / ML model specific channel state information. In some embodiments, the AI / ML model specific channel state information includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model procedure; an identifier of an AI / ML scenario; and / or an identifier of a reference signal.

[0011] In some embodiments, the indication to indicate reference signal usage includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model procedure; an identifier of an AI / ML scenario; an identifier of a reference signal; a string based indication; an ID based indication; a bitmap based indication; a hierarchical method based indication; and / or a sequential based indication. In some embodiments, by the sequential based indication, the usage of AI / ML model specific reference signals is indicated in different configuration order. In some embodiments, the AI / ML model specific reference signals include at least one of: a CSI-RS, a DMRS, a SRS, a PTRS, and / or a PRS. In some embodiments, the method further includes sending or transmitting, from the network device, an AI / ML model specific reporting information configuration, wherein the AI / ML model specific reporting information configuration includes an indication to indicate AI / ML model information. In some embodiments, the indication to indicate AI / ML model information includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model procedure; an identifier of an AI / ML scenario; and / or an identifier of a reference signal. In some embodiments, the AI / ML model specific reference signal configuration and / or the AI / ML model specific reporting information configuration is activated and / or deactivated by trigger information.

[0012] In some embodiments, the trigger information includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model procedure; an identifier of a reference signal; an identifier of a reporting quantity; an activation timing; and / or a deactivation timing. In some embodiments, the method further includes determining, by the UE, a report, wherein the report is configured to indicate, based on AI / ML model specific reference signal, measurement information between the network device and the UE. In some embodiments, the method further includes determining, by the UE, AI / ML model specific channel state information. In some embodiments, the AI / ML model specific channel state information includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model procedure; an identifier of an AI / ML scenario; and / or an identifier of a reference signal.

[0013] In a fifth aspect of the present application, a network device comprises a memory, a transceiver, and a processor coupled with the memory and the transceiver. The processor is configured to perform the above method.

[0014] In a sixth aspect of the present application, a UE comprises a memory, a transceiver, and a processor coupled with the memory and the transceiver. The processor is configured to perform the above method.

[0015] In a seventh aspect of the present application, a non-transitory machine-readable storage medium has stored thereon instructions, which when executed by a computer, cause the computer to perform the above method.

[0016] In an eighth aspect of the present application, a chip comprises a processor configured to invoke and run a computer program stored in a memory, so that a device installed with the chip performs the above method.

[0017] In a ninth aspect of the present application, a computer-readable storage medium has stored thereon a computer program, which causes a computer to perform the above method.

[0018] In a tenth aspect of the present application, a computer program product comprises a computer program, which causes a computer to perform the above method.

[0019] In an eleventh aspect of the present application, a computer program causes a computer to perform the above method. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or related art, the following briefly introduces the drawings to be described in the embodiments. Obviously, the drawings are only some embodiments of the present application, and a person of ordinary skill in the art can obtain other drawings from these drawings without paying creative labor.

[0021] Figure 1 An example schematic diagram of an AI / ML general framework in wireless communication is shown.

[0022] Figure 2 An example schematic diagram of a basic autoencoder model for enhancing CSI feedback according to an embodiment of the present application is shown.

[0023] Figure 3 A block diagram of a communication node in a communication network system according to an embodiment of the present application is shown.

[0024] Figure 4A A flowchart of an AI / ML wireless communication method according to an embodiment of the present application is shown.

[0025] Figure 4BA flowchart of an AI / ML wireless communication method according to embodiments of the application is shown.

[0026] Figure 5A A flowchart of an AI / ML wireless communication method performed by a network device according to embodiments of the application is shown.

[0027] Figure 5B A flowchart of an AI / ML wireless communication method performed by a network device according to embodiments of the application is shown.

[0028] Figure 6A A flowchart of an AI / ML wireless communication method performed by a UE according to embodiments of the application is shown.

[0029] Figure 7A A flowchart of an AI / ML wireless communication method performed by a UE according to embodiments of the application is shown.

[0030] Figure 8 A flowchart of downlink AI / ML model-specific resource configuration and feedback according to embodiments of the application is shown.

[0031] Figure 9 A flowchart of uplink AI / ML model-specific resource configuration and feedback according to embodiments of the application is shown.

[0032] Figure 10 A diagram of AI / ML model-specific activation / deactivation timing according to embodiments of the application is shown.

[0033] Figure 11 A diagram of AI / ML model-specific uplink feedback according to embodiments of the application is shown.

[0034] Figure 12A A block diagram of a communication device (e.g., a UE) according to embodiments of the application is shown.

[0035] Figure 12B A block diagram of a communication device (e.g., a network device) according to embodiments of the application is shown.

[0036] Figure 13A A block diagram of a communication device (e.g., a UE) according to embodiments of the application is shown.

[0037] Figure 13B A block diagram of a communication device (e.g., a network device) according to embodiments of the application is shown.

[0038] Figure 14 A block diagram of a wireless communication system according to embodiments of the application is shown. DETAILED DESCRIPTION

[0039] The technical content, structural features, implementation goals and effects of the embodiments of the present application will be described in detail with reference to the drawings. It should be particularly noted that the terms in the embodiments of the present application are only used for the purpose of describing specific embodiments and should not be regarded as limiting the present application.

[0040] As Figure 1 shown, data collection, model inference and feedback play a key role in the system, and these processes are specific to AI / ML models, and the relevant information between different models varies, as shown in Table 1. For example, for UE-based AI / ML positioning methods, channel measurement data such as power delay profile (PDP), channel impulse response (CIR), channel frequency response (CFR) or post-processed CIR, and the real coordinates of the UE are required. But for beam management, the following information is required, including the reporting behavior of the new reference signal received power (RSRP) and / or synchronization signal block (SSB) resource indicator (SSBRI) / CSI reference signal (CSI-RS) resource indicator (CRI). For example, a larger number of RSRPs need to be reported to generate labels and AI / ML inputs, or a larger number of beam IDs need to be reported as AI / ML outputs, rather than only reporting the best RSRP in the traditional mode. In addition, when set B is a subset of set A, it may be necessary to align the mapping relationship between set B and set A, such as which subset in set A is configured as set B. At the same time, for CSI prediction, its requirements are completely different from positioning and beam management, including real CSI of real DL / UL channel, original channel matrix or precoding matrix, etc. In order to obtain auxiliary information from measurement or feedback, dedicated reference signals are required. For example, the design of the reference signal can be enhanced to perform AI / ML-specific RSRP measurement, and the reference signal can be improved to more accurately measure data samples, which is required for AI / ML-based beam management. For CSI prediction, enhanced CSI-RS can be considered, especially for the data collection process, to generate a dataset with more accurate real CSI as samples. For example, by setting a higher power for the CSI-RS, or allocating more REs to the CSI-RS in the time / frequency domain for data collection, the UE can obtain more accurate downlink measurement channels as real CSI labels. Similarly, for positioning, dedicated reference signals are also required to improve accuracy to support AI / ML model inference or monitoring.

[0041] Table 1: AI model-specific reference signals and data requirements.

[0042]

[0043] In Table 1, for example, the AI / ML model functions include positioning, beam and CSI feedback enhancement. In positioning, the specific reference signal requirements include positioning-specific reference signals, higher density, wider beams, etc.; the data requirements include positioning-specific data, such as channel measurement data, e.g., PDP, CIR, CFR or post-processed CIR, and real coordinates of the UE.

[0044] In beam management, the specific reference signal requirements include BM-specific reference signals, higher power, higher density, different spatial filtering, etc.; the data requirements include BM-specific data, such as new RSRP and / or SSBRI / CRI reporting behavior. For example, a larger number of RSRPs need to be reported to generate labels and AI / ML inputs, or a larger number of beam IDs are reported as AI / ML outputs, etc.

