Model pairing for ai / ml based csi compression

Abstract

Various aspects of the present disclosure relate to user equipment, base stations, processors, and methods for model pairing for AI / ML based CSI compression. In one aspect, a user equipment (UE) receives a configuration for a first codebook. The UE transmits a first precoding matrix indicator (PMI) obtained based on the first codebook and a first channel state information (CSI) estimated by the UE, and at least one second PMI obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

Description

Technical Field

[0001] This disclosure relates to wireless communications, and more specifically to user equipment (UE), base stations, processors, and methods for model pairing for channel state information (CSI) compression based on artificial intelligence or machine learning (AI / ML). Background Technology

[0002] A wireless communication system may include one or more network communication devices, such as base stations, which may also be referred to as eNodeBs (eNBs), next-generation NodeBs (gNBs), or other suitable terms. Each network communication device (such as a base station) may support wireless communication with one or more user communication devices, which may also be referred to as user equipment (UEs), or other suitable terms. The wireless communication system may support wireless communication with one or more user communication devices by utilizing the resources of the wireless communication system (e.g., time resources (e.g., symbols, time slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers)). Additionally, the wireless communication system may support wireless communication across a variety of radio access technologies, including third-generation (3G) radio access technology, fourth-generation (4G) radio access technology, fifth-generation (5G) radio access technology, and other suitable radio access technologies other than 5G (e.g., sixth-generation (6G)).

[0003] CSI compression using two-sided models (such as artificial intelligence (AI) / machine learning (ML) models) has been introduced. In most use cases, CSI compression using two-sided models can provide performance improvements. However, some unresolved issues related to CSI compression using two-sided models remain that require further investigation in the future. Summary of the Invention

[0004] This disclosure relates to methods, apparatus, and systems for supporting model pairing for AI / ML-based CSI compression.

[0005] In a first aspect of the solution, a user equipment (UE) may include: a processor; and a transceiver coupled to the processor, wherein the processor is configured to: receive configuration about a first codebook from a base station via the transceiver; and transmit a first precoding matrix indicator (PMI) and at least one second PMI to the base station via the transceiver, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

[0006] In some implementations of the methods and apparatus described herein, at least one first AI / ML model may be associated with CSI feedback compression.

[0007] In some implementations of the methods and apparatus described herein, sending a first PMI and at least one second PMI may include: sending a request to a base station via a transceiver for determining whether a first AI / ML model in at least one first AI / ML model can be paired with a second AI / ML model of the base station; and sending a first PMI and a second PMI to the base station via a transceiver, the second PMI being associated with a first AI / ML model in at least one first AI / ML model.

[0008] In some implementations of the methods and apparatus described herein, the processor may also be configured to receive, via a transceiver, the pairing result of the second AI / ML model with the first AI / ML model from a base station.

[0009] In some implementations of the methods and apparatus described herein, sending a first PMI and at least one second PMI may include: receiving from a base station via a transceiver a request to determine whether a second AI / ML model of the base station can be paired with an active first AI / ML model in at least one first AI / ML model; and in response to the request, sending the first PMI and the second PMI, associated with the active first AI / ML model, to the base station via the transceiver.

[0010] In some implementations of the methods and apparatus described herein, the processor may also be configured to: receive an indication from the base station via a transceiver if the base station determines that the second AI / ML model is not paired with the activated first AI / ML model, the indication being used to deactivate the activated first AI / ML model if the base station determines that the second AI / ML model is not paired with the activated first AI / ML model.

[0011] In some implementations of the methods and apparatus described herein, at least one first AI / ML model may include multiple first AI / ML models, and at least one second PMI may include multiple second PMIs associated with the multiple first AI / ML models. Sending the first PMI and at least one second PMI may include: sending a request to a base station via a transceiver for selecting a first AI / ML model from the multiple first AI / ML models; and sending the first PMI and the multiple second PMIs to the base station via a transceiver.

[0012] In some implementations of the methods and apparatus described herein, the processor may also be configured to: receive from a base station via a transceiver an indication of the selection result, indicating the model identifier (ID) of the selected first AI / ML model when the first AI / ML model is selected by the base station from a plurality of first AI / ML models, and indicating a pairing failure when no first AI / ML model is selected by the base station from a plurality of first AI / ML models.

[0013] In some implementations of the methods and apparatus described herein, the processor may also be configured to: transmit via a transceiver to a base station an indication of the number of a plurality of first AI / ML models, and receive via a transceiver from the base station a CSI report configuration and a CSI reference signal (RS) configuration determined based on the indication of the plurality of first AI / ML models.

[0014] In some implementations of the methods and apparatus described herein, the configuration of the first codebook may be based on at least one AI / ML model or function at the UE, which may be associated with one or more AI / ML models.

[0015] In some implementations of the methods and apparatus described herein, the configuration of the first codebook may include scaling factors for parameters associated with a second codebook, which is indicated in a codebook configuration transmitted from a base station.

[0016] In some implementations of the methods and apparatus described herein, the configuration of the first codebook may be transmitted as function / model-related information during a function or model identification process between the UE and the base station, or the configuration of the first codebook may be transmitted via radio resource control (RRC) signaling.

[0017] In some implementations of the methods and apparatus described herein, parameters may include at least one of the following: the number of beams; the number of phase quantization sizes; the number of oversamples; the number of beam amplitude scaling factors for both broadband and subband; or the number of beam combination coefficients or phases in the beams, polarizations, and layers.

[0018] In some implementations of the methods and apparatus described herein, transmitting a first PMI and at least one second PMI may include transmitting the first PMI and at least one second PMI together via one of the following: uplink control information (UCI), media access control (MAC) control unit (CE), or radio resource control (RRC).

[0019] In some implementations of the methods and apparatus described herein, the CSI or eigenvalues ​​(EVs) of the channel matrix constructed based on the first codebook can be the input to at least one first AI / ML model.

[0020] In a second aspect of the solution, a base station may include: a processor; and a transceiver coupled to the processor, wherein the processor is configured to: transmit configuration about a first codebook to a user equipment (UE) via the transceiver; receive a first precoding matrix indicator (PMI) and at least one second PMI from the UE via the transceiver, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model; and determine whether a second AI / ML model of the base station can be paired with at least one first AI / ML model based on a third CSI and at least one fourth CSI, the third CSI being obtained based on the first codebook and the received first PMI, the at least one fourth CSI being obtained based on the second AI / ML model and the received at least one second PMI.

[0021] In some implementations of the methods and apparatus described in this paper, the second AI / ML model can be associated with CSI feedback decompression.

[0022] In some implementations of the methods and apparatus described herein, receiving a first PMI and at least one second PMI may include: receiving from a UE via a receiver a request for determining whether a first AI / ML model in at least one first AI / ML model can be paired with a second AI / ML model; and receiving from a UE via a transceiver a first PMI and a second PMI, the second PMI being associated with a first AI / ML model in at least one first AI / ML model.

[0023] In some implementations of the methods and apparatus described herein, the processor may also be configured to: compare a third CSI and a fourth CSI, the fourth CSI being obtained based on a second AI / ML model and a received second PMI; and transmit, via a transceiver, a pairing result of the second AI / ML model and the first AI / ML model to the UE, the pairing result being determined based on the comparison result.

[0024] In some implementations of the methods and apparatus described herein, if the mean squared error (MSE) between the third and fourth CSIs within a predetermined time period is greater than a threshold, the pairing result may indicate that the second AI / ML model is not paired with the first AI / ML model; and if the MSE is not greater than the threshold, the pairing result may indicate that the second AI / ML model is paired with the first AI / ML model, and the model ID or function ID associated with the second AI / ML model is indicated in the pairing result.

[0025] In some implementations of the methods and apparatus described herein, receiving a first PMI and at least one second PMI may include: sending a request to the UE via a transceiver to determine whether a second AI / ML model can be paired with an active first AI / ML model among at least one first AI / ML model; and receiving the first PMI and the second PMI from the UE via the transceiver, the second PMI being associated with the active first AI / ML model.

[0026] In some implementations of the methods and apparatus described herein, the processor may also be configured to send an indication to the UE via a transceiver to deactivate the activated first AI / ML model if the second AI / ML model is determined by the base station not to be paired with the activated first AI / ML model.

[0027] In some implementations of the methods and apparatus described herein, the processor may also be configured to: calculate the mean squared error (MSE) between a third CSI and a fourth CSI over a predetermined time period, wherein the fourth CSI is obtained based on a second AI / ML model and a second PMI associated with an activated first AI / ML model, and if the calculated MSE is greater than a threshold, the second AI / ML model is determined not to be paired with the activated first AI / ML model.

[0028] In some implementations of the methods and apparatus described herein, at least one second PMI may include a plurality of second PMIs associated with a plurality of first AI / ML models; and receiving the first PMI and at least one second PMI may include: receiving a request from the UE via a transceiver for selecting a first AI / ML model from the plurality of first AI / ML models; and receiving the first PMI and the plurality of second PMIs from the UE via a transceiver.

[0029] In some implementations of the methods and apparatus described herein, the processor may also be configured to: select a first AI / ML model from a plurality of first AI / ML models based on a third CSI and a plurality of fourth CSIs, the plurality of fourth CSIs being obtained based on a second AI / ML model and a plurality of second PMIs; and send an indication of the selection result to the UE via a transceiver, wherein if the first AI / ML model is selected from the plurality of first AI / ML models, the indication may specify the model identifier (ID) of the selected first AI / ML model, and if no first AI / ML model is selected, the indication may indicate a pairing failure.

[0030] In some implementations of the methods and apparatus described herein, selecting a first AI / ML model from a plurality of first AI / ML models may include: selecting a first AI / ML model from a plurality of first AI / ML models based on the mean squared error (MSE) between a third CSI and each of a plurality of fourth CSIs in a predetermined time period.

[0031] In some implementations of the methods and apparatus described herein, the processor may also be configured to: receive from the UE via a transceiver an indication of the number of a plurality of first AI / ML models, and transmit to the UE via the transceiver a CSI report configuration and a CSI reference signal (RS) configuration determined based on the indication of the plurality of first AI / ML models.

[0032] In some implementations of the methods and apparatus described herein, the configuration of the first codebook may be based on at least one AI / ML model or function at the UE, wherein the function is associated with one or more AI / ML models.

[0033] In some implementations of the methods and apparatus described herein, the configuration of the first codebook may include scaling factors for parameters associated with a second codebook, which is indicated in a codebook configuration transmitted from a base station.

[0034] In some implementations of the methods and apparatus described herein, the configuration of the first codebook may be transmitted as function / model-related information during a function or model identification process between the UE and the base station, or the configuration of the first codebook may be transmitted via radio resource control (RRC) signaling.

[0035] In some implementations of the methods and apparatus described herein, parameters may include at least one of the following: the number of beams; the number of phase quantization sizes; the number of oversamples; the number of beam amplitude scaling factors for both broadband and subband; or the number of beam combination coefficients or phases in the beams, polarizations, and layers.

[0036] In some implementations of the methods and apparatus described herein, the processor may be configured to receive a first PMI and at least one second PMI together via one of the following: uplink control information (UCI), media access control (MAC) control unit (CE), or radio resource control (RRC).

[0037] In some implementations of the methods and apparatus described herein, the CSI or eigenvalues ​​(EVs) of the channel matrix constructed based on the first codebook can be the input to at least one first AI / ML model.

[0038] In a third aspect of the solution, a processor for wireless communication may include: at least one memory; and a controller coupled to the at least one memory and configured such that the processor: receives configuration information about a first codebook from a base station via a transceiver; and transmits a first precoding matrix indicator (PMI) and at least one second PMI to the base station via the transceiver, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

[0039] In a fourth aspect of the solution, a method performed by a user equipment (UE) may include: receiving configuration of a first codebook from a base station via a transceiver; and transmitting a first precoding matrix indicator (PMI) and at least one second PMI to the base station via the transceiver, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

[0040] In a fifth aspect of the solution, a processor for wireless communication may include: at least one memory; and a controller coupled to the at least one memory and configured such that the processor: transmits configuration of a first codebook to a user equipment (UE) via a transceiver; receives a first precoding matrix indicator (PMI) and at least one second PMI from the UE via the transceiver, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model; and determines, based on a third CSI and at least one fourth CSI, whether a second AI / ML model of a base station can be paired with at least one first AI / ML model, the third CSI being obtained based on the first codebook and the received first PMI, the at least one fourth CSI being obtained based on the second AI / ML model and the received at least one second PMI.

[0041] In a sixth aspect of the solution, a method performed by a base station may include: transmitting configuration of a first codebook to a user equipment (UE) via a transceiver; receiving a first precoding matrix indicator (PMI) and at least one second PMI from the UE via the transceiver, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model; and determining, based on a third CSI and at least one fourth CSI, whether a second AI / ML model of the base station can be paired with at least one first AI / ML model, the third CSI being obtained based on the first codebook and the received first PMI, the at least one fourth CSI being obtained based on the second AI / ML model and the received at least one second PMI.

