Channel state information (CSI) compression-decompression in wireless communication systems
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2026-02-05
- Publication Date
- 2026-08-13
Smart Images
Figure IN2026050186_13082026_PF_FP_ABST
Abstract
Description
CHANNEL STATE INFORMATION (CSI) COMPRESSIONDECOMPRESSION IN WIRELESS COMMUNICATION SYSTEMSTECHNICAL FIELD
[0001] The present disclosure relates to a technical field of wireless communication systems. More particularly, the present disclosure relates to methods, apparatuses, and systems for performing Channel State Information (CSI) compression and decompression effectively in a wireless communication system (e.g., at a network side and a user equipment (UE) side).BACKGROUND OF THE DISCLOSURE
[0002] Wireless communication systems have been widely deployed for daily voice, video, data, and SMS service. Mobile communications have been routed through several stages of development, including 2G (Global System for Mobile Communications (GSM)), 3G (TD-SCDMA, UMTS), 4G (Long Term Evolution (LTE)), and have now entered and deployment stage of 5G / 6G (New Radio (NR)).
[0003] In the context of 5G / 6G networks, Channel State Information (CSI) plays a critical role in enabling network optimization and resource allocation. A typical wireless communication system involves exchange of the CSI between a user equipment (UE) and a network (network may include a network entity (for e.g., base station)). Within the exchange, the UE measures and reports its channel conditions as the CSI to the network for optimization of communication resources. Due to bandwidth limitations, the UE cannot transmit entire CSI towards the network. Thus, the UE may be required to compress the CSI and send the compressed CSI towards the network for the network to reconstruct / decompress the compressed CSI. Generally, an encoder is implemented at the UE side for CSI compression, and a decoder is implemented at the network side for CSI reconstruction / decompression.
[0004] However, aligning encoder and decoder models between the UE side and the network side remains a significant challenge. The challenge arises primarily due to mismatches in dataset distributions associated with the CSI between the UE and the network. In the context of CSI compression for 5G / 6G networks, data distribution mismatch refers to differences in how the data (specifically, data associated with the CSI) is represented, structured, or distributed between the UE (UE-side) and the network (network-side). The differences can create issues when trying to align the models (i.e., the encoder and decoder models) or methods used on both the UE side and the network side for various tasks such as compression, transmission, and reconstruction. Additionally, dynamic nature of unknown conditions associated with the UE-side, which may fluctuate due to environmental factors or user mobility, introduces further complexities in maintaining synchronization between the UE and the network.
[0005] Current methods for CSI compression fail to adequately address the UE-side data distribution mismatch and often introduce large overheads in terms of computational complexity and bandwidth usage, making the current methods unsuitable for deployment in high-performance 5G and 6G networks.
[0006] Thus, there exists a need for techniques which can address the above-discussed and other related challenges. Specifically, there exists a need for techniques for ensuring effective CSI compression at network side and user equipment (UE) side.SUMMARY OF THE DISCLOSURE
[0007] The following presents a simplified summary to provide a basic understanding of some aspects of the subject matter disclosed in this specification. This summary is not an extensive overview and is intended to neither identify key or critical elements nor delineate the scope of such elements. Its purpose is to present some concepts of the described features in a simplified form as a prelude to the more detailed description that is presented later.
[0008] In one non-limiting embodiment, the present disclosure describes a method of wireless communication at a network entity. The method comprises: training a network side encoder and a network side decoder using a first set of data associated with the network entity; transmitting at least one of: the network side encoder or at least a portion of the first set of data to a user equipment (UE); receiving compressed channel state information (CSI) from the UE; and reconstructing the received compressed CSI using the network side decoder to generate reconstructed CSI.
[0009] In another non-limiting embodiment, the present disclosure describes a method of wireless communication at a user equipment (UE). The method comprises: receiving at least one of: a network side encoder or at least a portion of a first set of data associated with a network entity; generating a UE side decoder based on at least one of: the network side encoder or the portion of the first set of data associated with the network entity; generating a UE side encoder based on the generated UE side decoder and a first set of data associated with the UE; compressing channel state information (CSI) using the generated UE side encoder; and transmitting the compressed CSI to the network entity.
[0010] In another non-limiting embodiment, the present disclosure describes a network entity. The network entity comprises a network side memory and one or more network side processors communicatively coupled with the network side memory. The one or more network side processors are configured to train a network side encoder and a network side decoder using a first set of data associated with the network entity. The one or more network side processors are configured to transmit at least one of: the network side encoder or at least a portion of thefirst set of data to a user equipment (UE). The one or more network side processors are configured to receive compressed channel state information (CSI) from the UE. The one or more network side processors are configured to reconstruct the received compressed CSI using the network side decoder to generate reconstructed CSI.
[0011] In another non-limiting embodiment, the present disclosure describes a user equipment (UE). The UE comprises a UE side memory and one or more UE side processors communicatively coupled with the UE side memory. The one or more UE side processors are configured to receive at least one of: a network side encoder or at least a portion of a first set of data associated with a network entity. The one or more UE side processors are configured to generate a UE side decoder based on at least one of: the network side encoder or the portion of the first set of data associated with the network entity. The one or more UE side processors are configured to generate a UE side encoder based on the generated UE side decoder and a first set of data associated with the UE. The one or more UE side processors are configured to compress channel state information (CSI) using the generated UE side encoder. The one or more UE side processors are configured to transmit the compressed CSI to the network entity.
[0012] In another non-limiting embodiment, the present disclosure describes a method of wireless communication. The method comprises: training, at a network entity, a network side encoder and a network side decoder using a first set of data associated with the network entity; transmitting, by the network entity, the network side encoder to a user equipment (UE); receiving, at the UE, the network side encoder from the network entity; compressing, at the UE, a channel state information (CSI) using the received network side encoder; transmitting, by the UE, the compressed CSI to the network entity; receiving, at the network entity, the compressed CSI from the UE; and reconstructing, at the network entity, the received compressed CSI using the network side decoder to generate a reconstructed CSI.
[0013] In another non-limiting embodiment, the present disclosure describes a wireless communication system. The wireless communication system comprises a network entity. The network entity is configured to train at least a network side encoder and a network side decoder using a first set of data associated with the network entity. The network entity is configured to transmit the network side encoder to a user equipment (UE). The wireless communication system comprises the UE. The UE is configured to receive the network side encoder from the network entity. The UE is configured to compress channel state information (CSI) using the received network side encoder. The UE is configured to transmit the compressed CSI to the network entity. The network entity is configured to receive the compressed CSI from the UE,and reconstruct the received compressed CSI using the network side decoder to generate a reconstructed CSI.
[0014] In another non-limiting embodiment, the present disclosure describes a method of wireless communication. The method comprises: transmitting, by a network entity, an indication comprising a model identifier (ID) to a user equipment (UE), the model ID comprising a pair of an encoder and a decoder between the network entity and the UE; receiving, at the UE, the indication comprising the model ID from the network entity; determining, at the UE, an encoder corresponding to the received model ID; transmitting, by the UE, compressed CSI to the network entity; receiving, at the network entity, the compressed CSI from the UE; and reconstructing, at the network entity, the received compressed CSI using a decoder corresponding to the model ID to generate a reconstructed CSI.