[0045] In CSI feedback enhancement, the specific reference signal requirements include CSI enhancement-specific reference signals, higher density, and measurement gaps, etc.; the data requirements include CSI enhancement-specific data, such as real CSI of real DL / UL channels, original channel matrix or precoding matrix, etc.

[0046] As can be seen from Table 1, the specific data required (for feedback, training, inference input, etc.) is highly related to the AI / ML function and / or reference signal, and it is very important to unify these configuration and data collection processes to support AI / ML model functions through effective signaling methods. At the same time, it can be seen that, as of the 111th meeting of RAN1, a lot of efforts have been made to simplify the life cycle management (LCM) process for the general framework, and a model ID solution has been proposed. In the consensus reached, the LCM process based on the premise that the AI / ML model has a model ID and related information and / or model function is studied, at least for part of the AI / ML operation. Further, in the UE-end / UE-side model, the following LCM process mechanisms are studied: for the function-based LCM process, the activation / deactivation / switching / backoff indication based on a single AI / ML function; for the model ID-based LCM process, the model selection / activation / deactivation / switching / backoff indication based on a single model ID.

[0047] Technical problems:

[0048] Based on the disclosure of the present application, if we want to combine the model ID with the AI model-specific reference signal and data collection, at least the following problems have been identified. The technical problems mentioned here are only examples. In addition to these technical problems, some embodiments of the present application can also solve other technical problems. The present application is not limited thereto.

[0049] Problem 1: How to pair AI model specific information with reference signal.

[0050] The CSI framework defined in NR includes CSI-RS resource setting and reporting setting, and is configured to the UE through CSI-ReportConfig in RRC signaling. The CSI-ReportConfig data structure is shown below, which defines three CSI-RS resource types related to reportQuantity for CSI measurement, including channel measurement using the first CSI-RS resource configuration, interference measurement using the second CSI-IM resource configuration, and another interference measurement using the NZP-CSI-RS resource configuration. As mentioned above, the information required by the AI / ML model is quite different from the traditional CSI framework, and the existing CSI-RS configuration method has not yet defined the AI / ML or other use cases, so how to pair the AI / ML model specific information with the reference signal configuration process needs to be further clarified.

[0051] Information element (IE):

[0052]

[0053]

[0054] Further, for SRS resource configuration, SRS resource set / SRS resource / TPC command / SRS resource has been defined for uplink channel measurement or positioning. However, compared with the CSI framework, SRS resource configuration has not yet been associated with any reporting information. As can be seen from the above analysis, when the AI / ML model is located at the UE side, some mandatory data or channel information is required, so a suitable process is needed to ensure the real-time sharing of network information with the UE.

[0055] IE:

[0056]

[0057]

[0058] Problem 2: How to pair AI model specific information with CSI reporting.

[0059] Under the traditional CSI-ReportConfig definition, CSI reporting is managed by the following IEs, including: reportQuantity, reportQuantity-r16, reportQuantity-r17, csi-ReportMode-r17, sharedCMR, reportFreqConfiguration, timeRestrictionForChannel / InterferenceMeasurements, etc. Among them, the three types of reportQuantity are used for different purposes and are highly related to the use scenarios: reportQuantity is used for traditional CSI feedback, reportQuantity-r16 is used for more accurate beam management by introducing L1-SINR information, and reportQuantity-r17 introduces CMR to reduce feedback overhead when the network uses multiple panels.

[0060] IEs:

[0061]

[0062] Currently, with the standardization process, various AI models have been discussed, including single-sided AI / ML models (located at the UE side or gNB side) and bilateral AI / ML models (gNB and UE each have partial functions to complete a specific AI task). Whether these AI / ML models are located only at the gNB side or are bilaterally deployed, additional feedback information is needed to ensure that the AI / ML model can work normally and safely. Similar to the traditional CSI feedback framework, AI / ML model-specific CSI feedback procedures are also needed. Therefore, how to define an AI model-specific CSI feedback method that can be compatible with different AI / ML models needs further discussion, and its application scenarios include but are not limited to positioning, beam management, CSI compression, etc.

[0063] In some embodiments of the present application, an AI / ML model-specific measurement framework is proposed, which involves reference signal configuration, CSI measurement, and CSI reporting. This has not been discussed in RAN1 meetings, but according to the current / future Tdocs discussion, this solution is a natural consequence. Through this concept, the wireless air interface can benefit in the following aspects: reducing signaling overhead, providing scalability for different model requirements under the same type of AI / ML model, and further unifying the processing process of single-sided or bilateral models through a single AI / ML model-specific procedure. The above is some summary of the present application, and some embodiments will be described in detail below.

[0064] Figure 1An example schematic diagram of a basic autoencoder model for enhanced CSI feedback according to embodiments of the application is shown. Figure 1 It is explained that in some embodiments, a basic autoencoder model is as follows: the encoder compresses the original CSI-RS values (original CSI for short) / maximum eigenvectors and outputs it to the gNB for decompression. The new CSI report is a CSI report containing enhanced CSI feedback by AI / ML model. At the UE end, the input is compressed and output to the channel. The input of the encoder can be (maximum) eigenvectors or channel matrix. The compressed output is used as the input of the decoder, and the reconstruction is performed at the gNB end.

[0065] Figure 2 It is explained that in some embodiments, at least one first node 10 (such as at least one user equipment UE), a second node 20 (such as a network device 20), and at least one third node 30 (such as other communication devices) are provided for communication in a communication network system 40. The communication network system 40 includes at least one first node 10 (such as at least one user equipment UE), a second node 20 (such as a network device 20), and at least one third node 30 (such as other communication devices). The at least one first node 10 can include a memory 12, a transceiver 13, and a processor 11 coupled with the memory 12 and the transceiver 13. The at least one second node 20 can include a memory 22, a transceiver 23, and a processor 21 coupled with the memory 22 and the transceiver 23. The at least one third node 30 can include a memory 32, a transceiver 33, and a processor 31 coupled with the memory 32 and the transceiver 33. The processor 11, 21, or 31 can be configured to implement the functions, procedures, and / or methods described in this specification. Layers of a radio interface protocol can be implemented by the processor 11, 21, or 31. The memory 12, 22, or 32 is operatively coupled with the processor 11, 21, or 31 and stores information related to an operation of the processor 11, 21, or 31. The transceiver 13, 23, or 33 is operatively coupled with the processor 11, 21, or 31 and the transceiver 13, 23, or 33 is used to transmit and / or receive a radio signal.

[0066] The processors 11, 21, or 31 can include application-specific integrated circuit (ASIC), other chip sets, logic circuit, and / or a data processing device. The memories 12, 22, or 32 can include read-only memory (ROM), random access memory (RAM), flash memory, storage cards, storage media, and / or other storage devices. The transceivers 13, 23, or 33 can include baseband circuitry for processing radio frequency signals. When the embodiments are implemented in software, the techniques described herein can be implemented using a suitably-programmed processing device (e.g., a computer) in combination with software. The software can be stored on a computer-readable medium, such as the memories 12, 22, or 32. The memories 12, 22, or 32 can be internal to the processors 11, 21, or 31, or external to the processors 11, 21, or 31, in which case the memories 12, 22, or 32 can be communicatively coupled to the processors 11, 21, or 31 via various means known in the art.

[0067] In some embodiments, the processor 11 is configured to determine an AI / ML model specific reference signal configuration, wherein the AI / ML model specific reference signal configuration comprises an indication for indicating an AI / ML model specific reference signal usage, the reference signal usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing process of single-sided models or double-sided models through a single AI / ML model specific procedure.

[0068] In some embodiments, the processor 21 is configured to configure an AI / ML model specific reference signal configuration to the UE 10, wherein the AI / ML model specific reference signal configuration comprises an indication for indicating an AI / ML model specific reference signal usage, the reference signal usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing process of single-sided models or double-sided models through a single AI / ML model specific procedure.