[0042] It should be understood that the content of this disclosure is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to be used to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0043] Figure 1 An example of a wireless communication system supporting model pairing for AI / ML-based CSI compression is illustrated according to various aspects of this disclosure.

[0044] Figure 2 An example of the underlying AI / ML model used in CSI compression associated with various aspects of this disclosure is illustrated.

[0045] Figure 3 An example of a general training process for a two-sided model for CSI compression, associated with various aspects of this disclosure, is illustrated.

[0046] Figures 4 to 6 An example of two-sided model training associated with various aspects of this disclosure is illustrated.

[0047] Figure 7 An example of a signaling process for model pairing for AI / ML-based CSI compression is illustrated according to various aspects of this disclosure.

[0048] Figure 8 An example conceptual diagram illustrating dataset alignment instructions and model pairing processes according to various aspects of this disclosure is provided.

[0049] Figure 9 The illustration shows an example process for model pairing between a base station and a UE via an air interface using an aligned dataset, according to various aspects of this disclosure.

[0050] Figure 10The illustration shows a schematic diagram of a method for checking whether a model on one side can be paired with a model on the other side, according to various aspects of this disclosure.

[0051] Figure 11 The illustration shows a schematic diagram of a method for selecting a model from a set of models on one side to pair with a target model on the other side, according to various aspects of the present disclosure.

[0052] Figure 12 and Figure 13 An example of a device supporting model pairing for AI / ML-based CSI compression is illustrated according to various aspects of this disclosure.

[0053] Figure 14 and Figure 15 An example of a processor supporting model pairing for AI / ML-based CSI compression is illustrated according to various aspects of this disclosure.

[0054] Figure 16 and Figure 17 The diagram illustrates a flowchart of a method for model pairing for AI / ML-based CSI compression, supported by various aspects of this disclosure. Detailed Implementation

[0055] The principles of this disclosure will now be described with reference to some embodiments. It should be understood that these embodiments are described for illustrative purposes only and to assist those skilled in the art in understanding and implementing this disclosure, and do not imply any limitation on the scope of this disclosure. This disclosure described herein can be implemented in various ways other than those described below.

[0056] In the following description and claims, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0057] References to "an embodiment," "example embodiment," "embodiment," "some embodiments," etc., in this disclosure indicate that the embodiments(s) described may include a particular feature, structure, or characteristic, but not every embodiment necessarily includes that particular feature, structure, or characteristic. Furthermore, such phrases do not necessarily refer to the same(s) embodiments(s). Additionally, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is believed that in conjunction with other embodiments (whether explicitly described or not) affecting such feature, structure, or characteristic is within the knowledge of those skilled in the art.

[0058] It should be understood that although the terms “first” and “second”, etc., may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used only to distinguish one element from another. For example, a first element may also be referred to as a second element without departing from the scope of the embodiments, and similarly, a second element may also be referred to as a first element. As used herein, the term “and / or” includes any and all combinations of one or more of the listed terms.

[0059] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments. As used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly specifies otherwise. It should also be understood that the terms “comprising,” “including,” “having,” “having,” “containing,” and / or “comprising” as used herein indicate the presence of the stated features, elements, and / or components, etc., but do not exclude the presence or addition of one or more other features, elements, components, and / or combinations thereof.

[0060] As used herein, the term "communication network" refers to a network that conforms to any suitable communication standard, such as 5G New Radio (NR), Long Term Evolution (LTE), LTE-Advanced (LTE-A), Wideband Code Division Multiple Access (WCDMA), High-Speed ​​Packet Access (HSPA), Narrowband Internet of Things (NB-IoT), etc. Furthermore, communication between terminal devices and network devices in a communication network can be performed according to any suitable generation of communication protocols, including but not limited to first-generation (1G), second-generation (2G), 2.5G, 2.75G, third-generation (3G), fourth-generation (4G), 4.5G, fifth-generation (5G) communication protocols, and / or any other currently known or to be developed in the future. Embodiments of this disclosure can be applied to various communication systems. Given the rapid development of communications, future types of communication technologies and systems will also exist in which this disclosure can be embodied. This should not be construed as limiting the scope of this disclosure to the aforementioned systems.

[0061] As used herein, the term "network device" generally refers to a node in a communication network through which terminal devices can access the communication network and receive services. Network devices can refer to base stations (BS) or access points (APs), such as Node B (NodeB or NB), Radio Access Network (RAN) nodes, Evolved Node B (eNodeB or eNB), NR NB (also known as gNB), Remote Radio Unit (RRU), Radio Head (RH), infrastructure equipment for V2X (vehicle-to-everything) communication, Transmit and Receive Point (TRP), Receive Point (RP), Remote Radio Head (RRH), relay, Integrated Access and Backhaul (IAB) nodes, and low-power nodes such as femtoBS, picoBS, etc., depending on the terminology and technology applied.

[0062] As used herein, the term "terminal device" generally refers to any terminal device capable of wireless communication. By way of example and not limitation, a terminal device may also be referred to as a communication device, user equipment (UE), end user equipment, subscriber station (SS), unmanned aerial vehicle (UAV), portable subscriber station, mobile station (MS), or access terminal (AT). Terminal devices may include, but are not limited to: mobile phones, cellular phones, smartphones, Voice over IP (VoIP) phones, wireless local loop phones, tablets, wearable terminal devices, personal digital assistants (PDAs), portable computers, desktop computers, image capture terminal devices (such as digital cameras), gaming terminal devices, music storage and playback devices, in-vehicle wireless terminal devices, wireless endpoints, mobile stations, laptop embedded devices (LEEs), laptop mounted devices (LMEs), USB dongles, smart devices, wireless customer premises equipment (CPEs), Internet of Things (IoT) devices, watches or other wearable devices, head-mounted displays (HMDs), vehicles, drones, medical devices (e.g., remote surgical equipment), industrial equipment (e.g., robots and / or other wireless devices operating in the context of industrial and / or automated processing chains), consumer electronics devices, devices operating on commercial and / or industrial wireless networks, etc. In the following description, the terms "terminal device," "communication device," "terminal," "user equipment," and "UE" are used interchangeably.

[0063] AI / ML is used to learn and perform certain tasks by training neural networks using massive amounts of data, and it has been successfully applied in the fields of computer vision (CV) and natural language processing (NLP). As a subset of ML, deep learning (DL) utilizes multi-layer neural networks (NNs) as an "AI model" to learn from massive amounts of data to solve problems and optimize performance. Due to the promising benefits demonstrated in numerous academic papers and field test results, AI / ML-based methods can achieve better performance than traditional methods when well trained.

[0064] Within 3GPP, there is currently discussion about introducing AI / ML into the air interface in NR Releases 18 and 19 for a number of selected use cases, including CSI feedback enhancement, beam management, and improved positioning accuracy, as well as some agreed-upon general frameworks, evaluation methods, and results.

[0065] For use cases involving enhanced CSI feedback, it is desirable to use AI / ML methods to reduce the overhead of CSI reporting, particularly the precoding matrix indicator (PMI) overhead, based on defined codebooks (such as Type I, Type II, and eType II). These codebooks allow the network to select downlink precoders that not only focus on the energy transmitted at the target device but also limit interference to other devices scheduled in parallel on the same time / frequency resources.

[0066] Type II CSI was introduced along with Type I CSI as part of 3GPP Release 15. Several significant extensions and enhancements to Type I CSI (known as eType II) were introduced in Release 16. The higher spatial granularity of PMI feedback comes at the cost of significantly higher signaling overhead. PMI reports for Type I CSI can consist of up to tens of bits, while those for Type II CSI can consist of hundreds of bits.

[0067] Version 15 Type II CSI precoder can be represented as the product of two matrices, as follows:

[0068] in Reported as a broadband precoder, it has the following structure:

[0069] In addition, matrix Provides amplitude values ​​(partial broadband and partial subband reports) and phase values ​​(for subband reports).

[0070] NR Release 16 introduces Enhanced Type II CSI, which shares the same basic principle as Release 15: reporting a set of beams on a broadband basis and a set of combining factors on a narrower basis. The reported beams are linearly combined using the combining factors to provide a set of precoder vectors, one for each layer. Therefore, a key feature of Release 16 Enhanced Type II CSI is the use of correlations in the frequency domain to reduce reporting overhead. Simultaneously, Release 16 Type II CSI allows for a twofold improvement in the frequency domain granularity of PMI reporting. This is achieved by introducing the concept of frequency domain (FD) units in conjunction with compression operations, where each FD unit corresponds to a subband or half-subband. Therefore, compared to one precoder per subband for Release 15 Type II CSI, Release 16 Type II CSI provides the transmitter side with a recommended precoder for each FD unit.

[0071] More specifically, for a given layer k, for version 16 Type II CSI, the precoder vector reported for all FD units can be represented as:

[0072] in N It is the number of FD units to be reported, and It is the pre-encoder vector for the nth FD unit of a given layer k. As described above for Version 15 Type II CSI, column B corresponds to L The selected beam. Furthermore, The report is the same for all FD units and for all layers. It is the size of M × N The compression matrix, which consists of a set of row vectors based on the DFT, and provides information from the corresponding data covered by the CSI report. N The transformation of the frequency domain of dimension N of each FD unit to a smaller delay domain of dimension M. This compression matrix is ​​frequency-independent (a common matrix for all FD units), but is reported separately for each layer. Furthermore, the number of rows in this matrix can be represented as... Where R is the number of FD units per subframe ( R =1 or 2), and p Configurable parameters for controlling the amount of compression. Additionally, the size is 2. L × M matrix Mapping from the delay domain to the beam domain is similar to the matrix-based approach for all sub-bands used in Version 15 Type II CSI. The matrix. Additionally, only... Total 2 L × M Fractions of each element β It is assumed to be a non-zero value and therefore needs to be reported, where β These are configurable parameters.

[0073] Therefore, the overhead reduction of the version 16 enhanced Type II CSI is due to the following two points: 1) The dimension of the delay space is smaller than the dimension of the frequency domain, as determined by the parameters. p Given; and 2). The finite fraction of non-zero elements is given by the parameter β Provided.

[0074] To further reduce the overhead of CSI feedback, AI / ML-based CSI compression was proposed as one of the use cases in Release 18 for studying AI / ML for air interface enhancement.

[0075] Figure 2 The illustration shows an example of the basic AI / ML model used in CSI compression. Figure 2 As shown, a dual-side AI / ML model is deployed to compress CSI / PMI. On the UE side, an AI model (i.e., the UE part model) is deployed, and the measured CSI (e.g., the estimated full CSI) is compressed via a neural network (NN) (e.g., a convolutional neural network (CNN) model, which may also be referred to as an encoder) after some preprocessing (e.g., eigenvalue decomposition (EVD)). After quantization, these bits are sent to the gNB as part of the CSI report (i.e., the compressed and quantized CSI) on the Physical Uplink Shared Channel (PUSCH) / Physical Uplink Control Channel (PUCCH). At the gNB, the received bits (i.e., the compressed and quantized CSI) are fed into another AI model (e.g., a CNN model, which may also be referred to as a decoder) deployed on the gNB side (i.e., the NW part model) to recover the CSI, thus obtaining the recovered full CSI.

[0076] In AI / ML-based CSI compression, performance is fully evaluated to identify the benefits from AI / ML and potential canonical impacts, particularly for model training, pairing, and monitoring of such two-sided models. For model training, training schemes for two-sided models discussed in Rel-18 are introduced, including Type 1, Type 2, and Type 3, which present different challenges in the air interface.

[0077] Figure 3An example of a general training process for a two-sided model for CSI compression, associated with various aspects of this disclosure, is illustrated. To train the two models (i.e., the NW part model and the UE part model), a shared dataset of CSI / Eigenvalues ​​(EVs) needs to be used for both models; for example, it can be used as input to the UE part model and labeled data for the NW part model. If, after multiple iterations between the two models, the mean squared error (MSE) between the labeled data and the output of the NW part model is sufficiently small, the two-sided model can be considered ready for deployment or inference.

[0078] Figures 4 to 6 An example of two-sided model training associated with various aspects of this disclosure is illustrated.

[0079] Figure 4 The diagram illustrates a Type 1 training scheme for a two-sided model, i.e., joint training on one side. In a Type 1 training scheme, the two models can be trained on one side (UE side or NW side), and after training is complete, one of the trained models can be transferred to the other side. For example... Figure 4 As shown, for example, the two models are trained on the NW side (402), and after the two trained models are ready for deployment / inference (404), the trained model for the encoder (i.e., the UE part model) is transferred (406) and deployed (410) to the UE side, and another trained model for the decoder is deployed (408) on the NW side.

[0080] However, for Type 1 training schemes, while the training dataset can be kept on one side without alignment, transferring the model to the other side is challenging in practice. The trained model cannot be optimized for this purpose because the software and hardware may be unavailable on the other side. On the other hand, supporting model transfer in version 19 is also problematic for the complex and diverse software / hardware platforms that support AI / ML.