[0015] In another non-limiting embodiment, the present disclosure describes a wireless communication system. The wireless communication system comprises a network entity. The network entity is configured to transmit an indication comprising a model identifier (ID) to a user equipment (UE). The model ID comprises a pair of an encoder and a decoder between the network entity and the UE. The wireless communication system comprises the UE. The UE is configured to receive the indication comprising the model ID from the network entity. The UE is configured to determine an encoder corresponding to the received model ID. The UE is configured to transmit compressed CSI to the network entity. The network entity is configured to receive the compressed CSI, and reconstruct the received compressed CSI using a decoder corresponding to the model ID to generate a reconstructed CSI.
[0016] The above summary is provided merely for the purpose of summarizing some example embodiments to provide a basic understanding of some aspects of the disclosure. Accordingly, it will be appreciated that the above-described embodiments are merely examples and should not be construed to narrow the scope or spirit of the disclosure in any way. It will be appreciated that the scope of the disclosure encompasses many potential embodiments in addition to those here summarized, some of which will be further described below.BRIEF DESCRIPTION OF DRAWINGS
[0017] The embodiments of the disclosure itself, as well as a preferred mode of use, further objectives and advantages thereof, will best be understood by reference to the following detailed description of an illustrative embodiment when read in conjunction with the accompanying drawings. One or more embodiments are now described, by way of example only, with reference to the accompanying drawings in which:
[0018] FIGS. 1A-1B illustrate an exemplary network environment or wireless communication system 100, in which the techniques of the present disclosure may be implemented;
[0019] FIGS. 2-3 illustrate exemplary scenarios 200, 300 of a first method for performing CSI compression and CSI reconstruction, in accordance with some embodiments of present disclosure;
[0020] FIG. 4 illustrate an exemplary scenario 400 of a second method for performing CSI compression and CSI reconstruction, in accordance with some embodiments of present disclosure;
[0021] FIG. 5 illustrate an exemplary scenario 500 of a third method for performing CSI compression and CSI reconstruction, in accordance with some embodiments of present disclosure;
[0022] FIG. 6 illustrates an exemplary structure 600 of a standardized artificial intelligence / machine learning (AI / ML) model 115, 125, in accordance with some embodiments of present disclosure;
[0023] FIG. 7 illustrates a flowchart illustrating an exemplary method 700 for performing CSI reconstruction, in accordance with some embodiments of the present disclosure;
[0024] FIG. 8 illustrates a flowchart illustrating an exemplary method 800 for performing CSI compression, in accordance with some embodiments of the present disclosure;
[0025] FIG. 9 illustrates a flowchart illustrating another exemplary method 900 for performing CSI compression and reconstruction, in accordance with some embodiments of the present disclosure; and
[0026] FIG. 10 illustrates a flowchart illustrating another exemplary method 1000 for performing CSI compression and reconstruction, in accordance with some embodiments of the present disclosure.DETAILED DESCRIPTION
[0027] Exemplary embodiments are described with reference to the accompanying drawings. Wherever convenient, the same reference numbers are used throughout the drawings to refer to the same or like parts. While examples and features of disclosed principles are described herein, modifications, adaptations, and other implementations are possible without departing from the spirit and scope of the disclosed embodiments. It is intended that the following detailed description be considered as exemplary only, with the true scope and spirit being indicated by the following claims. Additional illustrative embodiments are listed below.
[0028] In the present document, the word “exemplary” is used herein to mean “serving as an example, instance, or illustration.” Any embodiment or implementation of the present subject matter described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.
[0029] As used herein, the term “comprising” is not intended to be limiting, but may be a transitional term synonymous with “including,” “containing,” or “characterized by.” The term “comprising” may thereby be inclusive or open-ended and does not exclude additional, unrecited elements or method steps when used in a claim. For instance, in describing a method, “comprising” indicates that the claim is open-ended and allows for additional steps. In describing a device, “comprising” may mean that a named element(s) may be essential for an embodiment or aspect, but other elements may be added and still form a construct within the scope of a claim. In contrast, the transitional phrase “consisting of’ excludes any element, step, or ingredient not specified in a claim. This is consistent with the use of the term throughout the specification.
[0030] As discussed in the background, a typical wireless communication system involves exchange of the CSI between a user equipment (UE) and a network (network may include a network entity (for e.g., base station)). Due to bandwidth limitations, the UE cannot transmit entire CSI towards the network. Thus, the UE may be required to compress the CSI and send the compressed CSI towards the network. Generally, an encoder is implemented at the UE side for CSI compression, and a decoder is implemented at the network side for CSI reconstruction / decompression.
[0031] However, aligning encoder and decoder models between the UE side and the network side remains a significant challenge that arises primarily due to mismatches in dataset distributions associated with the CSI between the UE and the network. The differences can create issues when trying to align the models (i.e., the encoder and decoder models) or methods used on both the UE side and the network side for various tasks such as compression, transmission, and reconstruction. Additionally, dynamic nature of unknown conditions associated with the UE-side, which may fluctuate due to environmental factors or user mobility, introduces further complexities in maintaining synchronization between the UE and the network.
[0032] Current methods for CSI compression fail to adequately address the UE-side data distribution mismatch and often introduce large overheads in terms of computational complexity and bandwidth usage, making the current methods unsuitable for deployment in high-performance 5G and 6G networks. Thus, there exists a need for techniques which canaddress the above-discussed and other related challenges. Specifically, there exists a need for techniques for ensuring effective CSI compression at network side and user equipment (UE) side.
[0033] The present disclosure described various techniques for ensuring effective CSI compression while overcoming data distribution mismatches on both the network side and UE-side of a communication system. Further, the present disclosure aims to address challenges related to the alignment of data representations between different vendors (i.e., the UE and the network) and minimize the complexity associated with inter-vendor collaboration (i.e., collaboration between the UE and the network). In particular, the present disclosure describes various techniques for training a UE-side encoder using network-side shared information, optimizing the compression and reconstruction of the CSI. A detailed explanation of the proposed technique(s) is disclosed in the forthcoming paragraphs.
[0034] FIGS. 1A-1B illustrates an exemplary network environment or wireless communication system 100, in which the techniques of the present disclosure may be implemented.
[0035] As shown in FIG. 1A, the environment 100 may comprise a network entity 101, a plurality of UEs 103a, 103b, 103c (also referred to as UE 103), and a server 105 configured to communicate with the UE 103. Depending on the network type, the term network entity 101 may be referred to as any component (or collection of components) configured to provide wireless access to a network such as transmit point (TP) or base stations (evolved Node B (eNB), gNodeB (gNB), macrocells, microcells, small cells, etc.,) that may be configured to transmit and receive data to / from the plurality of UEs 103. The network entity 101 may be configured to provide wireless access in accordance with one or more wireless communication protocols e.g., 5G 3GPP new radio interface / access (NR), long term evolution (LTE), LTE advanced (LTE -A), etc. It will be apparent to a person skilled in the art that, although only a single network entity 101 is illustrated, any number of network entities may be employed as required, depending on the specific implementation and network configuration. Also, depending on the network type, the UE 103 may be referred to as any component such as mobile station, wireless terminal, user device, mobile device, etc. The network entity 101 may communicate with the UE 103 using one or more wireless communication techniques including at least 5G, LTE, LTE -A, Wi-Fi, or other wireless communication techniques. Further, the UE 103 may be coupled to the server 105.
