Enhanced data collection for CSI compression modeling

Abstract

A base station is configured to transmit, to a user equipment (UE), a training channel state information reference signal (CSI-RS) configuration including auxiliary information, wherein the auxiliary information includes an AI tag for a training CSI-RS to be measured; transmitting a start indication to the UE to prompt the UE to start measuring the training CSI-RS; and receiving a measurement report including a measurement result for the training CSI-RS.

Description

[0001] Priority / Incorporation by Reference

[0002] This application claims priority to U.S. Provisional Application Serial No. 63 / 381,171, filed on October 27, 2022, and entitled “Enhanced Data Collection for CSI Compression Modeling,” which is incorporated herein by reference in its entirety. Background Art

[0003] Recent 3rd Generation Partnership Project (3GPP) agreements have focused on channel state information (CSI) compression (i.e., encoding / decoding) with joint training for use with artificial intelligence (AI) and machine learning (ML) models to enhance CSI feedback behavior in both base stations and user equipment (UE).

[0004] The enhancements to model training can be narrowed down to three main use cases (i.e., scenarios). Type 1: joint training of dual-side models at a single side / entity (e.g., user equipment (UE) side or network side); Type 2: joint training of dual-side models at both the network side and the UE side; and Type 3: separate training at the network side and the UE side, where the UE-side CSI generation part and the network-side CSI reconstruction part are trained by the UE side and the network side, respectively (hereinafter referred to as split training). Summary of the invention

[0005] Some example embodiments relate to an apparatus of a base station, the apparatus comprising a processing circuit configured to configure a transceiver circuit to transmit a training channel state information reference signal (CSI-RS) configuration including auxiliary information to a user equipment (UE), wherein the auxiliary information includes an AI tag for a training CSI-RS to be measured; configure the transceiver circuit to transmit a start indication to the UE to prompt the UE to start measuring the training CSI-RS; and decode a measurement report including a measurement result for the training CSI-RS based on a signal received from the UE.

[0006] Other example embodiments relate to an apparatus of a base station, the apparatus comprising processing circuitry configured to configure transceiver circuitry to transmit a training sounding reference signal (SRS) configuration to a user equipment (UE), wherein the training SRS configuration comprises a training SRS to be transmitted via each transmit antenna; decode the training SRS based on a signal received from each antenna of the UE; measure the training SRS; and train a CSI compression model based on the training SRS measurements.

[0007] Another example embodiment relates to an apparatus of a user equipment (UE), the apparatus comprising a processing circuit configured to decode a training channel state information reference signal (CSI-RS) configuration including auxiliary information based on a signal received from a base station, wherein the auxiliary information includes an AI tag for a training CSI-RS to be measured; measure the training CSI-RS based on the CSI-RS configuration; and configure a transceiver circuit to transmit a measurement report including a measurement result for the training CSI-RS to the base station.

[0008] Additional example embodiments relate to an apparatus of a user equipment (UE), the apparatus comprising processing circuitry configured to decode a training sounding reference signal (SRS) configuration based on a signal received from a base station, wherein the training SRS configuration includes a training SRS to be transmitted via each transmit antenna, and configuring a transceiver circuit to transmit a training SRS based on the training SRS configuration. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 Example network arrangements are shown according to various example embodiments.

[0010] Figure 2 An example UE is shown according to various example embodiments.

[0011] Figure 3 An example base station is shown according to various example embodiments.

[0012] Figure 4 A first call flow for data collection and joint training at the UE side is shown according to various example embodiments.

[0013] Figure 5 A second call flow for data collection and joint training at the UE side is shown according to various example embodiments.

[0014] Figure 6 A first call flow for data collection and joint training at the network side is shown according to various example embodiments.

[0015] Figure 7 A second call flow for data collection and joint training at the network side is shown according to various example embodiments.

[0016] Figure 8 A first call flow for data collection and joint training at the UE and network side according to various example embodiments is shown.

[0017] Fig. 9A second call flow for data collection and joint training at the UE and network side according to various example embodiments is shown.

[0018] Fig.10 A first call flow for split training is shown that first occurs at the UE according to various example embodiments.

[0019] Fig.11 A second call flow for split training that first occurs at the network is shown according to various example embodiments. DETAILED DESCRIPTION

[0020] Example embodiments may be further understood with reference to the following description and associated drawings, wherein similar elements have the same reference numerals. Example embodiments relate to enhanced data collection operations used in training a channel state information (CSI) compression model.

[0021] The exemplary embodiments are described with respect to UE. However, reference to UE is provided for illustrative purposes only. The exemplary embodiments may be used with any electronic component that can establish a connection with a network and is configured with hardware, software and / or firmware for exchanging information and data with the network. Therefore, the UE described herein is used to represent any electronic component.

[0022] The exemplary embodiments are also described with reference to 5G New Radio (NR) networks. However, it should be understood that the exemplary embodiments may also be implemented in other types of networks, including but not limited to LTE networks, future evolutions of cellular protocols (e.g., 6G networks).