[0081] Figure 5 The diagram illustrates a Type 2 training scheme for a two-sided model, i.e., joint training on both sides. In the Type 2 training scheme, the two models are trained on both sides (including the UE side and the NW side).

[0082] like Figure 5 As shown, for example, assuming the datasets used for training are identical and aligned, the data during training (i.e., the values ​​of forward propagation (FP) and backward propagation (BP)) interacts between the UE and NW. Figure 5In this process, the two models are trained on the UE side and the NW side respectively (502, 504), and after the two trained models are ready for deployment / inference (506, 508), the trained models are deployed on the UE side and the NW side respectively (510, 512).

[0083] Although both models can be optimized on dedicated software / hardware platforms, the challenge of frequently exchanging FP / BP data remains, which would be a significant burden on the air interface. Furthermore, aligning the datasets used for training on both sides is also a major challenge in practice.

[0084] Figure 6 The diagram illustrates a Type 3 training scheme for a two-sided model, namely, sequential training. In the Type 3 training scheme, sequential training is employed, meaning that two models are first trained on one side (UE side or NW side). After training is complete, the generated data (i.e., compressed CSI) can be transferred to the other side along with the trained model, where training is then performed using the transferred data and an aligned dataset. For example, as... Figure 6 As shown, the two models are first trained on the NW side (602), and after the two trained models are training-ready (604), the trained model for the NW can be deployed (608) on the NW side. Furthermore, another trained model for the UE, along with compressed CSI, is transmitted to the UE side, where it is then trained using the transmitted data and aligned dataset (606). After the model for the UE is training-ready at the UE, it can be deployed (610) at the UE.

[0085] For Type 3 training schemes, both models can be potentially optimized based on the deployed software / hardware platform. Data generated from aligned datasets needs to be transferred, which incurs less overhead than in Type 2 and is platform-friendly across diverse environments. However, in practice, if the model on the NW side wants to support multiple models on the UE, it's best to use one model or a limited number of models paired with the models in the UE. Therefore, if training is required, Type 3 training schemes can be significantly better than the other two types (i.e., Type 1 and Type 2 training schemes). Furthermore, if models already exist on both sides, it's necessary to examine whether they can be paired in certain scenarios and how to monitor the performance of the paired models.

[0086] Some conclusions have already been captured in TR38.843. The following lists some potential specification impacts on CSI feedback enhancements for CSI compression using a two-sided model, including but not limited to: fallback mode, NW / UE alignment, model input / output, UE-side data collection, NW-side data collection, CSI configuration and reporting, and feasibility and methods to support traditional CSI reporting principles.

[0087] For example, regarding fallback modes, there may be potential regulatory implications for the coexistence and fallback mechanisms used to support AI / ML-based CSI feedback modes and traditional non-AI / ML-based CSI feedback modes.

[0088] For NW / UE alignment, there may be potential specification impacts on the alignment of quantization / dequantization methods and feedback message sizes between the network and the UE, including: (1) for vector quantization schemes, VQ codebook format and size, and the size and segmentation method of CSI generation model output; and (2) for scalar quantization schemes, uniform and non-uniform quantization and format, e.g., quantization granularity, which consists of the distribution of bits allocated to each floating point; and (3) quantization alignment using 3GPP-aware mechanisms.

[0089] For model input / output, there may be potential canonical implications for output-CSI-UE and input-CSI-NW, at least for the precoding matrix (e.g., the precoding matrix in the spatial-frequency domain or the precoding matrix represented using angle-delay domain projection). In the case of the precoding matrix represented using angle-delay domain projection, depending on the performance evaluation, the explicit channel matrix (i.e., the complete Tx) Rx MIMO channels have also been studied, where the original channel is located in the space-frequency domain or the original channel is located in the angle-delay domain.

[0090] For UE-side data collection, there may be potential regulatory impacts on the following aspects: enhancements to CSI-RS configuration to enable higher accuracy measurements; auxiliary information for UE data collection used to classify data in the form of IDs, for the purpose of distinguishing data characteristics caused by specific configurations, scenarios, sites, etc. (the feasibility of disclosing proprietary information to the other side should be considered when providing auxiliary information); and signaling used to trigger data collection.

[0091] For network-side data collection, there may be potential regulatory impacts on the following aspects: (1) enhancements to Sounding Reference Signals (SRS) and / or CSI-RS measurements and / or CSI reporting to enable higher accuracy measurements; (2) the content of ground-based real CSI, including data sample types (e.g., precoding matrix, channel matrix, etc.), data sample formats including scalar quantization and / or codebook-based quantization (e.g., e-type II-like), and auxiliary information (e.g., timestamps and / or cell IDs; auxiliary information for network data collection used to classify data in ID form, to distinguish due to specific configurations, The purpose of the data characteristics caused by the scene, site, etc.; and data quality indicators); (3) the latency requirements for data collection; (4) the signaling for triggering data collection; (5) the ground real CSI report for NW side data collection, which is used for model performance monitoring, the ground real CSI report includes scalar quantization for ground real CSI, codebook-based quantization for ground real CSI, radio resource control (RRC) signaling and / or L1 signaling procedures for enabling fast representation of AI / ML model performance, and non-periodic / semi-persistent or periodic ground real CSI reports; (6) Used for model training The document describes the ground truth CSI format, targeting either scalar quantization or codebook-based quantization that includes ground truth CSI. This paper considers the number of layers for which ground truth data is collected, and whether the number of layers used for ground truth CSI data collection is determined by the UE or the NW.

[0092] For CSI configuration and reporting, there may be potential specification impacts on the following aspects: (1) NW configuration used to determine the size of the CSI payload, such as possible CSI payload size, possible class limits and / or other related configurations; (2) how the UE determines / reports the actual CSI payload size and / or other CSI-related information within the constraints of the network configuration; and (3) the related uplink control information (UCI) format, which takes into account the traditional CSI reporting principle and CSI Part 1 and Part 2 as a starting point, where Part 1 has a fixed size configured by the network, while the size of Part 2 is dynamic and determined by the information in Part 1.

[0093] The feasibility and methodology used to support traditional CSI reporting principles may have potential regulatory implications for the following aspects: priority rules for CSI conflict handling and CSI omission, codebook subset constraints (e.g., input-CSI-NW / output-CSI-UE considered in the angle-delay domain, beam constraints that can be based on input CSI in the angle domain based on traditional SD base vectors), and CSI processing units.

[0094] In addition to the potential regulatory impacts mentioned above, there may be some potential regulatory enhancements to the following aspects: (1) CSI-RS configuration (excluding CSI-RS mode design enhancements); (2) CSI configuration (for the network to indicate information related to CSI reporting, for example, the gNB to indicate one or more of the following to the UE: information indicating the size of the CSI payload, information indicating the quantization method / granularity, level restrictions, and other payload-related aspects); (3) CSI reporting configuration (for the UE to determine / report the actual CSI payload size, and for the UE to report information related to the NW configuration); (4) CSI reporting UCI mapping / priority / omission; and (5) CSI processing procedures.

[0095] To support AI / ML-based CSI compression operations, new signaling needs to be introduced or existing signaling enhanced, and several issues need to be considered and addressed. The first issue is that the datasets used for model training on either side are very large and proprietary, and for privacy reasons, these datasets are not always permitted to be transmitted to the other side. Specifically, the training dataset plays a crucial role for any AI / ML-based method. As discussed above, the training datasets on either side can be proprietary, such as in terms of quantization level and data format. While the dataset could be transmitted to the other side, this approach consumes excessive radio resources and is unacceptable for privacy.

[0096] The second issue is that the performance of the two-sided models used for CSI compression is sensitive to model pairing. Evaluation results show that performance is highly correlated with the paired models. If the two-sided models are mismatched, performance can be severely degraded. There must be a canonical effect on the model pairing process for the two-sided models.

[0097] Furthermore, a third issue is how to monitor the performance of the two-sided model used for AI / ML-based CSI compression. Similar to other AI / ML-based methods, it is also necessary to monitor the performance of the two-sided model, which differs from the one-sided model due to the metric alignment for degradation.

[0098] Since the two AI / ML models are located on either side, a scheme needs to be designed to pair these models via an over-the-air interface to ensure performance and further monitoring.

[0099] This disclosure presents a solution for supporting model pairing for AI / ML-based CSI compression. In this solution, the UE and the base station can share a high-resolution codebook. By using the high-resolution codebook, the PMI can be obtained based on the CSI estimated by the UE, and is sent to the base station along with multiple PMIs generated via multiple models at the UE to evaluate whether the UE-side model is paired with the base station-side model. In this way, two models for AI / ML-based CSI compression can be better paired via the air interface, while the performance and further monitoring of AI / ML-based CSI compression are also ensured.

[0100] The aspects of this disclosure are described in the context of wireless communication systems.

[0101] Figure 1 An example of a wireless communication system 100 supporting CSI compression according to various aspects of this disclosure is illustrated. The wireless communication system 100 may include one or more network entities 102 (also referred to as network devices (NEs)), one or more UEs 104, a core network 106, and a packet data network 108. The wireless communication system 100 may support various radio access technologies. In some implementations, the wireless communication system 100 may be a 4G network, such as an LTE network or an LTE-advanced (LTE-A) network. In some other implementations, the wireless communication system 100 may be a 5G network, such as an NR network. In other implementations, the wireless communication system 100 may be a combination of 4G and 5G networks, or other suitable radio access technologies, including IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20. The wireless communication system 100 may support radio access technologies other than 5G. In addition, the wireless communication system 100 can support technologies such as time division multiple access (TDMA), frequency division multiple access (FDMA), or code division multiple access (CDMA).

[0102] One or more network entities 102 may be distributed throughout a geographic area to form a wireless communication system 100. The one or more network entities 102 described herein may be, include, or may be referred to as network nodes, base stations, network elements, radio access networks (RANs), base transceivers, access points, NodeBs, eNodeBs (eNBs), next-generation NodeBs (gNBs), or other suitable terms. Network entities 102 and UE 104 may communicate via communication link 110, which may be a wireless or wired connection. For example, network entities 102 and UE 104 may perform wireless communication (e.g., receive signaling, send signaling) via a Uu interface.

[0103] Network entity 102 may provide a geographic coverage area 112 for which it may support services (e.g., voice, video, packet data, messaging, broadcasting, etc.) for one or more UEs 104 within the geographic coverage area 112. For example, network entity 102 and UE 104 may support wireless communication of signals associated with services (e.g., voice, video, packet data, messaging, broadcasting, etc.) based on one or more radio access technologies. In some implementations, network entity 102 may be mobile, for example, a satellite associated with a non-terrestrial network. In some implementations, different geographic coverage areas 112 associated with the same or different radio access technologies may overlap, but different geographic coverage areas 112 may be associated with different network entities 102. The information and signals described herein may be represented using a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referenced throughout the description may be represented by voltage, current, electromagnetic waves, magnetic fields or particles, light fields or particles, or any combination thereof.

[0104] One or more UEs 104 may be distributed throughout the geographic area of ​​the wireless communication system 100. UE 104 may include or be referred to as a mobile device, wireless device, remote device, remote unit, handheld device, or subscriber device, or some other suitable term. In some implementations, UE 104 may be referred to as a unit, station, terminal, or client, among other examples. Alternatively or additionally, UE 104 may be referred to as an Internet of Things (IoT) device, an Internet of Everything (IoE) device, or a Machine Type Communication (MTC) device, among other examples. In some implementations, UE 104 may be stationary within the wireless communication system 100. In some other implementations, UE 104 may be mobile within the wireless communication system 100.

[0105] One or more UEs 104 can be devices in different forms or with different capabilities. Some examples of UEs 104 are shown in... Figure 1 The diagram shows that UE 104 can communicate with various types of devices, such as network entity 102, other UEs 104, or network devices (e.g., core network 106, packet data network 108, relay devices, integrated access and backhaul (IAB) nodes, or another network device). Figure 1 As shown in the diagram. Alternatively or concurrently, UE 104 may support communication with other network entities 102 or UE 104, which may act as relays in the wireless communication system 100.

[0106] UE 104 can also support direct wireless communication with other UE 104 via communication link 114. For example, UE 104 can support direct wireless communication with another UE 104 via a device-to-device (D2D) communication link. In some implementations (such as vehicle-to-vehicle (V2V) deployments, vehicle-to-everything (V2X) deployments, or cellular V2X deployments), communication link 114 may be referred to as a sidelink. For example, UE 104 can support direct wireless communication with another UE 104 via a PC5 interface.