[0036] In some embodiments, the network entity 101 may comprise one or more network side processors 107, a network side memory 109, a network side encoder (El) 111, a network side decoder (DI) 113, and an artificial intelligence / machine learning (AI / ML) model 115. Thenetwork side encoder (El) 111 may be trained at network side and may be responsible for CSI compression at network side. The network side decoder (DI) 113 may be responsible for CSI decompression at network side. Additionally, the network entity 101 may comprise a backhaul or network interface, multiple antennas, multiple Radio Frequency (RF) Transceivers, receive (Rx) processing circuitry, transmit (Tx) processing circuitry (not shown) as necessary for ensuring the effective CSI compression.
[0037] Similarly, the UE 103 may comprise one or more UE side processors 117, a UE side memory 119, a UE side encoder (E2) 121, a UE side decoder (D2) 123, and an AI / ML model 125. The UE side encoder (E2) 121 may be locally trained at UE side for the CSI compression. The UE side decoder (D2) 123 may be locally trained at the UE side for CSI reconstruction. The UE side decoder (D2) 123 may mimic the network side decoder 113 at the network side. Additionally, the UE 103 may also comprise a backhaul or network interface, multiple antennas, multiple Radio Frequency (RF) Transceivers, receive (Rx) processing circuitry, transmit (Tx) processing circuitry (not shown) as necessary for ensuring the effective CSI compression. Although, for illustrative purposes, the aforementioned components are depicted at the UE 103, it will be apparent to a person skilled in the art that these components may alternatively be comprised at the server 105 serving the UE 103. In some embodiments, the one or more network side processors 107 and the one or more UE side processors 117 may be configured to implement the functionalities of the present disclosure.
[0038] In one implementation, the one or more network side processors 107, the one or more UE side processors 117, the network side encoder 111, the UE side encoder 121, the network side decoder 113, and the UE side decoder 123 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, and / or any devices that manipulate signals based on operational instructions. In one embodiment, the one or more network side processors 107 and the one or more UE side processors 117 may be configured as standalone control unit or combination of one or more control units for implementing the subject matter of the present disclosure. Among other capabilities, the one or more network side processors 107 and the one or more UE side processors 117 may be configured to fetch and execute computer-readable instructions and other information stored in the respective memories. In some implementations, the data such as CSI information or CSI data may be stored within the network side memory 109 and the UE side memory 119 in the form of various data structures. Additionally, the CSI data may be organized using data models, such as relational or hierarchical data models. Theother data may include various temporary data and files generated by the units while performing various functions of the network entity 101 and the UE 103.
[0039] CSI refers to the information of the multipath wireless channel between the network entity 101 and the UE 103. The UE 103 measures downlink reference signals, compute downlink CSI, and provide a CSI report to the network entity 101, thereby facilitating downlink transmission. The exemplary environment 100 illustrated in a scenario of a typical wireless communication system which involves exchange of the CSI between the UE 103 and the network entity 101, where the UE 103 measures and reports its channel conditions to the network for optimization of communication resources. However, due to bandwidth limitations, the UE 103 may be required to compress the CSI using an encoder. Generally, an encoder is implemented at the UE side for CSI compression and a decoder is implemented at the network side for CSI reconstruction / decompression. However, aligning encoder-decoder models for effective CSI compression between the UE-side and NW-side remains a significant challenge, primarily due to mismatches in dataset distributions between the UE 103 and the network. Thus, various techniques are needed which can address these challenges, and ensure effective CSI compression at network side and UE side.
[0040] Now, referring to FIG. IB, the network entity 101 may be configured to transmit information such as dataset (i.e., a first set of data, a second set of data, etc.), a dataset ID, a model identifier (ID) with the UE 103 and the UE 103 may further transmit the information to the server 105. Prior to transmitting, the network entity 101 may be configured to split the dataset into one or more subsets based on UE’s 103 capability, and transmit the one or more subsets to the UE 103. Along with the one or more subsets, the network entity 101 may allocate the dataset ID and the model ID to the dataset to ensure proper tracking. Further, all the UEs 103 may be configured to establish and maintain a connection with the server 105. The server 105 may be responsible for aggregating the received information or dataset which may be utilized for training and validation of the UE side encoder (E2) 121 and the UE side decoder (D2) 123 at the UE side for performing the effective CSI compression. Utilizing the dataset for effective CSI compression will be explained in greater detail in conjunction with FIGS. 3-5 in the forthcoming paragraphs with the various methods (which may also be referred to as “directions”).
[0041] FIGS. 2-3 illustrate an exemplary scenarios 200, 300 of a first method for performing CSI compression and CSI reconstruction, in accordance with some embodiments of present disclosure. The first method may also be referred to as “direction A” or “technique 1”. The terms “first method”, “direction A”, and “technique 1” may have the same meaning and maybe used interchangeably through the specification, without departing from the scope of the disclosure. The first method may be implemented with the help of at least a first approach (also referred to as “approach 1”) and a second approach (also referred to as “approach 2”).
[0042] Technique 1.1 (Approach 1 of Technique 1):
[0043] In the first approach of the “direction A”, at the network side, the one or more network side processors 107 may be configured to train the network side encoder (El) 111 and the network side decoder (DI) 113 using a first set of data associated with the network entity 101. The first set of data may comprise first CSI (VI), first CSI feedback (Cl), second CSI (V2), and second CSI feedback (C2). The first CSI (Vl)may correspond to a target CSI (i.e., an input) at the UE 103. The first CSI feedback (Cl) may correspond to CSI feedback (i.e., an output) at the UE 103. The second CSI (V2) may correspond to a reconstructed target CSI. The second CSI feedback (C2) may correspond to a CSI feedback at the network entity 101. It may be understood by a person skilled in the art that the dataset (A) may be pre-existing or readily available at the network entity 101.
[0044] For training (as shown in FIG. 2) the network side encoder (El) 111 and the network side decoder (DI) 113, the one or more network side processors 107 may be configured to provide the first CSI (VI) to the network side encoder (El) 111 to generate the first CSI feedback (Cl). Further, the one or more network side processors 107 may be configured to provide the second CSI feedback (C2) to the network side decoder (DI) 113 to generate the second CSI (V2). The second CSI feedback (C2) may be generated by adding some noise 201 to the first CSI feedback (Cl). In one non-limiting example, the noise 201 may correspond to additive white gaussian noise (AWGN).