[0023] Throughout the specification, the term "joint training" is used. It should be understood that joint training means that the generation model and the reconstruction model can be trained in the same cycle for both forward propagation and backward propagation. Joint training can be performed at a single node or across multiple nodes. It should also be understood that separate training includes sequential training starting with UE-side training, or sequential training starting with network-side training, or parallel training at both UE and network.

[0024] It should be understood that throughout this description, it will be described that the CSI-RS will be sent by the base station and measured by the UE. However, as described above, the example implementation scheme relates to training CSI compression models. Therefore, the CSI-RS sent and measured is not necessarily the CSI-RS that is usually sent by the base station during normal operation, for example, a normal CSI-RS. That is, the CSI-RS sent may be a separate CSI-RS that is sent and measured for the purpose of training. These CSI-RS may be referred to as training CSI-RS. However, it should be understood that any reference to CSI-RS in the present disclosure may relate to a normal CSI-RS or a training CSI-RS.

[0025] Furthermore, although example embodiments are described with reference to data collection and training of a CSI compression model, example embodiments may be applicable to other types of reference signals with respect to data collection and model training.

[0026] As described above, a CSI compression model may be trained and used to improve the performance of a UE and / or a network. To train a CSI compression model, the UE first performs measurements on a CSI reference signal (CSI-RS). These measurements may then be used to train the CSI compression model. However, CSI-RS measurements may be performed on different CSI-RSs, and these measurements should be classified in a certain way so that the correct model is trained using the CSI-RS measurements. With appropriate classification, different models may be trained. Example implementations may provide information such classification may be used for CSI-RS measurements.

[0027] Example implementations relate to enhancements to data collection in three schemes / types (single entity, dual entity, and split). In a first aspect, type 1 single entity may be understood to facilitate UE-side training, validation, and testing of AI / ML models (version 1), or network-side training, validation, and testing of AI / ML models (version 2). In a second aspect, type 2 joint training of dual-side models on both the network side and the UE side is described. In a third aspect, type 3 split training is described. Each type of enhancement is described in more detail below.

[0028] Figure 1 An example network arrangement 100 according to various example embodiments is shown. The example network arrangement 100 includes a UE 110. Those skilled in the art will appreciate that the UE 110 may be any type of electronic component configured to communicate via a network, such as a mobile phone, a tablet computer, a desktop computer, a smart phone, a phablet, an embedded device, a wearable device, an Internet of Things (IoT) device, etc. It should also be appreciated that an actual network arrangement may include any number of UEs used by any number of users. Therefore, the example of a single UE 110 is provided for illustrative purposes only.

[0029] UE 110 may be configured to communicate with one or more networks. In the example of network configuration 100, the network with which UE 110 may wirelessly communicate is a 5G NR radio access network (RAN) 120. However, it should be understood that UE 110 may also communicate with other types of networks (e.g., 5G cloud RAN, next generation RAN (NG-RAN), traditional cellular networks, etc.), and UE 110 may also communicate with the network via a wired connection. Referring to the exemplary embodiment, UE 110 may establish a connection with 5G NR RAN 120. Therefore, UE 110 may have a 5G NR chipset to communicate with NR RAN 120.

[0030] The 5G NR RAN 120 may be part of a cellular network that may be deployed by a network operator (e.g., Verizon, AT&T, T-Mobile, etc.). The RAN 120 may include cells or base stations that are configured to transmit and receive traffic from UEs equipped with appropriate cellular chipsets. In this example, the 5G NR RAN 120 includes a gNB 120A. However, reference to a gNB is provided for illustrative purposes only, and any appropriate base station or cell (e.g., Node B, eNodeB, HeNB, eNB, gNB, gNodeB, macro cell, micro cell, small cell, femto cell, etc.) may be deployed.

[0031] Those skilled in the art will appreciate that any relevant process may be performed for UE 110 to connect to 5G NR RAN 120. For example, as described above, 5G NR RAN 120 may be associated with a particular network operator at which UE 110 and / or its user has protocol and credential information (e.g., stored on a SIM card). Upon detecting the presence of 5G NR RAN 120, UE 110 may send corresponding credential information in order to associate with 5G NR RAN 120. More specifically, UE 110 may be associated with a particular cell (e.g., gNB 120A).

[0032] The network arrangement 100 also includes a cellular core network 130, the Internet 140, an IP multimedia subsystem (IMS) 150, and a network service backbone 160. The cellular core network 130 manages traffic flowing between the cellular network and the Internet 140. The IMS 150 can be generally described as an architecture for delivering multimedia services to the UE 110 using IP protocols. The IMS 150 can communicate with the cellular core network 130 and the Internet 140 to provide multimedia services to the UE 110. The network service backbone 160 communicates directly or indirectly with the Internet 140 and the cellular core network 130. The network service backbone 160 can be generally described as a collection of components (e.g., servers, network storage arrangements, etc.) that implement a set of services that can be used to extend the functionality of the UE 110 to communicate with various networks.