[0107] Network entity 102 may support communication with core network 106, or with another network entity 102, or both. For example, network entity 102 may interface with core network 106 via one or more backhaul links 116 (e.g., via S1, N2, N2, or another network interface). Network entities 102 may communicate with each other via backhaul links 116 (e.g., via X2, Xn, or another network interface). In some implementations, network entities 102 may communicate directly with each other (e.g., between network entities 102). In some other implementations, network entities 102 may communicate with each other or indirectly (e.g., via core network 106). In some implementations, one or more network entities 102 may include sub-components, such as access network entities, which may be examples of access node controllers (ANCs). The ANC may communicate with one or more UEs 104 via one or more other access network transport entities (which may be referred to as radio heads, smart radio heads, or transmit-receive points (TRPs)).

[0108] In some implementations, network entity 102 can be configured with a decomposed architecture, which can be configured to utilize protocol stacks physically or logically distributed across two or more network entities 102, such as an Integrated Access Backhaul (IAB) network, an Open Radio Access Network (O-RAN) (e.g., a network configuration sponsored by the O-RAN Alliance), or a Virtualized RAN (vRAN) (e.g., a Cloud RAN (C-RAN)). For example, network entity 102 may include one or more of the following: CU, DU, Radio Unit (RU), RAN Intelligent Controller (RIC) (e.g., Near-RT RIC, Non-RT RIC), Service Management and Orchestration (SMO) system, or any combination thereof.

[0109] An RU can also be referred to as a radio head, intelligent radio head, remote radio head (RRH), remote radio unit (RRU), or transmit-receive point (TRP). One or more components of network entity 102 in the decomposed RAN architecture may be co-located, or one or more components of network entity 102 may be located in distributed locations (e.g., separate physical locations). In some implementations, one or more network entities 102 in the decomposed RAN architecture may be implemented as virtual units (e.g., virtual CU (VCU), virtual DU (VDU), virtual RU (VRU)).

[0110] The functional splitting among CU, DU, and RU can be flexible and can support different functions depending on which functions (e.g., network layer functions, protocol layer functions, baseband functions, radio frequency functions, and any combination thereof) are performed at the CU, DU, or RU. For example, protocol stack functional splitting can be adopted between the CU and DU, allowing the CU to support one or more layers of the protocol stack, while the DU can support one or more different layers of the protocol stack. In some implementations, the CU can host higher protocol layer (e.g., Layer 3 (L3), Layer 2 (L2)) functions and signaling (e.g., Radio Resource Control (RRC), Serving Data Adaptation Protocol (SDAP), Packet Data Convergence Protocol (PDCP)). The CU can connect to one or more DUs or RUs, and one or more DUs or RUs can host lower protocol layer functions and signaling, such as Layer 1 (L1) (e.g., Physical (PHY) layer) or L2 (e.g., Radio Link Control (RLC) layer, Media Access Control (MAC) layer), and each can be at least partially controlled by the CU 160.

[0111] Alternatively or concurrently, functional splitting of the protocol stack can be employed between the DU and RU, allowing the DU to support one or more layers of the protocol stack, while the RU can support one or more different layers of the protocol stack. The DU can support one or more different cells (e.g., via one or more RUs). In some implementations, functional splitting between the CU and DU, or between the DU and RU, can be within the protocol layer (e.g., some functions of the protocol layer can be performed by one of the CU, DU, or RU, while other functions of the protocol layer are performed by a different one of the CU, DU, or RU).

[0112] The CU can also be further functionally divided into CU control plane (CU-CP) and CU user plane (CU-UP) functions. The CU can be connected to one or more DUs via midhaul communication links (e.g., F1, F1-c, F1-u), while the DUs can be connected to one or more RUs via fronthaul communication links (e.g., open fronthaul (FH) interfaces). In some implementations, the midhaul or fronthaul communication links can be implemented based on interfaces (e.g., channels) between layers of a protocol stack supported by corresponding network entities 102 communicating via such communication links.

[0113] Core network 106 can support user authentication, access authorization, tracking, connectivity, and other access, routing, or mobility functions. Core network 106 can be an evolved packet core (EPC) or a 5G core (5GC), which may include control plane entities that manage access and mobility (e.g., a mobility management entity (MME) or access and mobility management function (AMF)) and user plane entities that route packets or interconnects to external networks (e.g., a serving gateway (S-GW), a packet data network (PDN) gateway (P-GW), or a user plane function (UPF)). In some implementations, the control plane entities may manage non-access stratum (NAS) functions, such as mobility, authentication, and bearer management (e.g., data bearers, signaling bearers, etc.) for one or more UEs 104 served by one or more network entities 102 associated with core network 106.

[0114] Core network 106 can communicate with packet data network 108 via one or more backhaul links 116 (e.g., via S1, N2, N2, or another network interface). Packet data network 108 may include application server 118. In some implementations, one or more UEs 104 may communicate with application server 118. UE 104 may establish a session (e.g., Protocol Data Unit (PDU) session, etc.) with core network 106 via network entity 102. Core network 106 can use the established session (e.g., an established PDU session) to route services (e.g., control information, data, etc.) between UE 104 and application server 118. A PDU session may be an example of a logical connection between UE 104 and core network 106 (e.g., one or more network functions of core network 106).

[0115] In the wireless communication system 100, network entity 102 and UE 104 can use the resources of the wireless communication system 100 (e.g., time resources (e.g., symbols, time slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers)) to perform various operations (e.g., wireless communication). In some implementations, network entity 102 and UE 104 can support different resource structures. For example, network entity 102 and UE 104 can support different frame structures. In some implementations, such as in 4G, network entity 102 and UE 104 can support a single frame structure. In some other implementations, such as in 5G and other suitable radio access technologies, network entity 102 and UE 104 can support various frame structures (i.e., multiple frame structures). Network entity 102 and UE 104 can support various frame structures based on one or more sets of parameters.

[0116] The wireless communication system 100 may support one or more parameter sets, and the parameter sets may include subcarrier spacing and cyclic prefixes. The first parameter set (e.g., μ =0) can be associated with the first subcarrier spacing (e.g., 15kHz) and the normal cyclic prefix. In some implementations, the first parameter set (e.g., ) associated with the first subcarrier spacing (e.g., 15kHz) is... μ =0) can utilize one time slot per subframe. The second parameter set (e.g., μ =1) can be associated with the second subcarrier spacing (e.g., 30kHz) and the normal cyclic prefix. The third parameter set (e.g., μ =2) can be associated with the third subcarrier spacing (e.g., 60 kHz) and a normal cyclic prefix or an extended cyclic prefix. The fourth parameter set (e.g., μ =3) can be associated with the fourth subcarrier spacing (e.g., 120 kHz) and the normal cyclic prefix. The fifth parameter set (e.g., μ =4) can be associated with the fifth subcarrier spacing (e.g., 240 kHz) and the normal cyclic prefix.

[0117] The time intervals of resources (e.g., communication resources) can be organized according to frames (also known as radio frames). Each frame can have a duration, for example, 10 milliseconds (ms). In some implementations, each frame can include multiple subframes. For example, each frame can include 10 subframes, and each subframe can have a duration, for example, 1 ms. In some implementations, each frame can have the same duration. In some implementations, each subframe of a frame can have the same duration.

[0118] Alternatively or concurrently, the time intervals of resources (e.g., communication resources) can be organized according to time slots. For example, a subframe may include a certain number (e.g., a certain quantity) of time slots. The number of time slots in each subframe may also depend on one or more parameter sets supported in the wireless communication system 100. For example, a first parameter set, a second parameter set, a third parameter set, a fourth parameter set, and a fifth parameter set (i.e., ...) associated with corresponding subcarrier intervals of 15 kHz, 30 kHz, 60 kHz, 120 kHz, and 240 kHz. μ =0、 μ =1、 μ =2、 μ =3、 μ =4) A single time slot per subframe, two time slots per subframe, four time slots per subframe, eight time slots per subframe, and 16 time slots per subframe can be used, respectively. Each time slot can include a certain number (e.g., a certain quantity) of symbols (e.g., OFDM symbols). In some implementations, the number (e.g., quantity) of time slots per subframe can depend on the parameter set. For a normal cyclic prefix, a time slot can include 14 symbols. For an extended cyclic prefix (e.g., for a 60kHz subcarrier spacing), a time slot can include 12 symbols. The relationship between the number of symbols per time slot, the number of time slots per subframe, and the number of time slots per frame for a normal or extended cyclic prefix can depend on the parameter set. It should be understood that for a first parameter set (e.g., ...) associated with a first subcarrier spacing (e.g., 15kHz) ... μ The reference of =0 can be used interchangeably between subframes and time slots.

[0119] In the wireless communication system 100, the electromagnetic (EM) spectrum can be divided into various classes, frequency bands, channels, etc., based on frequency or wavelength. For example, the wireless communication system 100 can support one or more operating frequency bands, such as frequency range names FR1 (410 MHz - 7.125 GHz), FR2 (24.25 GHz - 52.6 GHz), FR3 (7.125 GHz - 24.25 GHz), FR4 (52.6 GHz - 114.25 GHz), FR4a or FR4-1 (52.6 GHz - 71 GHz), and FR5 (114.25 GHz - 300 GHz). In some implementations, network entity 102 and UE 104 can perform wireless communication through one or more of these operating frequency bands. In some implementations, FR1 can be used by network entity 102, UE 104, and other devices or equipment for cellular communication services (e.g., control information, data). In some implementations, FR2 can be used by network entity 102, UE 104 and other devices or equipment for short-range, high data rate capabilities.

[0120] FR1 can be associated with one or more parameter sets (e.g., at least three parameter sets). For example, FR1 can be associated with the following: the first parameter set (e.g., μ =0), which includes a 15kHz subcarrier spacing; the second parameter set (e.g., μ =1), which includes a 30kHz subcarrier spacing; and a third parameter set (e.g., μ =2), which includes a 60kHz subcarrier spacing. FR2 can be associated with one or more parameter sets (e.g., at least two parameter sets). For example, FR2 can be associated with the following: a third parameter set (e.g., μ =2), which includes a 60kHz subcarrier spacing; and a fourth parameter set (e.g., μ =3), which includes a 120kHz subcarrier spacing.

[0121] Figure 7 An example of a signaling process 700 for model pairing for AI / ML-based CSI compression is illustrated according to various aspects of this disclosure. The signaling process 700 involves a base station 102 and a UE 104. In the following description, the base station 102 may also be referred to as an NW 103, gNB 102, etc.

[0122] like Figure 7 As shown, at step 702, base station 102 may send configuration information about the first codebook to UE 104. The configuration information about the first codebook is received by UE 104, so that base station 102 and UE 104 can share the same first codebook and perform the model matching process based on the first codebook.

[0123] In some embodiments of this disclosure, the first codebook may be a high-resolution codebook, and the configuration of the first codebook may be a high-resolution codebook indication, which includes scaling factors for parameters associated with a second codebook indicated in a codebook configuration transmitted from base station 102. Hereinafter, the second codebook may be the codebook currently used at UE 104 and base station 102 (e.g., the current Type II CSI codebook), and the parameters associated with the second codebook may include at least one of the following: the number of beams; the number of phase quantization sizes; the number of oversamples; the number of beam amplitude scaling factors for both broadband and subband; or the number of beam combination coefficients or phases in beams, polarizations, and layers.

[0124] In some embodiments of this disclosure, the configuration of the first codebook can be based on at least one AI / ML model or function at UE 104, which can be associated with one or more AI / ML models. In other words, the configuration of the first codebook can be function- or model-specific. Furthermore, the configuration of the first codebook can be transmitted as function / model-related information during a function or model identification process between UE 104 and base station 102, or the configuration of the first codebook can be transmitted via Radio Resource Control (RRC) signaling, which will combine... Figure 8 and Figure 9 Detailed explanation.

[0125] At step 704, UE 104 may transmit a first PMI and at least one second PMI, the first PMI being obtained based on a first codebook and a CSI estimated by the UE, and the at least one second PMI being obtained by at least one first AI / ML model of UE 104. In some embodiments of this disclosure, at least one first AI / ML model may be associated with CSI feedback compression. In some embodiments of this disclosure, UE 104 may transmit the first PMI and at least one second PMI together via one of the following: Uplink Control Information (UCI), Media Access Control (MAC) Control Unit (CE), or Radio Resource Control (RRC).

[0126] Base station 102 can receive a first PMI and at least one second PMI, and in step 706, based on the received first PMI and at least one received second PMI, for example, based on a third CSI and at least one fourth CSI, it determines whether a second AI / ML model of base station 102 can be paired with at least one first AI / ML model. The third CSI is obtained based on a first codebook and the received first PMI, and the at least one fourth CSI is obtained based on a second AI / ML model and at least one received second PMI. Thus, the model pairing process can be implemented. Some detailed operations of steps 704 and 706 are described below.

[0127] The model pairing process can be used to check whether an AI / ML model at one location on the UE side or the base station side can be paired with an AI / ML model at another location on the UE side or the base station side.