[0045] In some embodiments, the one or more network side processors 107 may be configured to transmit at least one of: the network side encoder (El) 111 or at least a portion of the first set of data to the UE 103. In a non-limiting example, the one or more network side processors 107 may be configured to transmit the network side encoder (El) 111 and the first CSI (VI) to the UE 103. Before transmitting, the one or more network side processors 107 may be configured to split the first set of data into one or more subsets based on UE capability. The one or more network side processors 107 may be configured to allocate an associate ID, a model ID to each of the one or more subsets. The associated ID may comprise additional conditions that may impact the performance of encoder(s) and decoder(s). The model ID may be utilized to pair the encoder and decoder between the network and the UE 103. The one or more network side processors 107 may be configured to transmit to the UE 103, a subset having the associate ID, the model ID corresponding to the UE 103 for CSI compression.
[0046] In some embodiments, the one or more network side processors 107 may be configured to transmit a performance target to the UE 103 along with the at least one of: the network side encoder (E 1) 111 or the portion of the first set of data. The one or more network side processors 107 may be configured to specify the type of input / output used to guide the UE 103 in adapting its own input and output accordingly. For example, Signal to Goal Compression score (SGCS) may be considered as the performance target for either end-to-end or only encoder performance at the UE side.
[0047] At the UE side, the one or more UE side processors 117 may be configured to receive (not shown) at least one of: the network side encoder (El) 111 or at least the portion of the first set of data associated with the network entity 101. In a non-limiting example, the one or more UE side processors 117 may be configured to receive the network side encoder (El) 111 and the first CSI (VI) from the network entity 101. Further, the one or more UE side processors 117 may be configured to receive, from the network entity 101, the subset corresponding to the UE 103 from one or more subsets of the first set of data. The subset may comprise at least the associate ID, the model ID for generating the UE side encoder (E2) 121.
[0048] In some embodiments, the one or more UE side processors 117 may be configured to generate the UE side decoder (D2) 123 based on at least one of: the network side encoder (El) 111 or the portion of the first set of data associated with the network entity 101. For generating the UE side decoder (D2) 123, the one or more UE side processors 117 may be configured to provide the first CSI (VI) and the network side encoder (El) 111 to determine the first CSI feedback (Cl). The one or more UE side processors 117 may be configured to add at least one type of noise (not shown) to the first CSI feedback (Cl) to determine a second CSI feedback (C2) for generating the UE side decoder (D2) 123. In one non-limiting example, the noise may correspond to additive white gaussian noise (AWGN). The one or more UE side processors 117 may be configured to provide the second CSI feedback (C2) to the UE side decoder (D2) 123 to determine (predict) the second CSI (V2). The one or more UE side processors 117 may be configured to determine a difference between the second CSI (V2) and the first CSI (VI) to determine a loss between the second CSI (V2) and the first CSI (VI). The one or more UE side processors 117 may be configured to fine-tune the UE side decoder (D2) 123 based on the determined loss. In one example, the one or more UE side processors 117 may be configured to fine tune weights of the UE side decoder (D2) 123 based on the determined loss. The generated UE side decoder (D2) 123 will mimic the network side decoder (DI) 113 at the network entity 101.
[0049] In some embodiments, the one or more UE side processors 117 may be configured to generate the UE side encoder (E2) 121 based on the generated UE side decoder (D2) 123 and the first set of data associated with the UE 103. The first set of data associated with the UE 103 may be a set of data that is captured at a different instance (e.g., time, position of the UE) from the first set of data associated with the network entity 101. The first set of data associated with the UE may depend upon the mobility of the UE, thus, may vary at different instances. Thus, generating the UE side encoder (E2) 121 may adapt the UE side additional conditions due to the first set of data associated with the UE. The one or more UE side processors 117 may be further configured to compress the CSI (i.e., the first CSI (VI)) using the generated UE side encoder (E2) 121. In one example, the first CSI (VI) may be provided as input to the generated UE side encoder (E2) 121. The generated UE side encoder (E2) 121 may compress the first CSI (VI) to generate a compressed CSI (i.e., the first CSI feedback (Cl)) (CCSI), as shown in FIG.3(a). Then, the one or more UE side processors 117 may be configured to transmit the compressed CSI to the network entity 101.
[0050] At the network side, the one or more network side processors 107 may be configured to receive the CCSI from the UE 103. The CCSI may be received along with some added noise (that may be added during the transmission) (i.e., receiving the second CSI feedback (C2)). The one or more network side processors 107 may be configured to reconstruct the received compressed CSI using the network side decoder (DI) 113 to generate a reconstructed CSI (RCSI) (i.e., the second CSI (V2)). In one example, the second CSI feedback (C2) may be provided as input to the network side decoder (DI) 113. The network side decoder (DI) 113 may reconstruct / decompress the second CSI feedback (C2) to generate the RCSI (i.e., the second CSI (V2)), as shown in FIG. 3(b).
[0051] Technique 1.2 (Approach 2 of Technique 1):
[0052] In the second approach of the “direction A”, at the network side, the one or more network side processors 107 may be configured to train the network side encoder (El) 111 and the network side decoder (DI) 113 using a first set of data associated with the network entity 101. The first set of data may comprise first CSI (VI), first CSI feedback (Cl), second CSI (V2), and second CSI feedback (C2). The first CSI (VI) may correspond to a target CSI at the UE 103. The first CSI feedback (Cl) may correspond to CSI feedback at the UE 103. The second CSI (V2) may correspond to a reconstructed target CSI. The second CSI feedback (C2) may correspond to a CSI feedback at the network entity 101.
[0053] For training (as shown in FIG. 2) the network side encoder (El) 111 and the network side decoder (DI) 113, the one or more network side processors 107 may be configured toprovide the first CSI (VI) to the network side encoder (El) 111 to generate the first CSI feedback (Cl). Further, the one or more network side processors 107 may be configured to provide the second CSI feedback (C2) to the network side decoder (DI) 113 to generate the second CSI (V2). The second CSI feedback (C2) may be generated by adding some noise 201 to the first CSI feedback (Cl). In one non-limiting example, the noise 201 may correspond to additive white gaussian noise (AWGN).
[0054] In some embodiments, the one or more network side processors 107 may be configured to transmit at least one of: the network side encoder (El) 111 or at least a portion of the first set of data to the UE 103. In a non-limiting example, the one or more network side processors 107 may be configured to transmit the first CSI (VI) and the first CSI feedback (Cl) to the UE 103. Before transmitting, the one or more network side processors 107 may be configured to split the first set of data into one or more subsets based on UE capability. The one or more network side processors 107 may be configured to allocate an associate ID, a model ID to each of the one or more subsets. The one or more network side processors 107 may be configured to transmit to the UE, a subset having the associate ID, the model ID corresponding to the UE 103 for CSI compression.
[0055] In some embodiments, the one or more network side processors 107 may be configured to transmit a performance target to the UE 103 along with the at least one of: the network side encoder (El) 111 or the portion of the first set of data.