[0033] Figure 2 An example UE 110 is shown according to various example embodiments. The UE 110 will refer to Figure 1100. UE 110 may represent any electronic device and may include a processor 205, a memory arrangement 210, a display device 215, an input / output (I / O) device 220, a transceiver 225, and other components 230. Other components 230 may include, for example, an audio input device, an audio output device, a battery providing a limited power source, a data acquisition device, a port for electrically connecting UE 110 to other electronic devices, a sensor for detecting the status of UE 110, and the like.

[0034] The processor 205 may be configured to execute multiple engines of the UE 110. For example, the engines may include a CSI-RS training data engine 235 for performing operations such as performing measurements on a CSI-RS or SRS and running / generating an AI / ML model on the measured or received CSI-RS or SRS.

[0035] The above-described engine as an application (e.g., program) executed by the processor 205 is merely exemplary. The functionality associated with the engine may also be represented as a separate combined component of the UE 110, or may be a modular component coupled to the UE 110, such as an integrated circuit with or without firmware. For example, the integrated circuit may include an input circuit for receiving a signal and a processing circuit for processing the signal and other information. The engine may also be embodied as one application or multiple separate applications. In addition, in some UEs, the functionality described for the processor 205 is split between two or more processors (such as a baseband processor and an application processor). The example implementation scheme may be implemented in any of these or other configurations of the UE.

[0036] Memory arrangement 210 may be a hardware component configured to store data related to operations performed by UE 110. Display device 215 may be a hardware component configured to display data to a user, and I / O device 220 may be a hardware component that enables a user to enter input. Display device 215 and I / O device 220 may be separate components or may be integrated together (such as a touch screen).

[0037] The transceiver 225 may be a hardware component configured to establish a connection with the 5G-NR RAN 120. Thus, the transceiver 225 may operate on a variety of different frequencies or channels (e.g., a continuous set of frequencies). The transceiver 225 includes a circuit configured to send and / or receive signals (e.g., a control signal, a data signal). Such signals may be encoded with information for implementing any of the methods described herein. The processor 205 may be operably coupled to the transceiver 225 and configured to receive signals from and / or send signals to the transceiver 225. The processor 205 may be configured to encode and / or decode signals (e.g., signaling from a base station of a network) for implementing any of the methods described herein.

[0038] Figure 3 An example base station 300 is shown according to various example embodiments. Base station 300 may represent a gNB 120A or any other access node that a UE 110 may use to establish a connection and manage network operations.

[0039] The base station 300 may include a processor 305, a memory arrangement 310, an input / output (I / O) device 315, a transceiver 320, and other components 325. These other components 325 may include, for example, an audio input device, an audio output device, a battery, a data acquisition device, ports for electrically connecting the base station 300 to other electronic devices and / or a power source, and the like.

[0040] The processor 305 may be configured to execute multiple engines of the UE 110. For example, the engines may include a CSI-RS training data engine 330 for performing operations such as performing measurements on a CSI-RS or SRS and running / generating an AI / ML model on the measured or received CSI-RS or SRS.

[0041] The memory 310 may be a hardware component configured to store data related to operations performed by the base station 300. The I / O device 315 may be a hardware component or port that enables a user to interact with the base station 300.

[0042] The transceiver 320 may be a hardware component configured to exchange data with the UE 110 and any other UE in the network arrangement 100. The transceiver 320 may operate on a variety of different frequencies or channels (e.g., a continuous frequency set). Therefore, the transceiver 320 may include one or more components (e.g., a radio device) to enable data exchange with various networks and UEs. The transceiver 320 includes a circuit configured to send and / or receive signals (e.g., a control signal, a data signal). Such signals may be encoded with information that implements any of the methods described herein. The processor 305 may be operably coupled to the transceiver 320 and configured to receive signals from the transceiver 320 and / or send signals to the transceiver. The processor 305 may be configured to encode and / or decode signals (e.g., signaling from the UE) for implementing any of the methods described herein.

[0043] As described above, in the first aspect, a Type 1 single entity may be understood as facilitating UE-side training, validation and testing of AI / ML models (version 1), or network-side training, validation and testing of AI / ML models (version 2).

[0044] In a first version of the first aspect of the example embodiments, enhanced data collection for type 1 training at the UE side is disclosed.

[0045] Figure 4 A first call flow for data collection and joint training at the UE side is shown according to various example embodiments. As described above, when CSI-RS measurements are used to train a model, it may be useful to classify the CSI-RS measurements. In a first option of the first version, when the CSI-RS is configured for training purposes, the network may provide auxiliary information to the UE 110. The auxiliary information may be an AI set identifier that is added to the CSI-RS configuration to assist in data collection, for example, the network provides an AI tag for the CSI-RS set for the UE 110 to appropriately store the CSI-RS measurements based on the AI ​​tag. In some example embodiments, the AI ​​tag may be an AI model identifier (ID), where data collected using the same model ID will be used to train the same CSI compression model.