[0069] In some embodiments, the transceiver 13 is configured to receive, from the network device 20, a downlink AI / ML model-specific reference signal configuration. The downlink AI / ML model-specific reference signal configuration comprises a resource type indication indicating at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model process; an identifier of an AI / ML scenario; and / or an identifier of a reference signal. The transceiver 13 is further configured to transmit, to the network device 20, downlink AI / ML model-specific CSI / reporting / feedback. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0070] In some embodiments, the transceiver 13 is configured to receive, from the network device 20, an uplink AI / ML model-specific reference signal configuration. The uplink AI / ML model-specific reference signal configuration comprises a resource type indication indicating at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model process; an identifier of an AI / ML scenario; and / or an identifier of a reference signal. The transceiver 13 is further configured to receive, from the network device 20, uplink AI / ML model-specific CSI / reporting / feedback. The transceiver 13 is further configured to receive, from the network device 20, feedback of AI / ML model-specific information. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0071] In some embodiments, the transceiver 23 is configured to transmit, to the UE 10, a downlink AI / ML model-specific reference signal configuration. The downlink AI / ML model-specific reference signal configuration comprises a resource type indication indicating at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model process; an identifier of an AI / ML scenario; and / or an identifier of a reference signal. The transceiver 23 is further configured to receive, from the UE 10, downlink AI / ML model-specific CSI / reporting / feedback. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0072] In some embodiments, the transceiver 23 is configured to transmit, to the UE 10, an uplink AI / ML model-specific reference signal configuration. The uplink AI / ML model-specific reference signal configuration comprises an indication of a resource type, indicating at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model process; an identifier of an AI / ML scenario; and / or an identifier of a reference signal. The transceiver 23 is further configured to transmit, to the UE 10, an uplink AI / ML model-specific CSI / reporting / feedback, and further configured to feedback, to the UE 10, AI / ML model-specific information. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0073] Figure 4A A flowchart of an AI / ML wireless communication method 800 according to an embodiment of the application is shown. In some embodiments, the method 800 comprises: step 802, configuring, to a UE, an AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration comprises an indication indicating an AI / ML model-specific reference signal usage, the reference signal usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0074] Figure 4B A flowchart of an AI / ML wireless communication method 900 according to an embodiment of the application is shown. In some embodiments, the method 900 comprises: step 902, determining, by a UE, an AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration comprises an indication indicating an AI / ML model-specific reference signal usage, the reference signal usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0075] Figure 5AA flowchart of an AI / ML wireless communication method 500 according to embodiments of the application is shown. In some embodiments, the method 500 comprises: step 502, transmitting, to a UE, an AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration comprises an indication for indicating a reference signal usage, the reference signal usage comprising at least one of: an AI / ML function, an AI / ML model, an AI / ML model procedure, and / or an AI / ML scenario; and / or step 504, receiving, from the UE, an AI / ML model-specific reporting information configuration, wherein the AI / ML model-specific reporting information configuration comprises an indication for indicating AI / ML model information. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0076] In some examples, the indication for indicating the reference signal usage comprises at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model procedure; an identifier of a reference signal; and a sequence-based indication. In some examples, the reference signals related to AI / ML are indicated in different configuration orders through the sequence-based indication. In some examples, the reference signals comprise at least one of: a CSI-RS, a DMRS, a SRS, a PTRS, and / or a PRS.

[0077] In some examples, the indication for indicating the AI / ML model-specific reporting information configuration comprises at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model procedure; and an identifier of a reference signal. In some examples, the AI / ML model-specific reference signal configuration and / or the AI / ML model-specific reporting information configuration is activated and / or deactivated through trigger information, wherein the trigger information comprises at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model procedure; an identifier of a reference signal; an identifier of a reporting quantity; an activation timing; and / or a deactivation timing.

[0078] Figure 5BA flowchart of an AI / ML wireless communication method 600 according to embodiments of the present application is shown. In some embodiments, the method 600 includes: step 602, transmitting, to a UE, an uplink AI / ML model-specific reference signal configuration and / or an uplink measurement report, wherein the uplink measurement report is configured to indicate uplink measurement information between the network device and the UE based on an uplink AI / ML model-specific reference signal transmitted from the network device to the UE; and / or step 604, transmitting, to the UE, an uplink AI / ML model-specific channel state information. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0079] In some examples, the uplink AI / ML model-specific reference signal configuration and / or the uplink measurement report includes an indication for indicating a reference signal usage, the indication including at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model procedure; an identifier of a reference signal; and / or a sequence-based indication. In some examples, the reference signals related to AI / ML are indicated in different configuration orders through the sequence-based indication. In some examples, the reference signals include at least one of: a CSI-RS, a DMRS, an SRS, a PTRS, and / or a PRS. In some examples, the uplink AI / ML model-specific channel state information includes at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model procedure; and / or an identifier of a reference signal.

[0080] Figure 6A A flowchart of an AI / ML wireless communication method 100 performed by a network device according to embodiments of the present application is shown. In some embodiments, the method 100 includes: step 102, transmitting, to a UE, a downlink AI / ML model-specific reference signal configuration, wherein the downlink AI / ML model-specific reference signal configuration includes a resource type indication for indicating at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model procedure; an identifier of an AI / ML scenario; and / or an identifier of a reference signal; and step 104, receiving, from the UE, a downlink AI / ML model-specific CSI / report / feedback. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of unilateral models or bilateral models through a single AI / ML model-specific procedure.

[0081] Figure 6B A flowchart of an AI / ML wireless communication method 200 performed by a network device according to an embodiment of the present application is shown. In some embodiments, the method 200 comprises: step 202, sending an uplink AI / ML model-specific reference signal configuration to a UE, wherein the uplink AI / ML model-specific reference signal configuration comprises at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model process; an identifier of an AI / ML scenario; and / or an identifier of a reference signal; step 204, sending an uplink AI / ML model-specific CSI / report / feedback to the UE; and step 206, feeding back AI / ML model-specific information to the UE. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of one-sided models or two-sided models through a single AI / ML model-specific procedure.

[0082] Figure 7A A flowchart of an AI / ML wireless communication method 300 performed by a UE according to an embodiment of the present application is shown. In some embodiments, the method 300 comprises: step 302, receiving a downlink AI / ML model-specific reference signal configuration from a network device, wherein the downlink AI / ML model-specific reference signal configuration comprises a resource type indication indicating at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model process; an identifier of an AI / ML scenario; and / or an identifier of a reference signal; and step 304, sending a downlink AI / ML model-specific CSI / report / feedback to the network device. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same type of AI / ML model, and unify the processing procedures of one-sided models or two-sided models through a single AI / ML model-specific procedure.

[0083] Figure 7BA flowchart of an AI / ML wireless communication method 400 performed by a UE according to an embodiment of the present application is shown. In some embodiments, the method 400 includes: step 402, receiving, from a network device, an uplink AI / ML model-specific reference signal configuration, wherein the uplink AI / ML model-specific reference signal configuration includes a resource type indication for indicating at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML model process; an identifier of an AI / ML scenario; and / or an identifier of a reference signal; step 404, receiving, from the network device, uplink AI / ML model-specific CSI / reporting / feedback; and step 406, receiving, from the network device, feedback of AI / ML model-specific information. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements under the same AI / ML model, and unify the processing procedures of single-sided models or double-sided models through a single AI / ML model-specific procedure.

[0084] To combine AI / ML model information with reference signaling and CSI measurement / reporting (including downlink and uplink), some embodiments of the present application propose the following overall solutions, as shown below. The overall solutions for downlink and uplink are respectively demonstrated for different AI / ML models (single-sided, double-sided, positioning, beam management, CSI compression, model inference / monitoring, etc.).