[0056] At the UE side, the one or more UE side processors 117 may be configured to receive (not shown) at least one of: the network side encoder (El) 111 or at least the portion of the first set of data associated with the network entity 101. In a non-limiting example, the one or more UE side processors 117 may be configured to receive the first CSI (VI) and the first CSI feedback (Cl) from the network entity 101. Further, the one or more UE side processors 117 may be configured to receive, from the network entity 101, the subset corresponding to the UE 103 from one or more subsets of the first set of data. The subset may comprise at least the associate ID, the model ID for generating the UE side encoder (E2) 121.
[0057] In some embodiments, the one or more UE side processors 117 may be configured to generate the UE side decoder (D2) 123 based on at least one of: the network side encoder (El) 111 or the portion of the first set of data associated with the network entity 101. For generating the UE side decoder (D2) 123, the one or more UE side processors 117 may be configured to provide the first CSI (VI) and the first CSI feedback (Cl) to determine the network side encoder (El) 111. The one or more UE side processors 117 may be configured to add at least one type of noise (not shown) to the first CSI feedback (Cl) to determine a second CSI feedback (C2)for generating the UE side decoder (D2) 123. In one non-limiting example, the noise may correspond to additive white gaussian noise (AWGN). The one or more UE side processors 117 may be configured to provide the second CSI feedback (C2) to the UE side decoder (D2) 123 to determine (predict) the second CSI (V2). The one or more UE side processors 117 may be configured to determine a difference between the first CSI (VI) and the second CSI (V2) to determine a loss between the first CSI (VI) and the second CSI (V2). The one or more UE side processors 117 may be configured to fine-tune the UE side decoder (D2) 123 based on the determined loss. In one example, the one or more UE side processors 117 may be configured to fine tune weights of the UE side decoder (D2) 123 based on the determined loss. The generated UE side decoder (D2) 123 will mimic the network side decoder (DI) 113 at the network entity 101.
[0058] In some embodiments, the one or more UE side processors 117 may be configured to generate the UE side encoder (E2) 121 based on the generated UE side decoder (D2) 123 and the first set of data associated with the UE 103. The first set of data associated with the UE 103 may be a set of data that is captured at a different instance (e.g., time, position of the UE) from the first set of data associated with the network entity 101. The first set of data associated with the UE may depend upon the mobility of the UE, thus, may vary at different instances. Thus, generating the UE side encoder (E2) 121 may adapt the UE side additional conditions due to the first set of data associated with the UE. The one or more UE side processors 117 may be further configured to compress the CSI (i.e., the first CSI (VI)) using the generated UE side encoder (E2) 121. In one example, the first CSI (VI) may be provided as input to the generated UE side encoder (E2) 121. The generated UE side encoder (E2) 121 may compress the first CSI (VI) to generate the CCSI (i.e., the first CSI feedback (Cl)), as shown in FIG. 3(a). Then, the one or more UE side processors 117 may be configured to transmit the compressed CSI to the network entity 101.
[0059] At the network side, the one or more network side processors 107 may be configured to receive the CCSI from the UE 103. The CCSI may be received along with some added noise (that may be added during the transmission) (i.e., receiving the second CSI feedback (C2)). The one or more network side processors 107 may be configured to reconstruct the received compressed CSI using the network side decoder (DI) 113 to generate the RCSI (i.e., the second CSI (V2)). In one example, the second CSI feedback (C2) may be provided as input to the network side decoder (DI) 113. The network side decoder (DI) 113 may reconstruct / decompress the second CSI feedback (C2) to generate the RCSI (i.e., the second CSI (V2)), as shown in FIG. 3(b).
[0060] FIG. 4 illustrate an exemplary scenario 400 of a second method for performing CSI compression and CSI reconstruction, in accordance with some embodiments of present disclosure. The second method may also be referred to as “direction B” or “technique 2”. The terms “second method”, “direction B”, and “technique 2” may have the same meaning and may be used interchangeably through the specification, without departing from the scope of the disclosure.
[0061] Technique 2:
[0062] In some embodiments, the second method may be implemented by a wireless communication system (e.g., the system 100 of FIG. 1). The wireless communication system 100 may comprise the network entity 101 and the UE 103. At the network side (the network entity 101), the one or more network side processors 107 may be configured to train at least a network side encoder (El) 111 and a network side decoder (DI) 113 using the first set of data associated with the network entity 101. The training may be same as the training discussed in the first method (direction A). In some embodiments, the one or more network side processors 107 may be configured to transmit the network side encoder (El) 111 to the UE 103.
[0063] At the UE side, the one or more UE side processors 117 may be configured to receive the network side encoder (El) 111 from the network entity 101. The one or more UE side processors 117 may be configured to compress the CSI (i.e., the first CSI (VI)) using the received network side encoder (El) 111. In one example, the first CSI (VI) may be provided as input to the received network side encoder (El) 111. The received network side encoder (El) 111 may compress the first CSI (VI) to generate a compressed CSI (i.e., the first CSI feedback (Cl)) (CCSI), as shown in FIG. 4(a). Further, the one or more UE side processors 117 may be configured to transmit the CCSI to the network entity 101.
[0064] At the network side, the one or more network side processors 107 may be configured to receive the CCSI from the UE 103. The CCSI may be received along with some added noise (that may be added during the transmission) (i.e., receiving the second CSI feedback (C2)). The one or more network side processors 107 may be configured to reconstruct the received CCSI using the network side decoder (DI) 113 to generate a reconstructed CSI (RCSI) (i.e., the second CSI (V2)). In one example, the second CSI feedback (C2) may be provided as input to the network side decoder (DI) 113. The network side decoder (DI) 113 may reconstruct / decompress the second CSI feedback (C2) to generate the RCSI (i.e., the second CSI (V2)), as shown in FIG. 4(b).
[0065] FIG. 5 illustrate an exemplary scenario 500 of a third method for performing CSI compression and CSI reconstruction, in accordance with some embodiments of presentdisclosure. The third method may also be referred to as “direction C” or “technique 3”. The terms “third method”, “direction C”, and “technique 3” may have the same meaning and may be used interchangeably through the specification, without departing from the scope of the disclosure.
[0066] Technique 3:
[0067] In some embodiments, the third method may be implemented by a wireless communication system (e.g., the system 100 of FIG. 1). The wireless communication system 100 may comprise the network entity 101 and the UE 103. At the network side (the network entity 101), the one or more network side processors 107 may be configured to transmit an indication comprising a model ID to the UE 103. The model ID may comprise a pair of an encoder 502 and a decoder 504 between the network entity 101 and the UE 103.
[0068] At the UE side, the one or more UE side processors 117 may be configured to receive the indication comprising the model ID from the network entity 101. The one or more UE side processors 117 may be configured to determine an encoder (e.g., the encoder 502) corresponding to the received model ID. The one or more UE side processors 117 may be configured to compress the CSI (i.e., the first CSI (VI)) using the determined encoder. In one example, the first CSI (VI) may be provided as input to the determined encoder (E) 502. The determined encoder (E) 502 may compress the first CSI (VI) to generate a compressed CSI (i.e., the first CSI feedback (Cl)) (CCSI), as shown in FIG. 5(a). Further, the one or more UE side processors 117 may be configured to transmit the CCSI to the network entity 101.