[0046] The classification (e.g., AI label) may be based on any factor related to the model being trained determined by the network. For example, the network may classify the CSI-RS transmitted by the cell by the antenna or combination of antennas of the cell, the frequency band (or frequency bands) of the CSI-RS, the transmit power of the CSI-RS, etc. To provide a simple non-limiting example, there may be two CSI compression models being trained, a first CSI compression model associated with a first cell and a second CSI compression model associated with a second cell. The network may transmit a CSI-RS configuration to the UE 110, the CSI-RS configuration including the CSI-RS transmitted by the first cell with the first AI label ID and the CSI-RS transmitted by the second cell with the second AI label ID. The UE 110 may perform CSI-RS measurements and store the measurements according to the AI ​​label so that the corresponding model can be trained using different CSI-RS measurements. It should be understood that the classification (e.g., AI label) based on the cell transmitting the CSI-RS is only an example, and any type of classification may be used. Other examples of classification / labels may include indoor / outdoor / rural / urban, specific network ports and channel configurations, etc. Typically, each of these different classifications may result in different channel statistics and may therefore be associated with a different CSI compression model.

[0047] At 410, the network (e.g., gNB 120A) sends a CSI-RS set configuration with an AI tag ID to UE 110. As described above, the AI ​​tag ID enables UE 110 to more effectively classify the measured CSI-RS. At 420, UE 110 performs CSI-RS measurements. At 430, UE 110 buffers CSI-RS measurements based on the AI ​​tag received from gNB 120A.

[0048] As described above, in these example embodiments, data is collected for Type 1 training at the UE side. Thus, the UE 110 may then use the CSI-RS measurements with the AI ​​tags to train one or more CSI compression models.

[0049] Figure 5A second call flow for data collection and joint training at the UE side according to various example embodiments is shown. In a second option of the first version of the first aspect, the network (e.g., gNB 120A) may indicate the start and stop of data collection and provide the UE 110 with a tag ID for measurement. The UE 110 may then buffer the collected results with the cell ID. The network may provide the UE 110 with auxiliary information on how to aggregate different data sets based on both the cell ID and the tag ID. To provide some further examples of classification, different data sets measured with different gNB vendors may not be grouped together. Similarly, different antenna panel configurations and virtualization schemes may not be grouped together. In these cases, different tag IDs may be assigned.

[0050] It should be understood that multiple UEs may provide measurement results to a UE-side server, which is used to aggregate data sets from different UEs and then train the model. Therefore, for the purpose of aggregation and model training, the UE-side server may receive auxiliary information.

[0051] At 510, gNB 120A sends a trigger to UE 110 to start data collection (i.e., measurements). Trigger 510 may also include a tag for UE 110 to use in classifying the measurements for further AI / ML training. At 520, UE 110 performs CSI-RS measurements. At 530, UE 110 buffers the CSI-RS measurements based on the received tag. At 540, the gNB sends a trigger to the UE to stop data collection.

[0052] As described above, in these example embodiments, data is collected for Type 1 training at the UE side. Thus, the UE 110 may then use the tagged CSI-RS measurements to train one or more CSI compression models corresponding to the classification associated with each tag.

[0053] In a second version of the first aspect of the example implementation, enhanced data collection for type 1 training at the network side is disclosed. In the second version, the network may perform joint training instead of the UE, as described with respect to the first version and Figures 4 to 5 Those skilled in the art will appreciate that training the model on the network side may provide greater computing power for model training than on the UE side, and may also save battery power of the UE because the UE does not need to perform computations.

[0054] Figure 6A first call flow for data collection and joint training at the network side is shown according to various example embodiments. In 610, the gNB 120A may send a trigger to the UE 110 to start data collection. The start trigger 610 may also include a CSI-RS set configuration. The CSI-RS set configuration of the start trigger 610 may configure a specific CSI-RS set for AI / ML data collection purposes. The gNB 120A may also configure a data format (e.g., number of bits per value, channel, eigenvectors, etc.) for data collection.

[0055] In 620, UE 110 performs CSI-RS measurements based on the received configuration. In some example embodiments, in 630, the UE buffers the measurements based on the received tags. The buffered CSI-RS measurements may be represented by different quantizations, such as 32-bit floating point, 16 bits, etc. In 640, gNB 120A sends a trigger for UE 110 to stop data collection / measurement. In 650, UE 110 sends the buffered CSI-RS measurements as a data payload to the gNB via a physical uplink shared channel (PUSCH).

[0056] In other example embodiments, after 620, UE 110 may omit buffering operation 630. Instead, UE 110 may send the measurement results as uplink control information (UCI) for each measurement (eg, continuously sending back the measurements as they are created).

[0057] As described above, in these example embodiments, data is collected for Type 1 training at the network side. Thus, the gNB 120A (or other network component) may then use the labeled CSI-RS measurements to train one or more CSI compression models corresponding to the classification associated with each label.

[0058] Figure 7 A second call flow for data collection and joint training at the network side according to various example embodiments is shown. In a second option of the second version of the first aspect of the example embodiment, the network may transmit a sounding reference signal (SRS) configuration to the UE 110 for data collection purposes. It should be understood that due to the correlation between the measured SRS and CSI-RS, the SRS configuration is still used to train the CSI compression model.