[0128] For example, in some embodiments of this disclosure, the model pairing process may be initiated by UE 104 to check whether a first AI / ML model in at least one first AI / ML model of UE 104 can be paired with a second AI / ML model of base station 102. In this case, step 704 may include further operations or actions. As an example, UE 104 may send a request to base station 102 to determine whether a first AI / ML model in at least one first AI / ML model of UE 104 can be paired with a second AI / ML model of base station 102, and send a first PMI and a second PMI to base station 102, the second PMI being associated with the first AI / ML model. The request to determine whether the first AI / ML model can be paired with the second AI / ML model can be regarded as a request to trigger the model pairing process. Base station 102 may receive the request sent from UE 104 and receive the first PMI and the second PMI, and at step 706, the received first PMI and the received second PMI may be used by base station 102 to determine whether the first AI / ML model is paired with the second AI / ML model. For example, at step 706, base station 102 may calculate a third CSI based on the first codebook and the received first PMI, and calculate a fourth CSI based on the second AI / ML model of base station 102 and the received second PMI, and then compare the third CSI and the fourth CSI to determine whether the first AI / ML model is paired with the second AI / ML model.

[0129] In some embodiments of this disclosure, base station 102 may send a pairing result (e.g., pairing failure or pairing success) between the second AI / ML model and the first AI / ML model to UE 104, the pairing result being determined based on the above comparison. For example, if the mean square error (MSE) between the third and fourth CSIs in a predetermined time period is greater than a predetermined or pre-configured threshold, the pairing result may indicate that the second AI / ML model is not paired with the first AI / ML model. However, if the MSE is not greater than the threshold, the pairing result may indicate that the second AI / ML model is paired with the first AI / ML model, and the model ID or function ID associated with the second AI / ML model may be indicated in the pairing result.

[0130] Alternatively or additionally, upon receiving the above request from UE 104, base station 102 may configure a CSI report configuration and / or a CSI reference signal (RS) configuration for model pairing and send them to UE 104. UE 104 may estimate the CSI from the configured CSI-RS based on the CSI-RS configuration and obtain a first PMI based on the first codebook and the estimated CSI. Similarly, UE 104 may obtain a second PMI by using a corresponding first AI / ML model, and the CSI or eigenvalue (EV) of the channel matrix constructed based on the first codebook from the estimated CSI is the input to the first AI / ML model.

[0131] As another example, in some embodiments of this disclosure, the model pairing process may be initiated by base station 102 to check whether a second AI / ML model of base station 102 can be paired with an active first AI / ML model of UE 104. In this case, step 704 may include further operations or actions. For example, base station 102 may send a request to UE 104 associated with determining whether a second AI / ML model can be paired with an active first AI / ML model among at least one first AI / ML model. UE 104 may receive the request and, in response to the request, send a first PMI and a second PMI to base station 102, the second PMI being associated with the active first AI / ML model. Base station 102 may receive the first PMI and the second PMI from UE 104, and at step 706, the received first PMI and the received second PMI may be used by base station 102 to determine whether the active first AI / ML model is paired with the second AI / ML model.

[0132] After base station 102 receives the first PMI and the second PMI associated with the activated first AI / ML model, at step 706, base station 102 may calculate a third CSI based on the first codebook and the received first PMI, and calculate a fourth CSI based on the second AI / ML model and the received second PMI, and then compare the third CSI and the fourth CSI to determine whether the activated first AI / ML model is paired with the second AI / ML model. For example, base station 102 may calculate the mean square error (MSE) between the third CSI and the fourth CSI over a predetermined time period, and if the calculated MSE is greater than a threshold, determine that the second AI / ML model is not paired with the activated first AI / ML model.

[0133] In some embodiments of this disclosure, base station 102 may instruct the deactivation of the activated first AI / ML model if the second AI / ML model is determined not to be paired with the activated first AI / ML model. Using this instruction, UE 104 may deactivate the activated first AI / ML model. Subsequently, for example, if more than one first AI / ML model is available at UE 104, UE 104 may determine another activated first AI / ML model that can be paired with the second AI / ML model; or if only one first AI / ML model exists at UE 104 (i.e., the deactivated first AI / ML model) or no first AI / ML model that can be paired with the second AI / ML model, UE 104 may employ a conventional CSI compression scheme. However, if the second AI / ML model is determined to be paired with the activated first AI / ML model, base station 102 may not send any instruction or information to UE 104. In other words, if UE 104 does not receive any feedback from base station 102 after sending the first PMI and the second PMI associated with the activated first AI / ML model, UE 104 may default to determining that the second AI / ML model is paired with the activated first AI / ML model.

[0134] Alternatively or additionally, base station 102 may configure a CSI report configuration and / or a CSI RS configuration for model pairing and send them to UE 104. UE 104 may estimate the CSI from the configured CSI-RS based on the CSI-RS configuration, and then obtain a first PMI based on the first codebook and the estimated CSI. Similarly, UE 104 may obtain a second PMI by using a corresponding first AI / ML model, and the CSI or EV from the estimated CSI, based on the channel matrix constructed from the first codebook, is the input to the first AI / ML model.

[0135] In addition to being used to check whether an AI / ML model at one of the UE side and the base station side can be paired with an AI / ML model at the other of the UE side and the base station side, the model pairing process can also be used to select a model from a set of models on one side to pair with a target model on the other side.

[0136] For example, if multiple first AI / ML models exist at UE 104 and UE 104 wants to select one from them, the model pairing process can be initiated by UE 104. In this case, step 704 above can include further operations or actions. As an example, UE 104 can send a request to base station 102 to select a first AI / ML model from the multiple first AI / ML models, and send a first PMI and multiple second PMIs associated with the multiple first AI / ML models to base station 102. Base station 102 can receive the request sent from UE 104 and receive the first PMI and multiple second PMIs, which can be used to determine which of the multiple first AI / ML models can be paired with a second AI / ML model. For example, at step 706, base station 102 can select a first AI / ML model from the multiple first AI / ML models based on a third CSI and multiple fourth CSIs obtained based on the second AI / ML model and multiple second PMIs. In some embodiments of this disclosure, base station 102 may select a first AI / ML model from a plurality of first AI / ML models based on the mean square error (MSE) between the third CSI and each of the plurality of fourth CSIs in a predetermined time period. For example, base station 102 may select a first AI / ML model with the smallest MSE (which is also no greater than the threshold described above).

[0137] In some embodiments of this disclosure, base station 102 may send an indication to UE 104 regarding the selection result. If a first AI / ML model is selected from a plurality of first AI / ML models, the indication may specify the model ID of the selected first AI / ML model; if no first AI / ML model is selected, the indication may indicate a pairing failure.

[0138] Alternatively or additionally, when selecting a first AI / ML model from multiple first AI / ML models of UE 104, UE 104 may also send an indication to base station 102 indicating the number of multiple first AI / ML models. Base station 102 may receive this indication and, with further consideration of the indication, configure CSI reporting configuration and / or CSIRS configuration for model pairing, and send the CSI reporting configuration and / or CSIRS configuration to UE 104. UE 104 may estimate CSI from the configured CSI-RS based on the CSI-RS configuration, and obtain a first PMI based on the first codebook and the estimated CSI. Similarly, UE 104 may obtain each second PMI by using the corresponding first AI / ML model, and the CSI or EV from the estimated CSI, constructed based on the first codebook, is the input to the first AI / ML model.

[0139] Although the foregoing description describes UE 104 or base station 102 using a request to trigger the model pairing process, this disclosure is not limited thereto. For example, the model pairing process may be triggered periodically, or using other types of requests, or using requests from devices other than UE 104 and base station 102.

[0140] Figure 8 An example conceptual diagram illustrating dataset alignment instructions and model pairing processes according to various aspects of this disclosure is provided. Figure 8 In this context, the UE can correspond to UE 104 as described above, and the NW can correspond to base station 102 as described above.

[0141] For the datasets on both the NW side and the UE side, the configuration of the high-resolution codebook (i.e., the configuration of the first codebook or high-resolution codebook indication as described above) can be shared between the NW and UE via the high-resolution codebook indication (step 802) and used to reconstruct the datasets on both sides. The reconstructed datasets on the NW side and the UE side can be generated by extending the currently used codebook (e.g., the current Type II CSI codebook) from at least one of the following aspects: (1) increasing the number of beams, L>4; (2) increasing the number of oversamples ( O 1 , O 2(3) Increase the number of beam amplitude scaling factors for both broadband and subband; and (4) increase the number of beam combination coefficients (phase) in the beam, polarization, and layers. The above aspects are merely examples, and the reconstructed dataset can be generated by extending the currently used codebook from other aspects, and there is no limitation on this in this disclosure. Furthermore, scaling factors can be applied to the corresponding numbers above to increase the resolution of the codebook, which means that the NR Rel-15 Type II codebook can be extended by extending RI∈{1,2,3,4} as the default codebook. In this way, the complex redesign of the codebook and PMI value calculation assumptions can be avoided. However, in addition to scaling factors, the configuration of the high-resolution codebook (which may also be referred to as codebook configuration below) can indicate some other ways to obtain the high-resolution codebook, and there is no limitation on this in this disclosure. Furthermore, the codebook configuration can be function- or model-specific.

[0142] Using a shared high-resolution codebook indicated by a high-resolution codebook indicator, both datasets can be reconstructed. However, in some embodiments of this disclosure, with Figure 8 Unlike other systems, the NW-side dataset can be pre-reconstructed to share high-resolution codebook indications with the UE. For example, the NW-side codebook configuration can be indicated directly as function / model-related information during AI / ML functions or model identification (which will be interpreted later). Alternatively, the NW-side codebook configuration can be sent from the NW to the UE as a higher-level parameter (different parameter sets can be associated with different codebook configuration IDs) when the pairing process is triggered by the NW or UE.

[0143] Based on the reconstructed datasets from both sides, the PMI corresponding to the configured high-resolution codebook can be generated by the UE and reported to the NW for the model pairing process (step 804). In some embodiments of this disclosure, a dedicated CSI-RS configuration and CSI reporting configuration can be configured for model pairing, enabling the UE to report the PMI corresponding to the high-resolution codebook.

[0144] continue Figure 8 In the model pairing process (step 804), on the UE side, the compressed value (i.e., the output PMI from the UE partial model, for example, a set) PMI AI ) can be generated and compared with PMI (e.g., a set) from a high-resolution codebook. PMI hrcb Together, they are reported to the NW. The inputs to the UE part of the model can be based on the reported data. PMI hrcb CSI / EV is constructed using a "channel matrix" consisting of a high-resolution codebook and a configured high-resolution codebook. PMI AIand PMI hrcb Paired. And reported together via L1 (UCI), L2 (MAC CE) or L3 (RRC), for example, in the same UCI report.

[0145] Furthermore, on the NW side, the CSI / EV recovered from the NW partial model and the high-resolution codebook can be compared. The input to the NW partial model is the received compressed value, i.e., the value from the UE partial model. PMI AI The reference CSI / EV is based on the received data. PMI hrcb The "channel matrix" is obtained by constructing a high-resolution codebook and configuration. The comparison between these two sets of CSI / EV can be achieved by calculating the output of the NW portion of the model within a given or predetermined time period and based on a set of received signals. PMI hrcb The MSE is derived from the obtained reference CSI / EV. For example, if the calculated MSE is greater than a threshold (e.g., a predetermined or pre-configured threshold), model pairing fails, followed by an indication of this failure to the UE. Otherwise, a successful pairing can be indicated to the UE along with the function ID or model ID of the NW part model. As another example, if the MSE is not greater than the threshold, the NW can select the model with the smallest MSE as the NW part model for requesting pairing with the UE's part model. The selection result can be indicated to the UE along with the function ID or model ID of the selected model.

[0146] Figure 9 An example process 900 is illustrated in which, according to various aspects of this disclosure, a model pairing is performed between a base station 102 (which may also be referred to below as NW 102 or gNB 102) and a UE 104 via an air interface using an aligned dataset.

[0147] like Figure 9 As shown, process 900 can be initiated after a function / model identifier (indicating relevant features for supporting AI / ML-based CSI compression using a dual-side model) between base station 102 and UE 104. In the function / model identifier, relevant information about the AI / ML function / model is shared and aligned between the base station and UE sides, and includes at least the following: (1) features or feature groups indicated by UE 104 for supporting AI / ML-based CSI compression, such as the number of supported ports, subbands, class, and quantization level of the output of the UE partial model in the UE capability report; and (2) the applicable conditions for the AI / ML function / model, such as the applicable signal-to-noise ratio (SNR), UE speed, etc. This information can be included in the UE capability report, UE auxiliary information request, and / or other new dedicated RRC signaling.