[0069] At the network side, the one or more network side processors 107 may be configured to receive the CCSI from the UE 103. The CCSI may be received along with some added noise (that may be added during the transmission) (i.e., receiving the second CSI feedback (C2)). The one or more network side processors 107 may be configured to reconstruct the received CCSI using a decoder (e.g., the decoder 504) corresponding to the model ID to generate a reconstructed CSI (RCSI) (i.e., the second CSI (V2)). In one example, the second CSI feedback (C2) may be provided as input to the determined decoder (D) 504. The determined decoder (D) 504 may reconstruct / decompress the second CSI feedback (C2) to generate the RCSI (i.e., the second CSI (V2)), as shown in FIG. 5(b).
[0070] Thus, the “direction A” shares the target CSI (VI), the CSI feedback (Cl), and the network side encoder (El) 111 that enables the UE 103 to train the UE side encoder (E2) 121, thereby ensuring high compression accuracy even under dynamic conditions specific to the UE 103. The “direction B” provides the pre-trained encoders (i.e., the network side encoder (El) 111) directly to the UE 103 thereby simplifying the implementation process while ensuringcompression accuracy through the use of the network side decoder (D2) 113 at the network side. Finally, the “direction C” related to standardized model structures and model IDs promotes compatibility and mitigates challenges related to inter-vendor integration.
[0071] FIG. 6 illustrates an exemplary structure 600 of a standardized artificial intelligence / machine learning (AI / ML) model 115, 125, in accordance with some embodiments of present disclosure.
[0072] In some embodiments, the AI / ML model 115, 125 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuitries, and / or any devices that manipulate signals based on operational instructions. The AI / ML model 115, 125 may be implemented as combination of one or more processors for performing the CSI compression and the CSI decompression / reconstruction. In some embodiments, the AI / ML model 115, 125 may comprise a dedicated processor for performing the CSI compression and the CSI decompression / reconstruction .
[0073] In some embodiments, the AI / ML model 115, 125 may comprise one or more input layers 601, a model backbone layer 603, and an output layer 605. The AI / ML model 115, 125 may also provide temporal domain adaptation. The one or more input layers 601 may process input data such as CSI data to extract relevant features while adapting to different input configurations. As an example, the one or more input layers 601 may include Convolutional Neural Network (CNN), or alike. Further, the model backbone layer 603 may comprise of the CNN or a transformer that serves as a core of the AI / ML model 115, 125. The model backbone layer 603 may be used process the CSI data and transform the processed data into a desired output (such as compressed CSI). Further, the output layer 605 may be a fully connected layer that process the compressed or transformed features and generates a final output.
[0074] Furthermore, the temporal domain adaptation refers to the ability of a model to adapt to temporal changes in the wireless channel over time. Thus, the temporal domain adaptation ensures that the AI / ML model 115, 125 may efficiently handle the dynamic changes and provides accurate CSI compression or reconstruction overtime. The standardized AI / ML model 115, 125 may support scalability and may be compatible with various input types, including the number of transmission (Tx) ports, CSI feedback payload sizes, and bandwidths. Additionally, the AI / ML model 115, 125 structure adapts to accommodate different configurations, ensuring its applicability in diverse communication environments. The AI / ML model 115, 125 may be allowed to have additional layers or parameters to scale with respect to the various input types, the numbers of Tx ports, the CSI feedback payload sizes, and the bandwidths. Thesehyperparameters (e.g., number of layers, layer sizes) of the AI / ML model 115, 125 may be standardized but adaptable for specific scenarios. Thus, temporal adaptation features and latent space expansion ensure scalability and reliable performance under varying bandwidths, payload sizes, and network configurations.
[0075] The present disclosure mitigates dataset mismatches issues by enabling UE-specific training through the first method and the second method, which directly address disparities between the UE and network datasets. Further, nominal decoders are employed to ensure alignment between both the network side and the UE side. Furthermore, proprietary information sharing may be streamlined by utilizing standardized structures as described in third method, thereby reducing the need for extensive proprietary information exchange and promoting better collaboration between vendors (i.e., the network entity 101 and the UE 103).
[0076] The present disclosure provides a solution that ensures efficient CSI compression by aligning data representations and reducing the impact of mismatches during CSI compression. By optimizing the CSI compression process, the present disclosure may minimize the need for complex interactions between different vendors’ systems, ensuring seamless compatibility and interoperability between equipment from various manufacturers. Further, various techniques of the present disclosure may provide with data segmentation and efficient transmission mechanisms that reduce the overhead associated with training and feedback exchanges. By introducing associated IDs and Model IDs, various techniques of the present disclosure resolves inconsistencies in phase normalization, improving reconstruction quality and resolving phase normalization issues.
[0077] Additionally, the present disclosure provides novel techniques for harmonizing datasets, reducing vendor-specific dependencies, and ensuring scalability of the solution in diverse multi-vendor environments. Various techniques of the present disclosure provides effective CSI compression even under various assumptions such as antenna configurations and down tilt angles. Thus, various techniques of the present disclosure may facilitate more efficient utilization of communication resources, enhance system performance, and streamline the deployment of multi-vendor solutions, thereby reducing overall system complexity.
[0078] FIG. 7 illustrates a flowchart illustrating an exemplary method 700 for performing CSI reconstruction, in accordance with some embodiments of the present disclosure. The various operations of the method 700 may be performed by the network entity 101 of FIG. 1A. The various operations of the method 700 may be described in accordance with the “direction A” discussed in the aforementioned paragraphs. FIG. 7 is described in conjunction with FIGS.1A-3.
[0079] The method 700 may include, at block 701, training the network side encoder (El) 111 and the network side decoder (DI) 113 using the first set of data associated with the network entity 101. The method 700 may include, at block 703, transmitting at least one of: the network side encoder (El) 111 or at least the portion of the first set of data to the UE 103. The method 700 may include, at block 705, receiving the CCSI from the UE 103. The method 700 may include, at block 707, reconstructing the received CCSI using the network side decoder (DI) 113 to generate the RCSI.
[0080] FIG. 8 illustrates a flowchart illustrating an exemplary method 800 for performing CSI compression, in accordance with some embodiments of the present disclosure. The various operations of the method 800 may be performed by the UE 103 of FIG. 1A. The various operations of the method 800 may be described in accordance with the “direction A” discussed in the aforementioned paragraphs. FIG. 8 is described in conjunction with FIGS. 1A-3.
[0081] The method 800 may include, at block 801, receiving at least one of: the network side encoder (E 1) 111 or at least the portion of the first set of data associated with the network entity 101. The method 800 may include, at block 803, generating the UE side decoder (D2) 123 based on at least one of: the network side encoder (El) 111 or the portion of the first set of data associated with the network entity 101. The method 800 may include, at block 805, generating the UE side encoder (E2) 121 based on the generated UE side decoder (D2) 123 and the first set of data associated with the UE 103. The method 800 may include, at block 807, compressing the CSI using the generated UE side encoder (E2) 121. The method 800 may include, at block 809, transmitting the CCSI to the network entity 101.