[0059] At 710, gNB 120A sends an SRS configuration to UE 110. It should be understood that similar to the discussion above for normal and training CSI-RS, this SRS configuration may differ in one or more characteristics compared to the "normal" SRS. The "training" SRS may be a separate configuration for data collection purposes, have a longer SRS periodicity, be capable of both time division duplex (TDD) and frequency division duplex (FDD), etc. UE 110 may transmit the SRS via each transmit antenna of the UE to ensure phase coherence between the transmit antennas used for each SRS transmission.

[0060] In 720, UE 110 transmits an SRS to gNB 120A for each of its antenna ports according to the received SRS configuration 710. Although four transmissions are shown for 720, this is only an example and other numbers of antennas are possible. In 730, gNB 120A measures the SRS received from UE 110.

[0061] As described above, in these example embodiments, data is collected for type 1 training at the network side. Thus, gNB 120A (or other network components) can then use SRS measurements to train one or more CSI compression models. As described above, there may be a correlation in the channel between SRS measurements and CSI-RS measurements. This correlation can be used when training the CSI compression model.

[0062] Now turning to Type 2: Joint training of dual-sided models at both the network side and the UE side, the second aspect of the example implementation is described. The second aspect relates to a scenario where the UE and the network train an AI model together. In the Type 2 scenario, model information is passed between the UE and the network during each iteration. The training data has two parts: an input data set used by the UE; and an output data set used by the network and a setting loss function.

[0063] Figure 8 A first call flow for data collection and joint training at the UE side and the network side according to various example embodiments is shown. Figure 8 It may be understood as describing the first option of the second aspect. In the first option, the UE 110 collects measurements and transmits a labeled CSI-RS output data set to the network for training.

[0064] In 810, gNB 120A transmits auxiliary information for collecting training data to UE 110. The auxiliary information 810 may include an AI tag and / or a CRI-RS configuration. The auxiliary information 810 may be, for example, the auxiliary information transmitted in 410 or 510, as described above with reference to Figure 4 and Figure 5 As described.

[0065] In 820, UE 110 performs CSI-RS measurements. In 830, UE 110 buffers CSI-RS measurements based on the tags received from 810. In 840, UE 110 performs pre-processing of the CSI-RS measurements. This processing may be understood as representing the UE side of the two-side training inherent to the type 2 scenario. In this example, the pre-processing may include marking each CSI-RS measurement result in the CSI-RS measurement results to be transmitted to the network.

[0066] At 860, UE 110 sends the labeled output data set to gNB 120A. Transmission 860 may be the pre-processed measurements 840, the principal eigenvectors of the measurement results, or some combination of the foregoing.

[0067] As mentioned above, the network will receive the labeled output dataset and apply the set loss function to further train the CSI compression model.

[0068] Fig. 9 A second call flow for data collection and joint training at the UE and network side according to various example embodiments is shown. Fig. 9 This may be understood as describing a second option of the second aspect. In the second option, the network may collect measured CSI-RS from UE 110 and send a labeled CSI-RS input data set back to UE 110 for training. Figure 6 The CSI-RS collection scheme described in the Fig. 9 However, it should be appreciated that the CSI-RS need not be buffered (e.g., measurements may be sent back simultaneously), or that the UE 110 may instead send an SRS to the gNB, as described with respect to Figure 7 As described.

[0069] like Figure 6 As shown, operations 910-950 are performed in a manner substantially similar to that described for operations 610-650, respectively. At 960, gNB 120A generates an input labeled data set for training at UE 110. At 970, gNB 120A sends the labeled data set to UE 110. UE 110 may then use the labeled data set for CSI compression model training.

[0070] In some example embodiments, UE 110 may represent a UE-side server, where data from multiple UEs may be aggregated to generate a synthetic dataset and output a label.

[0071] Turning to the third aspect of the example implementation, an improved means of handling data collection in a type 3 split training scenario is disclosed. In the third aspect, the UE may perform joint training (encoding / decoding) on ​​a data set. The UE may perform encoding / decoding model training offline to generate an intermediate data set. The intermediate data set may be understood as the encoder / decoder output after running through the training model. The UE may then send the intermediate data set to the network for the purpose of model training.

[0072] Fig.10 A first call flow for split training is shown that first occurs at the UE according to various example embodiments. Fig.10 It can be understood as describing the first option of the third aspect. The UE can be connected with Figures 4 to 5 The training auxiliary information is collected in a manner substantially similar to that described in . Fig.10 Included with Figure 4 The operations are substantially similar to those described in (ie, 1010-1030 correspond to 410-430), but it should be understood that any of the previously described operations for CSI-RS set configuration or SRS configuration transmitted to the UE may be used alternatively.

[0073] At 1040, UE 110 performs offline (e.g., without assistance from the network) encoding and decoding model training. At 1050, gNB 120A sends a CSI-RS set configuration for intermediate data set generation. For clarity, at 1050, UE 110 has performed offline modeling. After receiving CSI-RS set configuration 1050, UE 110 may further operate on the modeled data set.

[0074] In 1060, the UE 110 generates a labeled encoder / decoder output data set using the received CSI-RS set configuration 1050 and the offline modeled data set 1040. The labeled data set 1060 can also be understood as an intermediate data set. The labeled data set can be used by the network for training purposes.