[0148] Following the function / model identification, at step 901, base station 102 may instruct a high-resolution codebook configuration (which may also be referred to as a configuration about the high-resolution codebook) for UE 104 to reconstruct the dataset. In this step, the configuration about the high-resolution codebook for reconstructing the dataset may be instructed by base station 102 to UE 104 based on the identified function / model on the UE side, and it may be based on the current codebook used by UE 104 and base station 102 and may be a value calculation assumption. For example, the configuration about the high-resolution codebook may instruct at least one of the following: (1) scaling the number of beams, for example, regarding numberOfBeams (1) Two or three times the current configuration; (2) Scaling the number of phase sizes, for example, regarding phaseAlphabetSize (2) Two or four times the current configuration; (3) Enable subband amplitude quantization; and (4) Introduce an additional oversampling factor. O 1 , O 2 The above information regarding the configuration of the high-resolution codebook is merely an example, and the configuration may include more or different content used to expand the codebook. For example, the amplitude values ​​of the broadband and subband can also be scaled by a factor of two via adding one bit to each indication, which can also be indicated via RRC signaling to enable quantization with an additional bit. Furthermore, the above scaling factor can be indicated via RRC signaling, which can be used in the current... CodebookConfig Information elements, or new RRC IEs (e.g., HRSCodebook ConfigForModelPairing This is involved in ( ).

[0149] Using an aligned high-resolution codebook, there are at least two cases for performing model pairing: (1) checking whether a model on one side can be paired with a model on the other side; and (2) selecting a model from a set of models on one side to pair with other models on the other side.

[0150] Figure 10 The illustration shows a schematic diagram of a method for checking whether a model on one side can be paired with a model on the other side, according to various aspects of this disclosure. Figure 11 The illustration shows a schematic diagram of a method for selecting a model from a set of models on one side to pair with a target model on the other side, according to various aspects of this disclosure.

[0151] In cases where it is necessary to check whether a model on one side can be paired with a model on the other side (1), the model pairing process can be initiated by different sides. For example, if a model in UE 104 is newly deployed, updated, or needs to be monitored, UE 104 can initiate the model pairing process using a request to evaluate the model (step 902b), which can be considered case (1a). Furthermore, if a model in NW 102 is newly deployed, updated, or needs to be monitored, NW 102 can initiate the model pairing process using a request to evaluate the model (step 902a), which can be considered case (1b), and in this case, the NW102 model ID or function can be indicated to UE 104, for example, via the request, so that UE 104 can use a suitable model to generate CSI.

[0152] After initiation, NW 102 can send a configured CSI-RS for CSI measurement, used for model pairing (step 903), and UE 104 can estimate CSI based on the configured CSI-RS (step 903). Based on a high-resolution codebook, PMI hrcb It can be exported together with the regenerated (i.e., quantized) CSI (step 904). The former can be reported to NW 102, and the latter can be fed as input into the UE partial model to generate PMI AI Then, including { PMI hrcb , PMI AI The CSI can be reported to NW 102 (step 905), as follows. Figure 10 As shown in the image.

[0153] like Figure 10 As shown, upon receiving { PMI hrcb , PMI AI Following that, CSI from the high-resolution codebook And the CSI recovered from the NW part of the AI ​​model. The MSE is exported for comparison (step 906). The following MSE can be used as an example for comparison: Deviation =

[0154] If the calculated deviation is less than a predetermined or pre-configured threshold, then for case (1a), i.e., the model pairing process is initiated by UE 104, an indication can be sent to UE 104 to indicate successful pairing of the requested model (step 907). Otherwise, an indication is sent to UE 104 to indicate pairing failure (step 907). Furthermore, the model ID can be assigned for additional use of the requested model.

[0155] Furthermore, in case (1b), where the model pairing process is initiated from NW 102, if the calculated deviation is less than a predetermined or pre-configured threshold, no message is sent from NW 102 to UE 104. Otherwise, an indication to deactivate the current model can be sent from base station 102 to UE 104 to indicate pairing failure (step 907).

[0156] Similarly, in the case of selecting a model from a set of models on one side to pair with a target model on the other side (2), the model pairing process can be initiated by different sides. For example, UE 104 may want to select a model from a set of models to pair with a specific model or function on the NW side (this can be considered case (2a)), and therefore the model pairing process for this case can be initiated by UE 104. Similarly, NW 102 may want to select a model from a set of models to pair with a function with a specific identifier or a specific model on the UE side (this can be considered case (2b)), and therefore the model pairing process for this case can be initiated by NW 102.

[0157] Case (2a) may introduce new signaling designs, such as Figure 11 As shown in the diagram. First, UE 104 initiates a model pairing process for model selection, which may include a set of model sizes (i.e., the number of candidate models on the UE side) to align with the UCI report format / size (step 902b). The set of model sizes may be included in the request sent at step 902b to trigger the model pairing process; alternatively, at step 902b, the indication of the set of model sizes may be sent separately from the request to trigger the model pairing process. Base station 102 may configure CSI-RS and CSI reports to indicate the number of PMIs from candidate models (corresponding to the set of model sizes) (step 903). Additionally, different CSI-RS resources may be configured to be associated with different models on the UE side.

[0158] UE 104 can estimate CSI based on the configured CSI-RS. Based on a high-resolution codebook, PMI hrcbIt can be exported together with the regenerated (i.e., quantized) CSI (step 904). The former can be reported to NW 102, and the latter can be fed as input into the model on UE 104 to generate the output of the corresponding model. PMI AI Then, including { PMI hrcb , PMI AI,1 , PMI AI,2 , ..., PMI AI,k The CSI can be reported to NW 104 (step 905), as follows. Figure 11 As shown in the image.

[0159] In addition, refer to Figure 11 On the NW side, upon receiving { from UE 104 PMI hrcb , PMI AI,1 , PMI AI,2 , ..., PMI AI,k Following that, CSI from the high-resolution codebook And the CSI recovered from the NW part of the AI ​​model. The result is exported (step 906) to further calculate the following deviation:

[0160] have The model with the smallest deviation (which is also less than a predetermined or pre-configured threshold) can be selected as the pairing model on the UE side.

[0161] In case (2a), where the model pairing process is initiated by UE 104, if a paired model is selected, an indication of the selected model (e.g., model ID) can be sent to UE 104 (step 907). Otherwise, if no model is selected, an indication of pairing failure can be sent to UE 104 (step 907).

[0162] Similarly, in case (2b), where the model pairing process is initiated from NW 102 and no message is sent to UE 104, if no model is selected by NW 102, an indication to deactivate the currently activated model at UE 104 can be sent from NW 102 to UE 104 to indicate selection failure (step 907). UE 104 can then release the (multiple) UE partial models corresponding to the indicated deactivated model / function.

[0163] Furthermore, the process described in (1) above can be repeated in other scenarios, such as model monitoring to check whether the performance of the deployed model is still good enough. In addition, the embodiments described above are also applicable to other situations to perform model pairing using similar operations.

[0164] Figure 12 An example of device 1200 supporting model pairing for AI / ML-based CSI compression according to various aspects of this disclosure is illustrated. Device 1200 may be an example of UE 104 as described herein. Device 1200 may support wireless communication with one or more network entities 102, UE 104, or any combination thereof. Device 1200 may include components for bidirectional communication, including components for transmitting and receiving communications, such as processor 1202, memory 1204, transceiver 1206, and optional I / O controller 1208. These components may communicate electronically or be otherwise coupled (e.g., operative ground, communicative ground, functional ground, electronic ground, electrical ground) via one or more interfaces (e.g., bus).

[0165] Processor 1202, memory 1204, transceiver 1206, or various combinations thereof or various components thereof may be examples of components for performing aspects of the present disclosure as described herein. For example, processor 1202, memory 1204, transceiver 1206, or various combinations thereof or components thereof may support methods for performing one or more operations described herein.

[0166] In some implementations, processor 1202, memory 1204, transceiver 1206, or various combinations or components thereof may be implemented in hardware (e.g., in a communication management circuitry system). The hardware may include a processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof, configured to or otherwise supporting components for performing the functions described in this disclosure. In some implementations, processor 1202 and memory 1204 coupled to processor 1202 may be configured to perform one or more functions described herein (e.g., instructions stored in memory 1204 are executed by processor 1202).

[0167] For example, processor 1202 may support wireless communication at device 1200 according to the examples disclosed herein. Processor 1202 may be configured to operate to support: components for receiving configuration of a first codebook from base station 102 via transceiver 1206; and components for transmitting a first precoding matrix indicator (PMI) and at least one second PMI to base station 102 via transceiver 1206, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

[0168] Processor 1202 may include intelligent hardware devices (e.g., general-purpose processors, DSPs, CPUs, microcontrollers, ASICs, FPGAs, programmable logic devices, discrete gate or transistor logic components, discrete hardware components, or any combination thereof). In some implementations, processor 1202 may be configured to use a memory controller to operate a memory array. In some other implementations, the memory controller may be integrated into processor 1202. Processor 1202 may be configured to execute computer-readable instructions stored in memory (e.g., memory 1204) to cause device 1200 to perform various functions of this disclosure.

[0169] Memory 1204 may include random access memory (RAM) and read-only memory (ROM). Memory 1204 may store computer-readable, computer-executable code including instructions that, when executed by processor 1202, cause device 1200 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium such as system memory or another type of memory. In some implementations, the code may not be directly executed by processor 1202, but may cause a computer (e.g., when compiled and executed) to perform the functions described herein. In some implementations, memory 1204 may include a basic I / O system (BIOS) or similar system that controls basic hardware or software operations, such as interaction with peripheral components or devices.

[0170] I / O controller 1208 can manage input and output signals for device 1200. I / O controller 1208 can also manage peripheral devices not integrated into device M02. In some implementations, I / O controller 1208 can represent a physical connection or port to an external peripheral device. In some implementations, I / O controller 1208 can utilize an operating system such as iOS®, ANDROID®, MS-WINDOWS®, OS / 2®, UNIX®, LINUX®, or another known operating system. In some implementations, I / O controller 1208 can be implemented as part of a processor (such as processor 1206). In some implementations, a user can interact with device 1200 via I / O controller 1208 or via hardware components controlled by I / O controller 1208.

[0171] In some implementations, device 1200 may include a single antenna 1210. However, in other implementations, device 1200 may have more than one antenna 1210 (i.e., multiple antennas), including multiple antenna panels or antenna arrays capable of simultaneously transmitting or receiving multiple wireless transmissions. Transceiver 1206 may communicate bidirectionally via one or more antennas 1210, a wired link, or a wireless link as described herein. For example, transceiver 1206 may represent a wireless transceiver and may communicate bidirectionally with another wireless transceiver. Transceiver 1206 may also include a modem for modulating packets, providing modulated packets to one or more antennas 1210 for transmission, and demodulating packets received from one or more antennas 1210. Transceiver 1206 may include one or more transmit chains, one or more receive chains, or combinations thereof.

[0172] The transmission chain can be configured to generate and transmit signals (e.g., control information, data, packets). The transmission chain may include at least one modulator for modulating data onto a carrier signal to prepare a signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques, such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes such as phase shift keying (PSK) or quadrature amplitude modulation (QAM). The transmission chain may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level suitable for transmission over a wireless medium. The transmission chain may also include one or more antennas 1210 for transmitting the amplified signal into the air or the wireless medium.

[0173] The receiver chain can be configured to receive signals (e.g., control information, data, packets) via a wireless medium. For example, the receiver chain may include one or more antennas 1210 for receiving signals over the air or via a wireless medium. The receiver chain may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. The receiver chain may include at least one demodulator configured to demodulate the received signal and obtain the transmitted data by reversing the modulation technique applied during signal transmission. The receiver chain may include at least one decoder for decoding the demodulated signal to receive the transmitted data.

[0174] Figure 13 An example of device 1300 supporting model pairing for AI / ML-based CSI compression according to various aspects of this disclosure is illustrated. Device 1300 may be an example of base station 104 as described herein, and base station 102 may include multiple TRPs. Device 1300 may support wireless communication with one or more network entities 102, UE 104, or any combination thereof. Device 1300 may include components for bidirectional communication, including components for transmitting and receiving communications, such as processor 1302, memory 1304, transceiver 1306, and optional I / O controller 1308. These components may communicate electronically or be otherwise coupled (e.g., operative ground, communicative ground, functional ground, electronic ground, electrical ground) via one or more interfaces (e.g., bus).

[0175] Processor 1302, memory 1304, transceiver 1306, or various combinations thereof, or various components thereof, may be examples of components for performing aspects of the present disclosure as described herein. For example, processor 1302, memory 1304, transceiver 1306, or various combinations thereof, or components thereof, may support methods for performing one or more operations described herein.

[0176] In some implementations, processor 1302, memory 1304, transceiver 1306, or various combinations or components thereof may be implemented in hardware (e.g., in a communication management circuitry system). The hardware may include a processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof, configured to or otherwise supporting components for performing the functions described in this disclosure. In some implementations, processor 1302 and memory 1304 coupled to processor 1302 may be configured to perform one or more functions described herein (e.g., instructions stored in memory 1304 are executed by processor 1302).