[0082] FIG. 9 illustrates a flowchart illustrating another exemplary method 900 for performing CSI compression and reconstruction, in accordance with some embodiments of the present disclosure. The various operations of the method 900 may be performed by the network entity 101 and UE 103 of FIG. 1A. The various operations of the method 900 may be described in accordance with the “direction B” discussed in the aforementioned paragraphs. FIG. 9 is described in conjunction with FIGS. 1A-1B, 4 and 6.
[0083] The method 900 may include, at block 901, training, at the network entity 101, the network side encoder (El) 111 and the network side decoder (DI) 113 using the first set of data associated with the network entity 101. The method 900 may include, at block 903, transmitting, by the network entity 101, the network side encoder (El) 111 to the UE 103. The method 900 may include, at block 905, receiving, at the UE 103, the network side encoder (El) 111 from the network entity 101. The method 900 may include, at block 907, compressing, at the UE 103, the CSI using the received network side encoder (El) 111. The method 900 mayinclude, at block 909, transmitting, by the UE 103, the CCSI to the network entity 101. The method 900 may include, at block 911, receiving, at the network entity 101, the CCSI from the UE 103. The method 900 may include, at block 913, reconstructing, at the network entity 101, the received CCSI using the network side decoder (DI) 113 to generate the RCSI.
[0084] FIG. 10 illustrates a flowchart illustrating another exemplary method 1000 for performing CSI compression and reconstruction, in accordance with some embodiments of the present disclosure. The various operations of the method 1000 may be performed by the network entity 101 and UE 103 of FIG. 1A. The various operations of the method 1000 may be described in accordance with the “direction C” discussed in the aforementioned paragraphs. FIG. 10 is described in conjunction with FIGS. 1A-1B and 5-6.
[0085] The method 1000 may include, at block 1001, transmitting, by the network entity 101, the indication comprising the model ID to the UE 103, the model ID comprising the pair of the encoder 502 and the decoder 504 between the network entity 101 and the UE 103. The method 1000 may include, at block 1003, receiving, at the UE 103, the indication comprising the model ID from the network entity 101. The method 1000 may include, at block 1005, determining, at the UE 103, the encoder 502 corresponding to the received model ID. The method 1000 may include, at block 1007, transmitting, by the UE 103, the CCSI to the network entity 101. The method 1000 may include, at block 1009, receiving, at the network entity 101, the CCSI from the UE 103. The method 1000 may include, at block 1011, reconstructing, at the network entity 101, the received CCSI using the decoder 504 corresponding to the model ID to generate the RCSI.
[0086] The methods 700-1000 are merely provided for exemplary purposes, and embodiments are intended to include or otherwise cover any methods or procedures for CSI compression and CSI reconstruction. The various blocks of the methods 700-1000 shown in FIGS. 7-10 respectively have been arranged in a generally sequential manner for ease of explanation. However, it is to be understood that this arrangement is merely exemplary, and it should be recognized that the processing associated with methods 700-1000 (and the blocks shown in FIGS. 7-10 respectively) can occur in a different order. Additionally, individual blocks may be deleted from the methods without departing from the scope of the subject matter described herein. It may be noted here that the subject matter of some or all embodiments described with reference to FIGS. 1-6 may be relevant for the methods 700-1000, and the same is not repeated for the sake of brevity. Furthermore, the methods 700-1000 may be implemented in any suitable hardware, software, firmware, or combination thereof.
[0087] In an embodiment, one or more computer-readable storage media may be utilized in implementing embodiments consistent with the present disclosure. A computer-readable storage medium refers to any type of physical memory on which information or data readable by a processor may be stored. Thus, a computer-readable storage medium may store instructions for execution by one or more processors, including instructions for causing the processor(s) to perform steps or stages consistent with the embodiments described herein. The term “computer-readable medium” should be understood to include tangible items and exclude carrier waves and transient signals, i.e., be non-transitory. Examples include Random Access Memory (RAM), Read-Only Memory (ROM), volatile memory, non-volatile memory, hard drives, Compact Disc (CD) ROMs, DVDs, flash drives, disks, and any other known physical storage media.
[0088] The language used in the specification has been principally selected for readability and instructional purposes, and it may not have been selected to delineate or circumscribe the inventive subject matter. Accordingly, the disclosure of the embodiments of the disclosure is intended to be illustrative, but not limiting, of the scope of the disclosure.
Claims
We Claim:
1. A method (700) of wireless communication at a network entity (101), the method (700) comprising:training (701) a network side encoder (111) and a network side decoder (113) using a first set of data associated with the network entity (101);transmitting (703) at least one of: the network side encoder (111) or at least a portion of the first set of data to a user equipment (UE) (103);receiving (705) compressed channel state information (CSI) from the UE (103); and reconstructing (707) the received compressed CSI using the network side decoder (113) to generate reconstructed CSI.
2. The method (700) as claimed in claim 1, wherein the first set of data comprises first channel state information CSI (VI), first CSI feedback (Cl), second CSI (V2), and second CSI feedback (C2), wherein:the first CSI (VI) corresponds to target CSI at the UE (103),the first CSI feedback (Cl) corresponds to CSI feedback at the UE (103),the second CSI (V2) corresponds to reconstructed target CSI, andthe second CSI feedback (C2) corresponds to CSI feedback at the network entity (101).
3. The method (700) as claimed in claim 2, wherein transmitting (703) comprises one of:transmitting the network side encoder (111) and the first CSI (V 1) to the UE (103), or transmitting the first CSI (VI) and the first CSI feedback (Cl) to the UE (103).
4. The method (700) as claimed in claim 2, wherein training (701) the network side encoder (111) and the network side decoder (113) comprises:providing the first CSI (VI) to the network side encoder (111) to generate the first CSI feedback (Cl); andproviding the second CSI feedback (C2) to the network side decoder (113) to generate the second CSI (V2).
5. The method (700) as claimed in claim 1, further comprising:before transmitting, splitting the first set of data into one or more subsets based on UE capability;allocating an associate ID, a model ID to each of the one or more subsets; andtransmitting to the UE (103), a subset having the associate ID, the model ID corresponding to the UE (103) for CSI compression.
6. The method (700) as claimed in claim 1, further comprising:transmitting a performance target to the UE (103) along with the at least one of: the network side encoder (111) or the portion of the first set of data.
7. A method (800) of wireless communication at a user equipment (UE) (103), comprising:receiving (801) at least one of: a network side encoder (111) or at least a portion of a first set of data associated with a network entity (101);generating (803) a UE side decoder (123) based on at least one of: the network side encoder (111) or the portion of the first set of data associated with the network entity (101);generating (805) a UE side encoder (121) based on the generated UE side decoder (123) and a first set of data associated with the UE (103);compressing (807) channel state information (CSI) using the generated UE side encoder (121); andtransmitting (809) the compressed CSI to the network entity (101).