[0075] At 1070, UE 110 sends the intermediate data set to gNB 120A. UE 110 may send the intermediate data set as a buffered payload via PUSCH, or UE 110 may send per-CSI-RS measurements as UCI to gNB 120A.

[0076] There are possible additional options for operation 1070. UE 110 may feature a neural network (NN) encoder and quantizer, both of which are included in the labeled encoded output. gNB 120A (network) may feature a NN decoder and a dequantizer. The NN encoder may be used as the NN output, the quantizer may be used as the quantizer output, and the NN decoder may be used as the decoder output. Those skilled in the art will recognize that these inputs and outputs may be arranged in several ways to train the model. For example, the NN output may be used as the input to the model, and the quantizer output may be used as the desired output of the model.

[0077] Fig.11 A second call flow for split training that first occurs at the network is shown according to various example embodiments. Fig.11 It can be understood as describing the second option of the third aspect. It should be understood that the network can be connected to Figure 6 to Figure 7 The training auxiliary information is collected in a manner substantially similar to that described in . Fig.11 Included with Figure 6 The operations are substantially similar to those described in (ie, 1110-1150 correspond to 410-450), but it should be understood that any of the previously described operations for CSI-RS set configuration or SRS configuration transmitted to the UE may be used alternatively.

[0078] At 1160, gNB 120A performs offline model training based on the CSI-RS measurements received from UE 110. At 1170, gNB 120A generates a labeled encoder / decoder output data set (e.g., an intermediate data set). At 1180, gNB 120A sends the entire intermediate data set to UE 110, including the labeled encoder input and the labeled encoder output for UE-side encoder training.

[0079] Example

[0080] In a first embodiment, a method performed by a base station includes: transmitting a training channel state information reference signal (CSI-RS) configuration including auxiliary information to a user equipment (UE), wherein the auxiliary information includes an AI tag for a training CSI-RS to be measured; transmitting a start indication to the UE to prompt the UE to start measuring the training CSI-RS; and receiving a measurement report including a measurement result for the training CSI-RS from the UE.

[0081] In a second embodiment, the method according to the first embodiment further comprises transmitting a stop indication to the UE to prompt the UE to stop measuring the training CSI-RS.

[0082] In a third embodiment, according to the method of the second embodiment, the measurement report is received after the stop indication is transmitted.

[0083] In a fourth embodiment, the method according to the third embodiment, wherein the measurement report is received via a physical uplink shared channel (PUSCH).

[0084] In a fifth embodiment, the method according to the third embodiment, wherein the measurement report comprises a plurality of measurement reports received prior to transmitting the stop indication.

[0085] In a sixth embodiment, the method according to the fifth embodiment, wherein the measurement report is received via uplink control information (UCI).

[0086] In a seventh embodiment, according to the method of the first embodiment, the training CSI-RS configuration and the start indication are sent via the same message.

[0087] In an eighth embodiment, according to the method of the first embodiment, the auxiliary information further includes the number of bits per value of the CSI-RS, a channel, or an eigenvector.

[0088] In a ninth embodiment, according to the method of the first embodiment, the measurement report also includes a labeled output data set based at least on the CSI-RS measurement and the AI ​​tag, and the method further includes training a CSI compression model based on the labeled output data set.

[0089] In a tenth embodiment, the method according to the first embodiment further includes: generating a labeled input data set based at least on the measurement result and the AI ​​tag; and sending the labeled input data set to the UE.

[0090] In an eleventh embodiment, the method according to the first embodiment also includes: training an encoder model and a decoder model for processing CSI-RS; using the encoder model and the decoder model to process the measurement results to generate a labeled data set for the measurement results; and sending the labeled data set, the encoder input and the decoder output to the UE.

[0091] In a twelfth embodiment, according to the method of the first embodiment, the measurement result also includes a labeled data set based at least on the AI ​​label, an encoder input corresponding to the AI ​​label, and a decoder output corresponding to the AI ​​label.

[0092] In a thirteenth embodiment, according to the method of the first embodiment, the AI ​​tag includes an AI model identification (ID).

[0093] In a fourteenth embodiment, a processor is provided, wherein the processor is configured to execute any of the methods according to the first to thirteenth embodiments.

[0094] In a fifteenth embodiment, a base station comprises: a transceiver configured to communicate with a user equipment (UE); and a processor communicatively coupled to the transceiver and configured to perform any of the methods described in the first to thirteenth embodiments.

[0095] In a sixteenth embodiment, a method performed by a base station, the method comprising transmitting a training sounding reference signal (SRS) configuration to a user equipment (UE), wherein the training SRS configuration comprises a training SRS to be transmitted via each transmitting antenna; receiving a training SRS from each antenna of the UE; measuring the training SRS and training a CSI compression model based on the training SRS measurement.

[0096] In a seventeenth embodiment, according to the method of the sixteenth embodiment, the training SRS configuration comprises a time division duplex (TDD) configuration or a frequency division duplex (FDD) configuration.

[0097] In an eighteenth embodiment, a processor is provided, wherein the processor is configured to execute any of the methods according to the sixteenth to seventeenth embodiments.