[0085] Figure 8 A flowchart of downlink AI / ML model-specific resource configuration and feedback according to an embodiment of the present application is shown. Figure 8 It is explained that, in some embodiments, there are two steps. Step 1: the network sends AI / ML model-specific reference signal and CSI reporting configuration to the UE. Step 2: the UE sends AI / ML model-specific CSI / measurement result feedback to the network. Figure 8 It is further explained that, in some embodiments, for downlink configuration, AI / ML model-specific reference signals are adopted, including but not limited to CSI-RS, SSB, PRS, DMRS, etc. Corresponding to the AI / ML model-specific reference signals, AI / ML model-specific measurement and reporting results are proposed to support network-side AI / ML, and the reporting results can also be empty to support UE-side training or inference, etc.

[0086] Figure 9 A flowchart of uplink AI / ML model-specific resource configuration and feedback according to an embodiment of the present application is shown. Figure 9 It is explained that, in some embodiments, there are two steps. Step 1: the network sends AI / ML model-specific SRS configuration to the UE. Step 2 (optional): the network sends AI / ML model-specific measurement / data information to the UE. Figure 9Further, in some embodiments, for uplink configuration, similar to downlink AI / ML model specific reference signal configuration, AI / ML model specific SRS reference signal is defined in the system. Moreover, different from the fact that there is no corresponding CSI feedback from network to UE for legacy SRS resource configuration, additional uplink CSI feedback to UE is added in this invention to support AI / ML model (training, inference, monitoring, etc.) only at UE side, while also supporting bilateral AI / ML model, such as AI based modulation / coding / decoding, etc.

[0087] Further, in order to support the overall solution, detailed signaling procedures are also involved in the following embodiments, including downlink and uplink AI / ML model specific configuration and reporting.

[0088] Downlink AI / ML model specific signaling:

[0089] 1) Network device (e.g. gNB, AP, etc.) sends AI / ML model specific reference signaling (e.g. CSI-RS, SSB, DMRS, PRS, PTRS, etc.) configuration and related reporting configuration to UE. The implementation is shown in Embodiment 1 and Embodiment 2 below.

[0090] 2) UE receives the configuration from network through RRC or MAC-CE signaling, and performs measurement according to the configured reference signaling and reporting elements. The implementation is shown in Embodiment 3 below.

[0091] 3) UE feeds back the measurement results required by the reporting configuration, and transmits to network device through PUCCH, PUSCH or msg2 / 4 message. The implementation is shown in Embodiment 2 below.

[0092] Uplink AI / ML model specific signaling:

[0093] 1) Network device (e.g. gNB, AP, etc.) sends AI / ML model specific reference signal resource configuration (including SRS, DMRS, etc.) and related feedback information configuration to UE. The implementation is shown in Embodiment 4 and Embodiment 5 below.

[0094] 2) UE receives the configuration from network through RRC or MAC-CE signaling, and sends SRS according to the configured timing and frequency information. The implementation is shown in Embodiment 3 below.

[0095] 3) Network device can perform measurement according to SRS signal and required feedback information.

[0096] 4) Network device transmits AI / ML model specific measurement results to UE. The implementation is shown in Embodiment 5 below.

[0097] Embodiments:

[0098] Embodiment 1: Downlink AI / ML model specific reference signal configuration.

[0099] In addition to the conventional reference signals defined in Rel.15, Rel.16, Rel.17 and / or Rel.18, new reference signal resource types can be defined to support the regular operation of various AI / ML models in wireless systems. A flag can be added in the reference signal through which the UE can distinguish the purpose of the configured reference signal, whether it is used for non-AI / ML measurement or AI / ML measurement, etc. For example, an AI / ML model specific resource type can be used and indicated to the UE side, which can be represented as a string, ID or reference signal name, and the specific implementation is shown in Table 2.

[0100] In some examples, a base station configures an AI / ML model specific reference signal configuration, e.g., a downlink AI / ML model specific reference signal configuration, to a UE, wherein the AI / ML model specific reference signal configuration includes an indication to indicate an AI / ML model specific reference signal purpose, and the reference signal purpose includes at least one of: an AI / ML function, an AI / ML model, an AI / ML model procedure, and / or an AI / ML scenario.

[0101] In some examples, the indication to indicate the reference signal purpose includes at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model procedure; an identifier of a reference signal; and a sequence based indication. In some examples, the reference signals related to AI / ML are indicated in different configuration order through the sequence based indication.

[0102] Table 2: Direct downlink AI / ML model specific reference signal configuration.

[0103]

[0104] In Table 2, for example, the resource type indication method includes string, ID, reference signal name indication, etc. The string is used to distinguish the reference signal type, such as non-AI / ML model type and AI / ML model type. ID 0 indicates that the configuration is used for CSI measurement without AI mode. ID 1 indicates that the configuration is used for AI / ML model based operation, including AI / ML model training / monitoring, and CSI feedback. The reference signal name indication can include CSI-RS resource / resource set / resource setting and AI mode CSI-RS resource / resource set / resource setting.

[0105] In addition, considering that there are multiple AI / ML models for different purposes, such as positioning, CSI compression, beam management, etc., the definition of Table 2 needs to be further classified to support various AI / ML models (functions). These can also be represented as strings, IDs, bitmaps, or reference signal names, with detailed implementations as shown in Table 3.

[0106] Table 3: Direct signaling for distinguishing AI functions.

[0107]

[0108] In Table 3, for example, the resource type indication method includes strings, IDs, bitmaps, reference signal name indications, etc. Strings are used to distinguish AI functions, such as "positioning AI / ML model based", indicating that this configuration is used for positioning operations based on AI / ML models, and the same logic applies to "beam management", "CSI compression", etc. ID1: This configuration is used for positioning operations based on AI / ML models. ID2: This configuration is used for beam management operations based on AI / ML models. The same logic applies to other AI / ML-based functions. Bitmaps are used to distinguish AI functions, with each bit representing a specific AI function. Reference signal names are used to distinguish AI functions and can include AI positioning CSI-RS resources / resource sets / resource settings, etc.

[0109] In addition, for one AI function, multiple processes can be defined, including data collection, inference, monitoring, and training. This indication can also be an AI / ML model process for distinguishing different processes under one AI function, and can be represented by strings, IDs, bitmaps, or other methods. Taking bitmap as an example, "0001" indicates that this configuration is used for training, "0010" indicates that this configuration is used for monitoring, "0100" indicates that this configuration is used for inference, "1000" indicates that this configuration is used for data collection, etc.

[0110] In addition, for a specific AI function, there can be multiple AI / ML models, and the resource type will be structured using a hierarchical method to meet the needs of different scenarios, as shown in the following Table 4.

[0111] Table 4: Direct signaling for distinguishing AI functions.

[0112]

[0113]

[0114] In Table 4, in some examples, the AI function is distinguished using the function definition, ID-based indication, bitmap-based indication, and its content information #1, content information #2, and content information #3. Taking the ID-based indication and its content information #1, content information #2, and content information #3 as an example, in content information #1, ID 0 represents non-AI / ML model-based, and ID 1 represents AI / ML model-based; in content information #2, ID 0 represents a positioning AI / ML model, ID 1 represents a beam management AI / ML model, and ID 2 represents a CSI enhancement AI / ML model; in content information #3, ID 0 represents model 1 in function A, ID 1 represents model 2 in function A, and ID 2 represents model 3 in function A. This can achieve the differentiation of AI functions.

[0115] In some examples, the AI / ML model-specific reference signal includes at least one of the following: CSI-RS, DMRS, SRS, PTRS, PRS.

[0116] In addition to the above-mentioned method of directly indicating the use of UE configuration information, implicit indication methods should also be considered, as follows.