[0177] For example, processor 1302 may support wireless communication at device 1300 according to the examples disclosed herein. Processor 1302 may be configured to operate to support: components for transmitting configuration of a first codebook to user equipment (UE) 104 via transceiver 1306; components for receiving a first precoding matrix indicator (PMI) and at least one second PMI from UE 104 via transceiver 1306, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model; and components for determining whether a second AI / ML model of a base station can be paired with at least one first AI / ML model based on a third CSI and at least one fourth CSI, the third CSI being obtained based on the first codebook and the received first PMI, the at least one fourth CSI being obtained based on the second AI / ML model and the received at least one second PMI.

[0178] Processor 1302 may include intelligent hardware devices (e.g., general-purpose processors, DSPs, CPUs, microcontrollers, ASICs, FPGAs, programmable logic devices, discrete gate or transistor logic components, discrete hardware components, or any combination thereof). In some implementations, processor 1302 may be configured to use a memory controller to operate a memory array. In some other implementations, the memory controller may be integrated into processor 1302. Processor 1302 may be configured to execute computer-readable instructions stored in memory (e.g., memory 1304) to cause device 1300 to perform various functions of this disclosure.

[0179] Memory 1304 may include random access memory (RAM) and read-only memory (ROM). Memory 1304 may store computer-readable, computer-executable code including instructions that, when executed by processor 1302, cause device 1300 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium such as system memory or another type of memory. In some implementations, the code may not be directly executed by processor 1302, but may cause a computer (e.g., when compiled and executed) to perform the functions described herein. In some implementations, memory 1304 may include a basic I / O system (BIOS) or similar system that controls basic hardware or software operations, such as interaction with peripheral components or devices.

[0180] I / O controller 1308 can manage input and output signals for device 1300. I / O controller 1308 can also manage peripheral devices not integrated into device M02. In some implementations, I / O controller 1308 can represent a physical connection or port to an external peripheral device. In some implementations, I / O controller 1308 can utilize an operating system such as iOS®, ANDROID®, MS-WINDOWS®, OS / 2®, UNIX®, LINUX®, or another known operating system. In some implementations, I / O controller 1308 can be implemented as part of a processor (such as processor 1306). In some implementations, a user can interact with device 1300 via I / O controller 1308 or via hardware components controlled by I / O controller 1308.

[0181] In some implementations, device 1300 may include a single antenna 1310. However, in other implementations, device 1300 may have more than one antenna 1310 (i.e., multiple antennas), including multiple antenna panels or antenna arrays capable of simultaneously transmitting or receiving multiple wireless transmissions. Transceiver 1306 may communicate bidirectionally via one or more antennas 710, a wired link, or a wireless link as described herein. For example, transceiver 1306 may represent a wireless transceiver and may communicate bidirectionally with another wireless transceiver. Transceiver 1306 may also include a modem for modulating packets, providing modulated packets to one or more antennas 1310 for transmission, and demodulating packets received from one or more antennas 1310. Transceiver 1306 may include one or more transmit chains, one or more receive chains, or combinations thereof.

[0182] The transmission chain can be configured to generate and transmit signals (e.g., control information, data, packets). The transmission chain may include at least one modulator for modulating data onto a carrier signal to prepare a signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques, such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes such as phase shift keying (PSK) or quadrature amplitude modulation (QAM). The transmission chain may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level for transmission over a wireless medium. The transmission chain may also include one or more antennas 1310 for transmitting the amplified signal into the air or wireless medium.

[0183] The receiver chain can be configured to receive signals (e.g., control information, data, packets) via a wireless medium. For example, the receiver chain may include one or more antennas 1310 for receiving signals over the air or via a wireless medium. The receiver chain may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. The receiver chain may include at least one demodulator configured to demodulate the received signal and obtain the transmitted data by reversing the modulation technique applied during signal transmission. The receiver chain may include at least one decoder for decoding the demodulated signal to receive the transmitted data.

[0184] Figure 14 An example of a processor 1400 supporting model pairing for AI / ML-based CSI compression according to various aspects of this disclosure is illustrated. Processor 1400 may be an example of a processor configured to perform various operations according to the examples described herein. Processor 1400 may include a controller 1402 configured to perform various operations according to the examples described herein. Processor 1400 may optionally include at least one memory 1404, such as an L1 / L2 / L3 cache. Additionally or alternatively, processor 1400 may optionally include one or more arithmetic logic units (ALUs) 1400. One or more of these components may be electronically communicated or otherwise coupled (e.g., operatively, communicatively, functionally, electronically, electrically) via one or more interfaces (e.g., buses).

[0185] Processor 1400 may be a processor chipset and includes a protocol stack (e.g., a software stack) executed by the processor chipset to perform various operations (e.g., receive, acquire, retrieve, send, output, forward, store, determine, identify, access, write, read) according to the examples described herein. The processor chipset may include one or more cores, one or more caches (e.g., memory located locally in or included within the processor chipset (e.g., processor 1400), or other memory (e.g., random access memory (RAM), read-only memory (ROM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), static RAM (SRAM), ferroelectric RAM (FeRAM), magnetoresistive RAM (MRAM), resistive RAM (RRAM), flash memory, phase-change memory (PCM), etc.).

[0186] Controller 1402 can be configured to manage and coordinate various operations of processor 1400 (e.g., signaling, receiving, acquiring, retrieving, transmitting, outputting, forwarding, storing, determining, identifying, accessing, writing, and reading) to enable processor 1400 to support various operations of a base station according to the examples described herein. For example, controller 1402 can operate as a control unit of processor 1400 to generate control signals that manage the operation of various components of processor 1400. These control signals include enabling or disabling functional units, selecting data paths, initiating memory accesses, and coordinating the timing of operations.

[0187] Controller 1402 can be configured to fetch (e.g., retrieve, retrieve, receive) instructions from memory 1404 and determine subsequent instructions(s) to be executed, such that processor 1400 supports various operations according to the examples described herein. Controller 1402 can be configured to track the memory addresses of instructions associated with memory 1404. Controller 1402 can be configured to decode instructions to determine the operations to be performed and the operands involved. For example, controller 1402 can be configured to interpret instructions and determine control signals that will be output to other components of processor 1400, such that processor 1400 supports various operations according to the examples described herein. Additionally or alternatively, controller 1402 can be configured to manage data flow within processor 1400. Controller 1402 can be configured to control data transfers between registers, arithmetic logic unit (ALU), and other functional units of processor 1400.

[0188] Memory 1404 may include one or more caches (e.g., memory located locally on or included in the processor 1400) or other memories (such as RAM, ROM, DRAM, SDRAM, SRAM, MRAM, flash memory, etc.). In some implementations, memory 1404 may reside within or on the processor chipset (e.g., locally on the processor 1400). In some other implementations, memory 1404 may reside outside the processor chipset (e.g., remotely from the processor 1400).

[0189] Memory 1404 may store computer-readable, computer-executable code including instructions that, when executed by processor 1400, cause processor 1400 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium such as system memory or another type of memory. Controller 1402 and / or processor 1400 may be configured to execute the computer-readable instructions stored in memory 1404 to cause processor 1400 to perform various functions. For example, processor 1400 and / or controller 1402 may be coupled to or coupled to memory 1404, and processor 1400, controller 1402, and memory 1404 may be configured to perform the various functions described herein. In some examples, processor 1400 may include multiple processors, and memory 1404 may include multiple memories. One or more of the multiple processors may be coupled to one or more of the multiple memories, which may be configured individually or collectively to perform the various functions described herein.

[0190] One or more ALU 1400s can be configured to support various operations according to the examples described herein. In some implementations, one or more ALU 1400s may reside within or on a processor chipset (e.g., processor 1400). In some other implementations, one or more ALU 1400s may reside outside the processor chipset (e.g., processor 1400). One or more ALU 1400s can perform one or more computations on data, such as addition, subtraction, multiplication, and division. For example, one or more ALU 1400s can receive input operands and an opcode that determines the operation to be performed. One or more ALU 1400s are configured with various logic and arithmetic circuitry (including adders, subtractors, shifters, and logic gates) to process and manipulate data according to the operations. Alternatively or concurrently, one or more ALU 1400s may support logical operations such as AND, OR, XOR, NOR, and NAND, enabling one or more ALU 1400s to handle conditional operations, comparisons, and bitwise operations.

[0191] Processor 1400 may support wireless communication according to the examples disclosed herein. Processor 1400 may be configured or operable to support: components for receiving configuration of a first codebook from a base station via a transceiver; and components for transmitting a first precoding matrix indicator (PMI) and at least one second PMI to the base station via a transceiver, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

[0192] Figure 15 An example of a processor 1500 supporting model pairing for AI / ML-based CSI compression according to various aspects of this disclosure is illustrated. Processor 1500 may be an example of a processor configured to perform various operations according to the examples described herein. Processor 1500 may include a controller 1502 configured to perform various operations according to the examples described herein. Processor 1500 may optionally include at least one memory 1504, such as an L1 / L2 / L3 cache. Additionally or alternatively, processor 1500 may optionally include one or more arithmetic logic units (ALUs) 1500. One or more of these components may be electronically communicated or otherwise coupled (e.g., operatively, communicatively, functionally, electronically, electrically) via one or more interfaces (e.g., buses).

[0193] Processor 1500 may be a processor chipset and includes a protocol stack (e.g., a software stack) executed by the processor chipset to perform various operations (e.g., receive, acquire, retrieve, send, output, forward, store, determine, identify, access, write, read) according to the examples described herein. The processor chipset may include one or more cores, one or more caches (e.g., memory located locally in or included within the processor chipset (e.g., processor 1500), or other memory (e.g., random access memory (RAM), read-only memory (ROM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), static RAM (SRAM), ferroelectric RAM (FeRAM), magnetoresistive RAM (MRAM), resistive RAM (RRAM), flash memory, phase-change memory (PCM), etc.).

[0194] Controller 1502 can be configured to manage and coordinate various operations of processor 1500 (e.g., signaling processing, receiving, acquiring, retrieving, transmitting, outputting, forwarding, storing, determining, identifying, accessing, writing, and reading) to enable processor 1500 to support various operations of the UE according to the examples described herein. For example, controller 1502 can operate as a control unit of processor 1500 to generate control signals that manage the operation of various components of processor 1500. These control signals include enabling or disabling functional units, selecting data paths, initiating memory accesses, and coordinating the timing of operations.

[0195] Controller 1502 can be configured to fetch (e.g., retrieve, retrieve, receive) instructions from memory 1504 and determine subsequent instructions(s) to be executed, enabling processor 1500 to support various operations according to the examples described herein. Controller 1502 can be configured to track the memory addresses of instructions associated with memory 1504. Controller 1502 can be configured to decode instructions to determine the operations to be performed and the operands involved. For example, controller 1502 can be configured to interpret instructions and determine control signals that will be output to other components of processor 1500, enabling processor 1500 to support various operations according to the examples described herein. Additionally or alternatively, controller 1502 can be configured to manage data flow within processor 1500. Controller 1502 can be configured to control data transfers between registers, arithmetic logic unit (ALU), and other functional units of processor 1500.

[0196] Memory 1504 may include one or more caches (e.g., memory located locally on or included in the processor 1500) or other memories (such as RAM, ROM, DRAM, SDRAM, SRAM, MRAM, flash memory, etc.). In some implementations, memory 1504 may reside within or on the processor chipset (e.g., locally on the processor 1500). In some other implementations, memory 1504 may reside outside the processor chipset (e.g., remotely from the processor 1500).

[0197] Memory 1504 may store computer-readable, computer-executable code including instructions that, when executed by processor 1500, cause processor 1500 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium such as system memory or another type of memory. Controller 1502 and / or processor 1500 may be configured to execute the computer-readable instructions stored in memory 1504 to cause processor 1500 to perform various functions. For example, processor 1500 and / or controller 1502 may be coupled to or coupled to memory 1504, and processor 1500, controller 1502, and memory 1504 may be configured to perform the various functions described herein. In some examples, processor 1500 may include multiple processors, and memory 1504 may include multiple memories. One or more of the multiple processors may be coupled to one or more of the multiple memories, which may be configured individually or collectively to perform the various functions described herein.

[0198] One or more ALU 1500s can be configured to support various operations according to the examples described herein. In some implementations, one or more ALU 1500s may reside within or on a processor chipset (e.g., processor 1500). In some other implementations, one or more ALU 1500s may reside outside the processor chipset (e.g., processor 1500). One or more ALU 1500s can perform one or more computations on data, such as addition, subtraction, multiplication, and division. For example, one or more ALU 1500s can receive input operands and an opcode that determines the operation to be performed. One or more ALU 1500s are configured with various logic and arithmetic circuitry (including adders, subtractors, shifters, and logic gates) to process and manipulate data according to the operations. Alternatively or concurrently, one or more ALU 1500s may support logical operations such as AND, OR, XOR, NOR, and NAND, enabling one or more ALU 1500s to handle conditional operations, comparisons, and bitwise operations.