[0098] In a nineteenth embodiment, a base station comprises: a transceiver configured to communicate with a user equipment (UE); and a processor communicatively coupled to the transceiver and configured to perform any of the methods described in the sixteenth to seventeenth embodiments.

[0099] In a twentieth embodiment, a method performed by a user equipment (UE), the method comprising: receiving a training channel state information reference signal (CSI-RS) configuration including auxiliary information, wherein the auxiliary information includes an AI tag for a training CSI-RS to be measured; measuring the training CSI-RS based on the CSI-RS configuration; and transmitting a measurement report including measurement results for the training CSI-RS to a network.

[0100] In a twenty-first embodiment, the method according to the twentieth embodiment further includes receiving a start indication from the network to start measuring the training CSI-RS.

[0101] In a twenty-second embodiment, according to the method of the twenty-first embodiment, the training CSI-RS configuration and the start indication are received in the same message.

[0102] In a twenty-third embodiment, according to the method of the twenty-first embodiment, the method further comprises receiving a stop instruction from the network to stop measuring the training CSI-RS.

[0103] In a twenty-fourth embodiment, according to the method of the twenty-third embodiment, the method further comprises buffering the measurement results for the training CSI-RS based on the AI ​​tag, wherein the measurement report is transmitted after receiving the stop indication.

[0104] In a twenty-fifth embodiment, the method according to the twenty-fourth embodiment, wherein the measurement report is transmitted via a physical uplink shared channel (PUSCH).

[0105] In a twenty-sixth embodiment, the method according to the twenty-fourth embodiment, wherein the measurement report includes a plurality of measurement reports transmitted before receiving the stop indication.

[0106] In a twenty-seventh embodiment, the method of the twenty-sixth embodiment, wherein the measurement report is transmitted via uplink control information (UCI).

[0107] In a twenty-eighth embodiment, according to the method of the twentieth embodiment, the auxiliary information further includes the number of bits per value of the CSI-RS, a channel, or an eigenvector.

[0108] In a twenty-ninth embodiment, according to the method of the twentieth embodiment, the method further includes generating a labeled output data set based at least on the measurement results and the AI ​​tags, and sending the labeled output data set to the network.

[0109] In a thirtieth embodiment, the method according to the twentieth embodiment also includes receiving a labeled output data set from the network based at least on the measurement results and the AI ​​label, and training a CSI compression model based on the labeled output data set.

[0110] In a thirty-first embodiment, the method according to the twentieth embodiment also includes: training an encoder model and a decoder model for processing CSI-RS; using the encoder model and the decoder model to process the measurement results to generate a labeled data set based at least on the measurement results and the AI ​​label; and sending the labeled data set, the encoder input, and the decoder output to the network.

[0111] In a thirty-second embodiment, the method according to the twentieth embodiment also includes receiving a labeled data set, an encoder input, and a decoder output, and training a CSI compression model based on the labeled data set, the encoder input, and the decoder output.

[0112] In a thirty-third embodiment, according to the method of the twentieth embodiment, the AI ​​tag includes an AI model identification (ID).

[0113] In a thirty-fourth embodiment, a processor is configured to perform any of the methods described in accordance with the twentieth to thirty-third embodiments.

[0114] In a thirty-fifth embodiment, a user equipment (UE) comprises: a transceiver configured to communicate with a base station; and a processor communicatively coupled to the transceiver and configured to perform any of the methods described in the twentieth to thirty-third embodiments.

[0115] In a thirty-sixth embodiment, a method performed by a user equipment (UE), the method comprising receiving a training sounding reference signal (SRS) configuration from a network, wherein the training SRS configuration comprises training an SRS to be sent via each transmitting antenna and sending the training SRS based on the training SRS configuration.

[0116] In a thirty-seventh embodiment, according to the method of the thirty-sixth embodiment, the training SRS configuration comprises a time division duplex (TDD) configuration or a frequency division duplex (FDD) configuration.

[0117] In a thirty-eighth embodiment, a processor is configured to perform any of the methods described in embodiments thirty-six to thirty-seven.

[0118] In a thirty-ninth embodiment, a user equipment (UE) comprises: a transceiver configured to communicate with a base station; and a processor communicatively coupled to the transceiver and configured to perform any of the methods described in the thirty-sixth to thirty-seventh embodiments.

[0119] Those skilled in the art will appreciate that the example embodiments described above may be implemented with any suitable software configuration or hardware configuration or combination thereof. Example hardware platforms for implementing the example embodiments may include, for example, Intel x86-based platforms with compatible operating systems, Windows OS, Mac platforms and MAC OS, mobile devices with operating systems such as iOS, Android, etc. In another example, the example embodiments of the above method may be embodied as a program including lines of code stored on a non-transitory computer-readable storage medium, which, when compiled, may be executed on a processor or microprocessor.

[0120] Although the present application describes various combinations of aspects each having different features, those skilled in the art will understand that any feature of one aspect may be combined with features of other aspects or features that are not functionally or logically inconsistent with the operation or function of the device of the aspects disclosed in the present invention in any manner not publicly denied.