[0117] IE:

[0118]

[0119]

[0120] In embodiment 1, the use of the reference signal is distinguished using the configuration order. Taking CSI-RS as an example (the same applies to other reference signals): the first resource configuration is used for channel measurement of non-AI / ML model; the second resource configuration is used for interference measurement based on CSI-IM resource; the third resource configuration is used for interference measurement based on NZP-CSI-RS; the fourth resource configuration is used for CSI operation (measurement, monitoring, etc.) based on AI / ML model; the fifth resource configuration is used for beam management operation (measurement, monitoring, etc.) based on AI / ML model. In some examples, through the order-based indication, the AI / ML-related reference signal in the AI / ML model-specific reference signal configuration is distinguished in different configuration orders.

[0121] Embodiment 2: Downlink AI / ML model-specific CSI / report / feedback to the network.

[0122] In some examples, the base station receives (or the UE sends) an AI / ML model specific reporting information configuration from the UE, where the AI / ML model specific reporting information configuration includes an indication to indicate the AI / ML model information. The AI / ML model specific reporting information configuration includes, for example, downlink AI / ML model specific CSI / reporting / feedback. In some examples, the indication to indicate the AI / ML model information includes at least one of the following: an identifier of the AI / ML model; an identifier of the AI / ML type; an identifier of the AI / ML function; an identifier of the AI / ML scenario; an identifier of the AI / ML model process; and / or an identifier of the reference signal.

[0123] For AI / ML model specific information feedback needs, there are multiple ways to distinguish the needs of AI functions, as shown below.

[0124] 1) Define a dedicated reporting quantity for AI / ML model specific operation.

[0125] 2) Dedicated reporting configuration, which contains AI / ML model specific information.

[0126] 3) Dedicated measurement configuration, which contains AI / ML model specific information.

[0127] 4) For 2) and 3), AI / ML model specific information can be represented as AI / ML model ID, AI mode type, AI / ML model function, AI / ML model process, etc.

[0128] In particular, for the implementation method of 1), similar to the currently defined reporting configuration, a dedicated reporting quantity is defined and dedicated to AI / ML model operation, and under this AI dedicated reporting quantity, function specific reporting information is contained, and its example data structure and detailed description are shown as follows.

[0129] IE:

[0130]

[0131] None (None): indicates that the UE does not need feedback, mainly for training, monitoring, updating / switching / backoff, inference or other purposes of the UE side AI / ML model.

[0132] Positioning: feedback information required for AI / ML model based positioning and AI / ML model specific information located at the network side, where the AI / ML model specific information can be represented as AI / ML model ID, AI mode type, AI / ML model function or reference signal index (CSI-RS resource index, DMRS resource index, SSB resource index), etc.

[0133] Beam management: Feedback information needed for AI / ML model based management and AI / ML model specific information are located at network side, where AI / ML model specific information can be represented as AI / ML model ID, AI pattern type, AI / ML model function or reference signal index (CSI-RS resource index, DMRS resource index, SSB resource index) etc.

[0134] CSI: Feedback information needed for AI / ML model based CSI and AI / ML model specific information are located at network side, where AI / ML model specific information can be represented as AI / ML model ID, AI pattern type, AI / ML model function or reference signal index (CSI-RS resource index, DMRS resource index, SSB resource index) etc.

[0135] Embodiment 3: On-demand AI / ML model specific operation triggering.

[0136] In some examples, AI / ML model specific reference signal configuration and / or AI / ML model specific reporting information configuration are activated / deactivated by triggering information.

[0137] Considering the power consumption and computing resource limitation of UE, AI / ML model operation is only activated in certain specific scenarios, thus the transmission and measurement or reporting of reference signal needs to be on-demand with non-periodic or semi-static operation. In order to reduce air interface overhead, the following information can be carried by MAC-CE, PDCCH or RRC signaling:

[0138] 1) Activation / deactivation information for specific AI / ML model, carried by MAC-CE or PDCCH.

[0139] 2) Activation / deactivation timing for AI / ML model, can be predefined or carried by MAC-CE / RRC.

[0140] For 1), the indication information can be a bitmap, an AI / ML model ID, an AI / ML model type, or an AI function name. Taking the bitmap as an example, assuming there are 4 AI / ML models (positioning, beam management, CSI-1, CSI-2, etc.) in the system, if the bitmap is “1111”, it means that all AI / ML models are in the active state; if the bitmap is “0000”, it means that no AI / ML model is used in the system. For 2), the activation / deactivation timing is AI / ML model specific, for example, the activation timing of the beam management AI / ML model is 10 slots, and the deactivation timing is also 10 slots; while the activation timing of the CSI AI / ML model is 20 slots, and the deactivation timing is also 20 slots, because it needs a longer data collection time, etc. In addition, there can be multiple AI / ML models performing the same function, and the activation / deactivation timing will be different due to the model ID, as shown in Table 1. Figure 10 In some examples, the trigger information includes at least one of the following: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of a reference signal; an identifier of an AI / ML model process; an identifier of a reporting quantity; an activation timing; and / or a deactivation timing.

[0141] Figure 10 A schematic diagram of AI / ML model specific activation / deactivation timing is shown according to an embodiment of the present application. Figure 10 It is explained that, in some embodiments, at time slot #1, the UE receives a configuration from the network device; at time slot #10, the activation timing of the beam management AI / ML model is 10 slots, and the deactivation timing is also 10 slots; at time slot #20, for the CSI AI / ML model, the activation timing is 20 slots, and the deactivation timing is also 20 slots, because a longer data collection time is needed, etc.

[0142] Embodiment 4: Uplink AI / ML model specific reference signal configuration.

[0143] In some examples, the base station configures the UE with an AI / ML model specific reference signal configuration, such as an uplink AI / ML model specific reference signal configuration, wherein the AI / ML model specific reference signal configuration includes an indication indicating an AI / ML model specific reference signal usage, and the reference signal usage includes at least one of the following: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario.

[0144] In some examples, the indication for indicating reference signal usage includes at least one of: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model procedure; an identifier of a reference signal; and a sequence-based indication. In some examples, the reference signals related to AI / ML are differentiated by different configuration order through the sequence-based indication.

[0145] For a conventional uplink reference signal configuration, taking SRS as an example, its data structure is shown as follows. In addition to general information (such as resource set / resource / timing / combination, etc.), the usage of SRS has been introduced in Rel. 17. For the usage item, it indicates whether the SRS resource set is used for beam management, codebook-based or non-codebook-based transmission, or antenna switching; for the usage PDC, it indicates that the SRS resource set is used for propagation delay compensation.

[0146] IE:

[0147]

[0148]

[0149] Similar to the interaction of downlink AI / ML model-specific reference signals, a flag can be added in the reference signal to enable the UE to distinguish the transmission mode of the configured reference signal, such as for non-AI measurement or AI / ML measurement, etc. In one example, an AI / ML model-specific resource type is used and indicated to the UE side, which can be represented as a string, ID or reference signal name, as shown in Table 5.

[0150] Table 5: Direct downlink AI / ML model-specific reference signal configuration.

[0151]

[0152] In Table 5, the resource type indication method can include string and ID indication, etc. The string is used to distinguish the type or usage of the reference signal, such as non-AI / ML model type and AI / ML model type. ID 0 can represent that the configuration is used for a traditional mode, while ID 1 is used for AI / ML model-based operation, including AI / ML model training / monitoring.

[0153] In addition, considering that there are multiple AI / ML models for different purposes, such as positioning, CSI compression, beam management, etc., the definition of Table 5 needs to be further classified to support various AI / ML models (functions). These can also be represented as a string, ID, bitmap or reference signal name, as shown in Table 6.

[0154] Table 6: Direct signaling for distinguishing AI functions.