[0199] Processor 1500 may support wireless communication according to the examples disclosed herein. Processor 1500 may be configured or operable to support: components for transmitting configuration of a first codebook to a user equipment (UE) via a transceiver; components for receiving a first precoding matrix indicator (PMI) and at least one second PMI from the UE via the transceiver, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model; and components for determining whether a second AI / ML model of a base station can be paired with at least one first AI / ML model based on a third CSI and at least one fourth CSI, the third CSI being obtained based on the first codebook and the received first PMI, the at least one fourth CSI being obtained based on the second AI / ML model and the received at least one second PMI.

[0200] Figure 16 A flowchart illustrating a method 1600 supporting model pairing for AI / ML-based CSI compression according to various aspects of this disclosure is provided. Various operations of method 1600 can be implemented by a device or components thereof as described herein. For example, operations of method 1600 can be performed by a UE 104 as described herein. In some implementations, the device can execute a set of instructions to control functional elements of the device to perform the described functions. Alternatively or additionally, the device can use dedicated hardware to perform aspects of the described functions.

[0201] At 1605, the method may include: receiving configuration information about the first codebook from the base station via a transceiver. The operation of 1605 can be performed according to the examples described herein. In some implementations, aspects of the operation of 1605 can be derived from references... Figure 1 The device described is used to perform this action.

[0202] At 1610, the method may include: transmitting a first precoding matrix indicator (PMI) and at least one second PMI to a base station via a transceiver, the first PMI being obtained based on a first codebook and first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model. The operation of 1610 can be performed according to the examples described herein. In some implementations, aspects of the operation of 1610 may be derived from references... Figure 1 The device described is used to perform this action.

[0203] Figure 17 A flowchart illustrating a method 1700 supporting model pairing for AI / ML-based CSI compression according to various aspects of this disclosure is provided. Various operations of method 1700 can be implemented by devices or components thereof as described herein. For example, various operations of method 1700 can be performed by a base station 104 as described herein. In some implementations, the device can execute a set of instructions to control the functional elements of the device to perform the described functions. Alternatively or additionally, the device can use dedicated hardware to perform aspects of the described functions.

[0204] At 1705, the method may include: sending a configuration of the first codebook to the user equipment (UE) via a transceiver. The operation of 1705 can be performed according to the examples described herein. In some implementations, aspects of the operation of 1705 can be found in the references... Figure 1 The device described is used to perform this action.

[0205] At 1710, the method may include: receiving from the UE via a transceiver a first precoding matrix indicator (PMI) and at least one second PMI, the first PMI being obtained based on a first codebook and first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model. The operation of 1710 can be performed according to the examples described herein. In some implementations, aspects of the operation of 1710 may be derived from references... Figure 1 The device described is used to perform this action.

[0206] At 1715, the method may include: determining whether a second AI / ML model of the base station can be paired with at least one first AI / ML model based on a third CSI and at least one fourth CSI, wherein the third CSI is obtained based on a first codebook and a received first PMI, and the at least one fourth CSI is obtained based on a second AI / ML model and at least one received second PMI. The operation of 1715 can be performed according to the examples described herein. In some implementations, aspects of the operation of 1715 may be derived from references... Figure 1 The device described is used to perform this action.

[0207] It should be noted that the methods described in this paper describe possible implementations, and the operations and steps can be rearranged or otherwise modified, and other implementations are possible. Furthermore, aspects from two or more of these methods can be combined.

[0208] The various illustrated blocks and components described in this disclosure can be implemented or performed using the following devices designed to perform the functions described herein: general-purpose processors, DSPs, ASICs, CPUs, FPGAs or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination thereof. A general-purpose processor may be a microprocessor, but alternatively, the processor may be any processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices (e.g., a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors combined with a DSP core, or any other such configuration).

[0209] The functions described herein can be implemented in hardware, software executed by a processor, firmware, or any combination thereof. If implemented in software executed by a processor, these functions can be stored on or transmitted via a computer-readable medium as one or more instructions or code. Other examples and implementations are also within the scope of this disclosure and the appended claims. For example, due to the nature of software, the functions described herein can be implemented using software executed by a processor, hardware, firmware, hardwired, or any combination thereof. Features implementing the functions can also be physically located in various locations, including being distributed such that portions of the function are implemented in different physical locations.

[0210] Computer-readable media include both non-transitory computer storage media and communication media, including any media that facilitates the transfer of a computer program from one place to another. Non-transitory storage media can be any available medium that can be accessed by a general-purpose computer or a special-purpose computer. For example, non-transitory computer-readable media can include RAM, ROM, electrically erasable programmable ROM (EEPROM), flash memory, optical disc (CD) ROM or other optical disc storage, magnetic disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a general-purpose computer or a special-purpose computer, or a general-purpose processor or a special-purpose processor.

[0211] As used herein and included in the claims, the article “a” preceding an element is not limited and is understood to mean “at least one” or “one or more” of those elements. The terms “a,” “at least one,” “one or more,” and “at least one of one or more” are used interchangeably. As used herein and included in the claims, “or” as used in a list of items (e.g., a list of items prefixed with phrases such as “at least one of…” or “one or more of…” or “one or two of…”) indicates an inclusive list, such that a list of at least one of, for example, A, B, or C means A or B or C or AB or AC or BC or ABC (i.e., A, B, and C). Furthermore, as used herein, the phrase “based on” should not be construed as a reference to a closed set of conditions. For example, an example step described as “based on condition A” could be based on both condition A and condition B without departing from the scope of this disclosure. In other words, as used herein, the phrase “based on” should be interpreted in the same manner as the phrase “at least partially based on.” Furthermore, as used herein and included in the claims, “set” can include one or more elements.

[0212] The description herein is provided to enable those skilled in the art to make or use this disclosure. Various modifications to this disclosure will be apparent to those skilled in the art, and the general principles defined herein can be applied to other variations without departing from the scope of this disclosure. Therefore, this disclosure is not limited to the examples and designs described herein, but should be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A user equipment (UE), comprising: processor; as well as A transceiver, which is coupled to the processor, The processor is configured as follows: Receive configuration information about the first codebook from the base station via the transceiver; and The transceiver transmits a first precoding matrix indicator (PMI) and at least one second PMI to the base station, the first PMI being obtained based on the first codebook and the first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

2. The UE of claim 1, wherein sending the first PMI and the at least one second PMI comprises: The transceiver sends a request to the base station to determine whether the first AI / ML model in the at least one first AI / ML model can be paired with the second AI / ML model of the base station; as well as The first PMI and the second PMI are transmitted to the base station via the transceiver, wherein the second PMI is associated with the first AI / ML model in the at least one first AI / ML model.

3. The UE according to claim 2, wherein the processor is further configured to: The transceiver receives the pairing result of the second AI / ML model and the first AI / ML model from the base station.

4. The UE of claim 1, wherein sending the first PMI and the at least one second PMI comprises: The transceiver receives from the base station a request associated with determining whether the base station's second AI / ML model can be paired with an active first AI / ML model among the at least one first AI / ML model, and In response to the request, the first PMI and the second PMI are sent to the base station via the transceiver, the second PMI being associated with the activated first AI / ML model.

5. The UE of claim 1, wherein the at least one first AI / ML model comprises a plurality of first AI / ML models, and the at least one second PMI comprises a plurality of second PMIs associated with the plurality of first AI / ML models, and Sending the first PMI and the at least one second PMI includes: The transceiver sends a request to the base station for selecting a first AI / ML model from the plurality of first AI / ML models; as well as The first PMI and the plurality of second PMIs are transmitted to the base station via the transceiver.

6. The UE of claim 5, wherein the processor is further configured to: receive an indication of the selection result from the base station via the transceiver. When a first AI / ML model is selected by the base station from the plurality of first AI / ML models, the indication specifies the model identifier (ID) of the selected first AI / ML model, and If no first AI / ML model is selected by the base station from the plurality of first AI / ML models, the indication indicates that pairing has failed.

7. The UE of claim 1, wherein the configuration regarding the first codebook is transmitted as function / model-related information during a function or model identification process between the UE and the base station, or The configuration of the first codebook is transmitted via Radio Resource Control (RRC) signaling.

8. A base station, comprising: processor; as well as A transceiver, which is coupled to the processor, The processor is configured as follows: The transceiver transmits the configuration of the first codebook to the user equipment (UE); The transceiver receives a first precoding matrix indicator (PMI) and at least one second PMI from the UE, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model; and Based on a third CSI and at least one fourth CSI, it is determined whether the second AI / ML model of the base station can be paired with the at least one first AI / ML model. The third CSI is obtained based on the first codebook and the received first PMI, and the at least one fourth CSI is obtained based on the second AI / ML model and the received at least one second PMI.

9. The base station of claim 8, wherein receiving the first PMI and the at least one second PMI comprises: The receiver receives a request from the UE to determine whether a first AI / ML model in the at least one first AI / ML model can be paired with a second AI / ML model; as well as The first PMI and the second PMI are received from the UE via the transceiver, wherein the second PMI is associated with the first AI / ML model in the at least one first AI / ML model.

10. The base station according to claim 9, wherein the processor is further configured to: Comparing the third CSI and the fourth CSI, the fourth CSI being obtained based on the second AI / ML model and the received second PMI; and The transceiver sends the pairing result of the second AI / ML model and the first AI / ML model to the UE, the pairing result being determined based on the comparison result.

11. The base station of claim 8, wherein receiving the first PMI and the at least one second PMI comprises: Sending a request to the UE via the transceiver, associated with determining whether the second AI / ML model can be paired with an active first AI / ML model among the at least one first AI / ML models; and The first PMI and the second PMI are received from the UE via the transceiver, and the second PMI is associated with the activated first AI / ML model.

12. The base station of claim 8, wherein the at least one second PMI comprises a plurality of second PMIs associated with a plurality of first AI / ML models, and Receiving the first PMI and the at least one second PMI includes: The transceiver receives a request from the UE for selecting a first AI / ML model from the plurality of first AI / ML models; as well as The first PMI and the plurality of second PMIs are received from the UE via the transceiver.

13. The base station of claim 12, wherein the processor is further configured to: Based on the third CSI and multiple fourth CSIs, a first AI / ML model is selected from the multiple first AI / ML models, wherein the multiple fourth CSIs are obtained based on the second AI / ML model and the multiple second PMIs. The transceiver sends an indication of the selection result to the UE. in When a first AI / ML model is selected from the plurality of first AI / ML models, the indication specifies the model identifier (ID) of the selected first AI / ML model, and If no first AI / ML model is selected, the indication indicates that the pairing has failed.

14. The base station of claim 8, wherein the configuration with respect to the first codebook includes a scaling factor for parameters associated with a second codebook, the second codebook being indicated in a codebook configuration transmitted from the base station.

15. The base station of claim 8, wherein the configuration of the first codebook is transmitted as function / model-related information during a function or model identification process between the UE and the base station, or The configuration of the first codebook is transmitted via Radio Resource Control (RRC) signaling.

16. The base station of claim 14, wherein the parameters include at least one of the following: The number of beams; The number of phase quantization magnitudes; Number of oversamples; The number of beam amplitude scaling factors for both broadband and subband; or The number of beam combination coefficients or phases in the beam, polarization, and layer.

17. A processor for wireless communication, comprising: At least one memory; as well as A controller, coupled to the at least one memory, and configured such that the processor: Receive configuration information about the first codebook from the base station via transceiver; as well as The transceiver transmits a first precoding matrix indicator (PMI) and at least one second PMI to the base station, the first PMI being obtained based on the first codebook and the first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

18. A method performed by a user equipment (UE), the method comprising: Receive configuration information about the first codebook from the base station via transceiver; as well as The transceiver transmits a first precoding matrix indicator (PMI) and at least one second PMI to the base station, the first PMI being obtained based on the first codebook and the first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model.

19. A processor for wireless communication, comprising: At least one memory; as well as A controller, coupled to the at least one memory, and configured such that the processor: The configuration of the first codebook is sent to the user equipment (UE) via the transceiver; The transceiver receives a first precoding matrix indicator (PMI) and at least one second PMI from the UE, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model; and Based on a third CSI and at least one fourth CSI, it is determined whether the base station's second AI / ML model can be paired with the at least one first AI / ML model. The third CSI is obtained based on the first codebook and the received first PMI, and the at least one fourth CSI is obtained based on the second AI / ML model and the received at least one second PMI.

20. A method performed by a base station, the method comprising: The configuration of the first codebook is sent to the user equipment (UE) via the transceiver; The transceiver receives a first precoding matrix indicator (PMI) and at least one second PMI from the UE, the first PMI being obtained based on the first codebook and first channel state information (CSI) estimated by the UE, and the at least one second PMI being obtained by at least one first artificial intelligence or machine learning (AI / ML) model; and Based on a third CSI and at least one fourth CSI, it is determined whether the base station's second AI / ML model is paired with the at least one first AI / ML model. The third CSI is obtained based on the first codebook and the received first PMI, and the at least one fourth CSI is obtained based on the second AI / ML model and the received at least one second PMI.

Images