[0121] It is understood that the use of personally identifiable information should be subject to privacy policies and practices that are generally recognized to meet or exceed industry or government requirements for maintaining user privacy. Specifically, personally identifiable information data should be managed and processed to minimize the risk of unintentional or unauthorized access or use, and the nature of the authorized use should be clearly stated to users.

[0122] It will be apparent to those skilled in the art that various modifications may be made to the present disclosure without departing from the spirit or scope of the present disclosure. Therefore, it is intended that the present disclosure covers modifications and variations of the present disclosure as long as they fall within the scope of the appended claims and their equivalents.

Claims

1. A base station device, the device comprising a processing circuit, the processing circuit being configured to: configuring the transceiver circuitry to transmit a training channel state information reference signal (CSI-RS) configuration including assistance information to a user equipment (UE), wherein the assistance information includes an AI tag for a training CSI-RS to be measured; The transceiver circuit is configured to transmit a start indication to the UE to prompt the UE to start measuring the training CSI-RS; as well as A measurement report including a measurement result for the training CSI-RS is decoded based on a signal received from the UE.

2. The apparatus according to claim 1, wherein the processing circuit is further configured to: The transceiver circuit is configured to transmit a stop indication to the UE to prompt the UE to stop measuring the training CSI-RS. The apparatus of claim 2 , wherein the measurement report is received after the transmitting the stop indication. 4 . The apparatus of claim 3 , wherein the measurement report comprises a plurality of measurement reports received prior to the transmitting the stop indication. 5 . The apparatus of claim 1 , wherein the training CSI-RS configuration and the start indication are sent via the same message. 6 . The apparatus according to claim 1 , wherein the auxiliary information further includes an eigenvector, a channel, or a number of bits per value of the CSI-RS.

7. The apparatus of claim 1 , wherein the measurement report further comprises a labeled output data set based at least on the CSI-RS measurement and an AI tag, wherein the processing circuit is further configured to: A CSI compression model is trained based on the labeled output dataset.

8. The apparatus of claim 1 , wherein the processing circuit is further configured to: generating a labeled input data set based at least on the measurements and the AI ​​labels; and The transceiver circuitry is configured to transmit the marked input data set to the UE.

9. The apparatus of claim 1 , wherein the processing circuit is further configured to: Train the encoder and decoder models for processing CSI-RS; processing the measurements using the encoder model and the decoder model to generate a labeled dataset for the measurements; and Transceiver circuitry is configured to transmit the marked data set, the encoder input, and the decoder output to the UE.

10. The apparatus of claim 1, wherein the measurement result further comprises a labeled data set based at least on the AI ​​label, an encoder input corresponding to the AI ​​label, and a decoder output corresponding to the AI ​​label.

11. A base station device, the device comprising a processing circuit, the processing circuit being configured to: configuring the transceiver circuitry to transmit a training sounding reference signal (SRS) configuration to a user equipment (UE), wherein the training SRS configuration comprises a training SRS to be transmitted via each transmit antenna, wherein the training SRS configuration comprises a time division duplex (TDD) configuration or a frequency division duplex (FDD) configuration; decoding a training SRS based on a signal received from each antenna of the UE; measuring the training SRS; as well as A CSI compression model is trained based on the training SRS measurements.

12. An apparatus of a user equipment (UE), the apparatus comprising a processing circuit, the processing circuit being configured to: decoding a training channel state information reference signal (CSI-RS) configuration including auxiliary information based on a signal received from a base station, wherein the auxiliary information includes an AI tag for a training CSI-RS to be measured; measuring the training CSI-RS based on the CSI-RS configuration; as well as The transceiver circuit is configured to transmit a measurement report including a measurement result for the training CSI-RS to the base station.

13. The apparatus of claim 12, wherein the processing circuit is further configured to: A start indication for starting to measure the training CSI-RS is decoded based on a signal received from the base station.

14. The apparatus of claim 13, wherein the training CSI-RS configuration and the start indication are received in the same message.

15. The apparatus of claim 13, wherein the processing circuit is further configured to: A stop indication for stopping measuring the training CSI-RS is decoded based on a signal received from the base station.

16. The apparatus of claim 15, wherein the processing circuit is further configured to: The measurement results for the training CSI-RS are buffered based on the AI ​​tag, wherein the measurement report is transmitted after receiving the stop indication.

17. The apparatus of claim 15, wherein the measurement report comprises a plurality of measurement reports transmitted prior to receiving the stop indication.

18. The apparatus according to claim 12, wherein the auxiliary information further includes an eigenvector, a channel, or a number of bits per value of the CSI-RS.

19. The apparatus of claim 12, wherein the processing circuit is further configured to: generating a labeled output data set based at least on the measurements and the AI ​​labels; and The transceiver circuit is configured to transmit the labeled output data set to the base station.

20. The apparatus of claim 12, wherein the processing circuit is further configured to: decoding a labeled output data set based on at least the measurements and the AI ​​tags based on a signal received from the base station; and A CSI compression model is trained based on the labeled output dataset.

Images