[0155]

[0156] In Table 6, the resource type indication method can include string, ID, and bitmap indication, etc. In the string, “positioning AI / ML model based” means that the configuration is used for AI / ML model based positioning operation. ID 1 means that the configuration is used for AI / ML model based positioning operation. ID 2 means that the configuration is used for AI / ML model based beam management operation, and so on. The same logic applies to “beam management”, “CSI compression”, etc. The bitmap is used to distinguish AI functions, each bit represents a specific AI function, 0 means invalid, and 1 means valid. For example, in the bitmap, “1000”: the last bit is “1”, which means that the configuration is used for AI / ML model based positioning operation.

[0157] Embodiment 5: Uplink AI / ML model specific CSI / reporting / feedback to UE.

[0158] In some examples, the base station configures one report to the UE, which is used to indicate the measurement information based on AI / ML model specific reference signal between the base station and the UE. In some examples, the base station configures AI / ML model specific channel state information to the UE. In some examples, the AI / ML model specific channel state information includes at least one of the following: AI / ML model identifier; AI / ML type identifier; AI / ML function identifier; AI / ML scenario identifier; AI / ML model process identifier; and / or reference signal identifier.

[0159] Since the uplink data transmission is controlled by the network device, the UE does not need to obtain the channel state information about the uplink transmission. But for the AI / ML model operation located at the UE side (e.g. single-sided or double-sided), as mentioned above, the network side measurement information of the uplink reference signal is necessary. To solve the above problem, the following framework is proposed:

[0160] 1) Define a new report configuration issued by the network device to the UE.

[0161] 2) The report configuration includes at least AI / ML model specific uplink reference signal configuration, network device to UE feedback information, and feedback behavior (timing, frequency, etc.). The feedback information is from the network device to the UE, which is AI / ML model specific, where the AI / ML model specific information can be represented as AI / ML model ID, AI mode type, AI / ML model function, etc.

[0162] Embodiment 6: AI / ML model specific information feedback from network to UE.

[0163] Figure 11 is a schematic diagram of AI / ML model specific uplink feedback according to an embodiment of the present application. Figure 11 It is explained that in some embodiments, step 1: UE receives AI / ML model specific configuration from network device; step 2: network device receives AI / ML model specific reference signal from UE; step 3: network device performs measurement based on AI / ML model specific reference signal; step 4: network device transmits AI / ML model specific information feedback to UE.

[0164] As shown in Figure 11 , for AI / ML model specific information feedback from network device to UE side, the following information will be included:

[0165] 1) AI / ML model related description information.

[0166] 2) AI / ML model specific data obtained based on uplink reference signal measurement or calculation.

[0167] For AI / ML model related description information, it can be represented as the following content:

[0168] 1) AI / ML model ID, AI mode type, AI / ML model function, etc. These directly reveal the feedback content from the network, and the UE can perform AI operation based on the decoded information.

[0169] 2) AI / ML model specific SRS resource index. SRS resource index is also valid because AI / ML model information is part of SRS resource configuration.

[0170] Figure 12Ais a block diagram of a wireless communication device 1000 (e.g., a UE) according to embodiments of the application. The wireless communication device 1000 can be a first node, e.g., a UE. The wireless communication device 1000 includes a transceiver 1001 configured to receive, from a network device, a downlink AI / ML model-specific reference signal configuration, wherein the downlink AI / ML model-specific reference signal configuration comprises a resource type indication indicating at least one of: an identifier of an AI / ML model, an identifier of an AI / ML type, an identifier of an AI / ML function, an identifier of an AI / ML model process, an identifier of an AI / ML scenario, and / or an identifier of a reference signal. The transceiver 1001 is further configured to transmit, to the network device, a downlink AI / ML model-specific CSI / reporting / feedback. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model needs of one AI / ML model type, and unify the procedures of single-sided models or double-sided models through a single AI / ML model-specific process.

[0171] In some examples, the resource type indication comprises at least one of: a string-based indication, an ID-based indication, a bitmap-based indication, a reference signal name-based indication, and / or a configuration order-based indication. In some examples, the resource type indication adopts a hierarchical method structure to meet different scenario needs of multiple AI / ML models for a certain specific AI / ML function. In some examples, the one or more reference signal types comprise at least one of: a configuration for CSI measurement without AI / ML mode, and / or a configuration for one or more AI / ML model operations including AI / ML model training / monitoring and / or CSI feedback. In some examples, the one or more AI / ML functions comprise at least one of: positioning based on one or more AI / ML model operations, beam management based on one or more AI / ML model operations, and / or CSI compression based on one or more AI / ML model operations.

[0172] In some examples, the configuration order comprises at least one of: a first resource configuration for non-AI based channel measurement; a second resource configuration for interference measurement (IM) based on CSI-IM resource; a third resource configuration for interference measurement based on non-zero power channel state information reference signal (NZP-CSI-RS); a fourth resource configuration for CSI operation based on AI / ML model; and / or a fifth resource configuration for beam management operation based on AI / ML model. In some examples, the downlink AI / ML model specific CSI / reporting / feedback comprises one dedicated indication for indicating one or more AI / ML functionalities. In some examples, the dedicated indication comprises at least one of: a dedicated reporting quantity for one or more AI / ML model based operations, a dedicated reporting configuration containing AI / ML model specific information, and / or a dedicated measurement configuration containing AI / ML model specific information.

[0173] In some examples, the AI / ML model specific information comprises at least one of: AI / ML model ID, AI / ML mode type, AI / ML model functionality, and / or reference signal index. In some examples, the one or more AI / ML model based operations comprises downlink AI / ML model specific operation triggered by aperiodic on-demand operation or semi-static on-demand operation. In some examples, the downlink AI / ML model specific operation is triggered by activation / deactivation information for a specific AI / ML model, which is carried by media access control-control element (MAC-CE) signaling or physical downlink control channel (PDCCH) signaling. In some examples, the downlink AI / ML model specific operation is triggered by AI / ML model specific activation / deactivation timing, which is predefined, or carried by MAC-CE signaling or radio resource control (RRC) signaling.

[0174] In another embodiment, the transceiver 1001 is configured to receive, from the network device, an uplink AI / ML model-specific reference signal configuration, wherein the uplink AI / ML model-specific reference signal configuration comprises a resource type indication indicating at least one of: an identifier of an AI / ML model, an identifier of an AI / ML type, an identifier of an AI / ML function, an identifier of an AI / ML model procedure, an identifier of an AI / ML scenario, and / or an identifier of a reference signal. The transceiver 1001 is further configured to receive, from the network device, an uplink AI / ML model-specific CSI / reporting / feedback. The transceiver 1001 is further configured to receive, from the network device, feedback of AI / ML model-specific information. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements of one AI / ML model type, and unify the procedures of one-sided models or two-sided models through a single AI / ML model-specific procedure.

[0175] In some examples, the resource type indication comprises at least one of: a string-based indication, an ID-based indication, a bitmap-based indication, a reference signal name-based indication, and / or a configuration order-based indication. In some examples, the resource type indication adopts a hierarchical method structure to meet different scenario requirements of multiple AI / ML models for a certain specific AI / ML function. In some examples, the one or more reference signal types comprise at least one of: a configuration for CSI measurement without AI / ML mode, and / or a configuration for one or more AI / ML model operations including AI / ML model training / monitoring and / or CSI feedback. In some examples, the one or more AI / ML functions comprise: positioning based on one or more AI / ML model operations, beam management based on one or more AI / ML model operations, and / or CSI compression based on one or more AI / ML model operations.

[0176] In some examples, the configuration order includes: a first resource configuration for non-AI channel measurement; a second resource configuration for CSI-IM resource based interference measurement; a third resource configuration for NZP-CSI-RS based interference measurement; a fourth resource configuration for AI / ML model based CSI operation; and / or a fifth resource configuration for AI / ML model based beam management operation. In some examples, the uplink AI / ML model specific CSI / reporting / feedback includes AI / ML model specific uplink reference signal configuration, feedback information from network device to UE, and / or feedback behavior. In some examples, the AI / ML model specific information includes AI / ML model related description information and / or AI / ML model specific data from uplink reference signal measurement or calculation. In some examples, the AI / ML model related description information includes AI / ML model ID, AI / ML mode type, AI / ML model function, reference signal index, and / or AI / ML model specific sounding reference signal resource index.

[0177] Figure 12B is a block diagram of a wireless communication device 1100 (e.g., a network device) according to embodiments of the application. The wireless communication device 1100 can be a second node, e.g., a network device. The wireless communication device 1100 includes a transceiver 1101 configured to transmit, to a user equipment (UE), a downlink AI / ML model specific reference signal configuration, wherein the downlink AI / ML model specific reference signal configuration includes a resource type indication for indicating one or more reference signal types, one or more AI / ML functions, and / or reference signal usages. The transceiver 1101 is also configured to receive, from the UE, downlink AI / ML model specific CSI / reporting / feedback. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements of one AI / ML model type, and unify the procedures of single-sided models or double-sided models through a single AI / ML model specific procedure.

[0178] In another embodiment, the transceiver 1101 is configured to transmit, to a UE, an uplink AI / ML model specific reference signal configuration, wherein the uplink AI / ML model specific reference signal configuration includes a resource type indication for indicating one or more reference signal types, one or more AI / ML functions, and / or reference signal usages. The transceiver 1101 is also configured to transmit, to the UE, uplink AI / ML model specific CSI / reporting / feedback. The transceiver 1101 is also configured to feed back, to the UE, AI / ML model specific information. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements of one AI / ML model type, and unify the procedures of single-sided models or double-sided models through a single AI / ML model specific procedure.

[0179] Figure 13Ais a block diagram of a wireless communication device 1200 (e.g., UE) according to embodiments of the application. The wireless communication device 1200 can be a first node, e.g., a UE. The wireless communication device 1200 includes a determiner 1201 configured to determine an AI / ML model specific reference signal configuration, wherein the AI / ML model specific reference signal configuration includes an indication for indicating a reference signal usage of an AI / ML model specific reference signal, the reference signal usage including at least one of: an AI / ML function, an AI / ML model, an AI / ML model process, and / or an AI / ML scenario. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model needs of one AI / ML model type, and unify the procedures of single-sided or double-sided models through a single AI / ML model specific process.

[0180] In some examples, the indication for indicating the reference signal usage includes: an identifier of the AI / ML model; an identifier of the AI / ML type; an identifier of the AI / ML function; an identifier of the AI / ML scenario; an identifier of the AI / ML model process; an identifier of the reference signal; a string-based indication; an ID-based indication; a bitmap-based indication; a hierarchical method-based indication; and / or a sequential-based indication. In some examples, through the sequential-based indication, the AI / ML related reference signals in the AI / ML model specific reference signal configuration are indicated in different configuration sequences. In some examples, the AI / ML model specific reference signal includes: a CSI-RS, a DMRS, a SRS, a PTRS, and / or a PRS. In some examples, the method further includes receiving, from the UE, an AI / ML model specific reporting information configuration, wherein the AI / ML model specific reporting information configuration includes an indication for indicating AI / ML model information. In some examples, the indication for indicating the AI / ML model information includes: an identifier of the AI / ML model; an identifier of the AI / ML type; an identifier of the AI / ML function; an identifier of the AI / ML scenario; an identifier of the AI / ML model process; and / or an identifier of the reference signal. In some examples, the AI / ML model specific reference signal configuration and / or the AI / ML model specific reporting information configuration is activated / deactivated through trigger information.

[0181] In some examples, the trigger information includes: an identifier of the AI / ML model; an identifier of the AI / ML type; an identifier of the AI / ML function; an identifier of the AI / ML scenario; an identifier of the AI / ML model procedure; an identifier of the reference signal; an identifier of the reporting quantity; an activation timing; and / or a deactivation timing. In some examples, the method further includes configuring the UE with a report, where the report is configured to indicate measurement information between the network device and the UE based on the AI / ML model specific reference signal. In some examples, the method further includes configuring the UE with AI / ML model specific channel state information. In some examples, the AI / ML model specific channel state information feedback includes: an identifier of the AI / ML model; an identifier of the AI / ML type; an identifier of the AI / ML function; an identifier of the AI / ML scenario; an identifier of the AI / ML model procedure; and / or an identifier of the reference signal.

[0182] Figure 13B is a block diagram of a wireless communication device 1300 (e.g., a network device) according to embodiments of the application. The wireless communication device 1300 can be a second node, e.g., a network device. The wireless communication device 1300 includes a configurator 1301 configured to configure a UE with an AI / ML model specific reference signal configuration, where the AI / ML model specific reference signal configuration includes an indication indicating a reference signal usage of the AI / ML model specific reference signal, the reference signal usage including at least one of: an AI / ML function, an AI / ML model, an AI / ML model procedure, and / or an AI / ML scenario. This can solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements of one AI / ML model type, and unify the procedure of one-sided or two-sided models through a single AI / ML model specific procedure.

[0183] In some examples, the indication to indicate reference signal usage includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model process; an identifier of a reference signal; a string-based indication; an ID-based indication; a bitmap-based indication; a hierarchical method-based indication; and / or a sequential-based indication. In some examples, by the sequential-based indication, AI / ML related reference signals of AI / ML model specific reference signal configurations are indicated in different configuration sequences. In some examples, the AI / ML model specific reference signal includes: a CSI-RS, a DMRS, an SRS, a PTRS, and / or a PRS. In some examples, the method further includes transmitting, to the network device, an AI / ML model specific reporting information configuration, wherein the AI / ML model specific reporting information configuration includes an indication to indicate AI / ML model information. In some examples, the indication to indicate the AI / ML model information includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model process; and / or an identifier of a reference signal.

[0184] In some examples, the AI / ML model specific reference signal configuration and / or the AI / ML model specific reporting information configuration is activated / deactivated by trigger information. In some examples, the trigger information includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model process; an identifier of a reference signal; an identifier of a reporting quantity; an activation timing; and / or a deactivation timing. In some examples, the method further includes determining, by the UE, a report, wherein the report is configured to indicate, based on the AI / ML model specific reference signal, information between the network device and the UE. In some examples, the method further includes determining, by the UE, an AI / ML model specific channel state information feedback. In some examples, the AI / ML model specific channel state information includes: an identifier of an AI / ML model; an identifier of an AI / ML type; an identifier of an AI / ML function; an identifier of an AI / ML scenario; an identifier of an AI / ML model process; and / or an identifier of a reference signal.

[0185] In the above embodiments of the present application, the term “ / ” can be interpreted to mean “and / or”.

[0186] In summary, in some embodiments of the present application, downlink AI / ML model specific reference signal configuration, downlink AI / ML model specific CSI / reporting / feedback to network, on-demand triggering of downlink AI / ML model specific operation, uplink AI / ML model specific reference signal configuration, uplink AI / ML model specific CSI / reporting / feedback to UE, and / or AI / ML model specific information feedback from network to UE are disclosed to solve the problems in the prior art, reduce signaling overhead, provide scalability for different model requirements of one AI / ML model type, and unify the process of unilateral or bilateral models through a single AI / ML model specific process.

[0187] Figure 14 FIG. 7 is a block diagram of a wireless communication example system 700 according to an embodiment of the present application. The embodiments described herein can be implemented into this system with any suitable configuration of hardware and / or software. Figure 14 The system 700 is shown to include radio frequency (RF) circuitry 710, baseband circuitry 720, application circuitry 730, a memory / storage 740, a display 750, a camera 760, a sensor 770, and an input / output (I / O) interface 780 coupled together at least via the bus 792. The application circuitry 730 can include circuitry such as one or more processor cores, and / or one or more hardware modules.

[0188] The monitoring of the ML model is not limited to the UE or the gNB, but can also be performed in a third node, and the related signaling and data need to be reported to the third node. The third node can be a UE, a gNB, or a server. The embodiment methods in the present application are all applicable. In this way, the signaling overhead between the gNB and the UE can be reduced.

[0189] While this application has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the application is not to be limited to the disclosed embodiments, but on the contrary, is intended to cover various arrangements modification and equivalents thereof that fall within the scope of the appended claims which do not depart from the spirit or scope of the application.

Claims

1. A wireless communication method for artificial intelligence (AI) / machine learning (ML), characterized in that, include: Configure AI / ML model-specific reference signal configuration to user equipment (UE), wherein the AI / ML model-specific reference signal configuration includes an indication of the purpose of the reference signal for indicating the AI / ML model-specific reference signal, the purpose of the reference signal including at least one of the following: AI / ML function, AI / ML model, AI / ML model process and / or AI / ML scenario.

2. The method according to claim 1, characterized in that, The indication used to indicate the purpose of the reference signal includes at least one of the following: The identifier of the AI / ML model; Identifiers of AI / ML type; The identifier of the AI / ML function; The identifier for the AI / ML scenario; The identifier of the AI / ML model process; Identifier of the reference signal; String-based instructions; ID-based indication; Bitmap-based indication; Instructions based on a hierarchical approach; and / or Sequence-based instructions.

3. The method according to claim 2, characterized in that, The sequence-based indications specify the purpose of the AI / ML model-specific reference signals in different configuration orders.

4. The method according to any one of claims 1 to 3, characterized in that, The AI / ML model-specific reference signal includes at least one of the following: Channel State Information Reference Signal (CSI-RS), Demodulation Reference Signal (DMRS), Sound Reference Signal (SRS), Phase Tracking Reference Signal (PTRS), and / or Positioning Reference Signal (PRS).

5. The method according to any one of claims 1 to 4, characterized in that, Also includes: The UE receives AI / ML model-specific report information configuration, wherein the AI / ML model-specific report information configuration includes an indication for indicating AI / ML model information.

6. The method according to claim 5, characterized in that, Indications used to indicate information about the AI / ML model include at least one of the following: The identifier of the AI / ML model; Identifiers of AI / ML type; The identifier of the AI / ML function; The identifier of the AI / ML model process; The identifier of the AI / ML scenario; and / or The identifier of the reference signal.

7. The method according to claim 5 or 6, characterized in that, The AI / ML model-specific reference signal configuration and / or the AI / ML model-specific reporting information configuration are activated and / or deactivated by trigger information.

8. The method according to claim 7, characterized in that, The triggering information includes at least one of the following: The identifier of the AI / ML model; Identifiers of AI / ML type; The identifier of the AI / ML function; The identifier for the AI / ML scenario; The identifier of the AI / ML model process; Identifier of the reference signal; Identifier for the reported quantity; Activation sequence; and / or Deactivate timing.

9. The method according to any one of claims 1 to 4, characterized in that, Also includes: A report is configured to be sent to the UE, wherein the report is configured to indicate measurement information between the network device and the UE based on a specific reference signal of the AI / ML model.

10. The method according to any one of claims 1 to 4, characterized in that, Also includes: Configure AI / ML model-specific channel state information for the UE.

11. The method according to claim 10, characterized in that, in, The AI / ML model-specific channel state information includes at least one of the following: The identifier of the AI / ML model; Identifiers of AI / ML type; The identifier of the AI / ML function; The identifier of the AI / ML model process; The identifier of the AI / ML scenario; and / or The identifier of the reference signal.

12. A wireless communication method for artificial intelligence (AI) / machine learning (ML), characterized in that, include: The user equipment (UE) determines the AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration includes an indication of the purpose of the reference signal for indicating the AI / ML model-specific reference signal, the purpose of which includes at least one of the following: AI / ML function, AI / ML model, AI / ML model process and / or AI / ML scenario.

13. The method according to claim 12, characterized in that, The indication used to indicate the purpose of the reference signal includes at least one of the following: The identifier of the AI / ML model; Identifiers of AI / ML type; The identifier of the AI / ML function; The identifier for the AI / ML scenario; The identifier of the AI / ML model process; Identifier of the reference signal; String-based instructions; ID-based indication; Bitmap-based indication; Instructions based on a hierarchical approach; and / or Sequence-based instructions.

14. The method according to claim 13, characterized in that, The sequence-based indications specify the purpose of the AI / ML model-specific reference signals in different configuration orders.

15. The method according to any one of claims 12 to 14, characterized in that, The AI / ML model-specific reference signal includes at least one of the following: Channel State Information Reference Signal (CSI-RS), Demodulation Reference Signal (DMRS), Sound Reference Signal (SRS), Phase Tracking Reference Signal (PTRS), and / or Positioning Reference Signal (PRS).

16. The method according to any one of claims 12 to 15, characterized in that, Also includes: The configuration for sending or transmitting AI / ML model-specific reporting information from or to a network device includes instructions for indicating AI / ML model information.

17. The method according to claim 16, characterized in that, Indications used to indicate information about the AI / ML model include at least one of the following: The identifier of the AI / ML model; Identifiers of AI / ML type; The identifier of the AI / ML function; The identifier of the AI / ML model process; The identifier of the AI / ML scenario; and / or The identifier of the reference signal.

18. The method according to claim 16 or 17, characterized in that, The AI / ML model-specific reference signal configuration and / or the AI / ML model-specific reporting information configuration are activated and / or deactivated by trigger information.

19. The method according to claim 18, characterized in that, The triggering information includes at least one of the following: The identifier of the AI / ML model; Identifiers of AI / ML type; The identifier of the AI / ML function; The identifier for the AI / ML scenario; The identifier of the AI / ML model process; Identifier of the reference signal; Identifier for the reported quantity; Activation sequence; and / or Deactivate timing.

20. The method according to any one of claims 12 to 15, characterized in that, Also includes: The UE determines a report, wherein the report is configured to indicate measurement information between the network device and the UE based on a specific reference signal of the AI / ML model.

21. The method according to any one of claims 12 to 15, characterized in that, Also includes: The UE determines the specific channel state information of the AI / ML model.

22. The method according to claim 21, characterized in that, The AI / ML model-specific channel state information includes at least one of the following: The identifier of the AI / ML model; Identifiers of AI / ML type; The identifier of the AI / ML function; The identifier of the AI / ML model process; The identifier of the AI / ML scenario; and / or The identifier of the reference signal.

23. A network device, characterized in that, include: Memory; transceiver; as well as A processor coupled to the memory and the transceiver; The processor is configured to perform the method according to any one of claims 1 to 11.

24. A user equipment (UE), characterized in that, include: Memory; transceiver; as well as A processor coupled to the memory and the transceiver; The processor is configured to perform the method of any one of claims 12 to 22.

25. A wireless communication device, characterized in that, include: The configurator is configured to configure an AI / ML model-specific reference signal configuration to a user equipment (UE), wherein the AI / ML model-specific reference signal configuration includes an indication of the purpose of the reference signal for indicating the AI / ML model-specific reference signal, the purpose of which includes at least one of the following: AI / ML function, AI / ML model, AI / ML model process and / or AI / ML scenario.

26. A wireless communication device, characterized in that, include: The determiner is configured to determine an AI / ML model-specific reference signal configuration, wherein the AI / ML model-specific reference signal configuration includes an indication of the purpose of the reference signal for indicating the AI / ML model-specific reference signal, the purpose of the reference signal including at least one of: AI / ML function, AI / ML model, AI / ML model process and / or AI / ML scene.

Images