8. The method (800) as claimed in claim 7, wherein the first set of data associated with the network entity (101) comprises first channel state information CSI (VI), first CSI feedback (Cl), second CSI (V2), and second CSI feedback (C2), wherein:the first CSI (VI) corresponds to target CSI at the UE (103),the first CSI feedback (Cl) corresponds to CSI feedback at the UE (103),the second CSI (V2) corresponds to reconstructed target CSI, andthe second CSI feedback (C2) corresponds to CSI feedback at the network entity (101).
9. The method (800) as claimed in claim 8, wherein receiving (801) comprises one of: receiving the network side encoder (111) and the first CSI (VI) from the network entity (101), orreceiving the first CSI (VI) and the first CSI feedback (Cl) from the network entity (101).
10. The method (800) as claimed in claim 8, wherein generating (803) the UE side decoder (123) comprises:providing the first CSI (VI) and the network side encoder (111) to determine the first CSI feedback (Cl);adding at least one type of noise (201) to the first CSI feedback (Cl) to determine a second CSI feedback (C2);providing the second CSI feedback (C2) to the UE side decoder (123) to determine the second CSI (V2);determining a difference between the second CSI (V2) and the first CSI (VI) to determine a loss between the second CSI (V2) and the first CSI (VI); andfine-tuning the UE side decoder (123) based on the determined loss.
11. The method (800) as claimed in claim 8, wherein generating (803) the UE side decoder (123) comprises:providing the first CSI (VI) and the first CSI feedback (Cl) to determine the network side encoder (111);adding at least one type of noise 201 to the first CSI feedback (C 1) to determine a second CSI feedback (C2);providing the second CSI feedback (C2) to the UE side decoder (123) to determine the second CSI (V2);determining a difference between the first CSI (VI) and the second CSI (V2) to determine a loss between the first CSI (VI) and the second CSI (V2); andfine-tuning the UE side decoder (123) based on the determined loss.
12. The method (800) as claimed in claim 7, further comprising:receiving, from the network entity (101), a subset corresponding to the UE (103) from one or more subsets of the first set of data, the subset comprising at least the associate ID, the model ID for generating the UE side encoder 121.
13. A network entity (101), comprising :a network side memory (109);one or more network side processors (107) communicatively coupled with the network side memory (109), the one or more network side processors (107) configured to:train a network side encoder (111) and a network side decoder (113) using a first set of data associated with the network entity (101);transmit at least one of: the network side encoder (111) or at least a portion of the first set of data to a user equipment (UE) (103);receive compressed channel state information (CSI) from the UE (103); and reconstruct the received compressed CSI using the network side decoder (113) to generate reconstructed CSI.
14. The network entity (101) as claimed in claim 13, wherein the first set of data comprises first channel state information CSI (VI), first CSI feedback (Cl), second CSI (V2), and second CSI feedback (C2), wherein:the first CSI (VI) corresponds to target CSI at the UE (103),the first CSI feedback (Cl) corresponds to CSI feedback at the UE (103),the second CSI (V2) corresponds to reconstructed target CSI, andthe second CSI feedback (C2) corresponds to CSI feedback at the network entity (101).
15. The network entity (101) as claimed in claim 14, wherein the one or more network side processors (107) are configured to:transmit the network side encoder (111) and the first CSI (VI) to the UE (103), or transmit the first CSI (VI) and the first CSI feedback (Cl) to the UE (103).
16. The network entity (101) as claimed in claim 14, wherein for training the network side encoder (111) and the network side decoder (113), the one or more network side processors (107) are configured to:provide the first CSI (VI) to the network side encoder (111) to generate the first CSI feedback (Cl); andprovide the second CSI feedback (C2) to the network side decoder (113) to generate the second CSI (V2).
17. The network entity (101) as claimed in claim 13, wherein the one or more network side processors (107) are further configured to:split the first set of data into one or more subsets based on UE capability before transmitting;allocate an associate ID, a model ID to each of the one or more subsets; and transmit to the UE (103), a subset having the associate ID, the model ID corresponding to the UE (103) for CSI compression.
18. The network entity (101) as claimed in claim 13, wherein the one or more network side processors (107) are further configured to:transmit a performance target to the UE (103) along with the at least one of: the network side encoder ( 111) or the portion of the first set of data.
19. A user equipment (UE) (103), comprising:a UE side memory (109);one or more UE side processors (117) communicatively coupled with the UE side memory (109), the one or more UE side processors (117) configured to:receive at least one of: a network side encoder ( 111) or at least a portion of a first set of data associated with a network entity (101);generate a UE side decoder (123) based on at least one of: the network side encoder (111) or the portion of the first set of data associated with the network entity (101);generate a UE side encoder (121) based on the generated UE side decoder (123) and a first set of data associated with the UE (103);compress channel state information (CSI) using the generated UE side encoder (121); andtransmit the compressed CSI to the network entity (101).
20. The UE (103) as claimed in claim 19, wherein the first set of data associated with the network entity (101) comprises first channel state information CSI (VI), first CSI feedback (Cl), second CSI (V2), and second CSI feedback (C2), wherein:the first CSI (VI) corresponds to target CSI at the UE (103),the first CSI feedback (Cl) corresponds to CSI feedback at the UE (103),the second CSI (V2) corresponds to reconstructed target CSI, andthe second CSI feedback (C2) corresponds to CSI feedback at the network entity (101).
21. The UE (103) as claimed in claim 20, wherein the one or more UE side processors (117) are configured to:receive the network side encoder (111) and the first CSI (VI) from the network entity (101), orreceive the first CSI (VI) and the first CSI feedback (Cl) from the network entity (101).
22. The UE (103) as claimed in claim 20, wherein for generating the UE side decoder (123), the one or more UE side processors (117) are further configured to:provide the first CSI (VI) and the network side encoder ( 111) to determine the first CSI feedback (Cl);add at least one type of noise (201) to the first CSI feedback (Cl) to determine a second CSI feedback (C2);provide the second CSI feedback (C2) to the UE side decoder (123) to determine the second CSI (V2);determine a difference between the second CSI (V2) and the first CSI (VI) to determine a loss between the second CSI (V2) and the first CSI (VI); andfine-tune the UE side decoder (123) based on the determined loss.
23. The UE (103) as claimed in claim 20, wherein for generating the UE side decoder (123), the one or more UE side processors (117) are further configured to:provide the first CSI (VI) and the first CSI feedback (Cl) to determine the network side encoder (111);add at least one type of noise (201) to the first CSI feedback (Cl) to determine a second CSI feedback (C2);provide the second CSI feedback (C2) to the UE side decoder (123) to determine the second CSI (V2);determine a difference between the first CSI (V 1) and the second CSI (V2) to determine a loss between the first CSI (VI) and the second CSI (V2); andfine-tune the UE side decoder (123) based on the determined loss.
24. The UE (103) as claimed in claim 19, wherein the one or more UE side processors (117) are further configured to:receive, from the network entity (101), a subset corresponding to the UE (103) from one or more subsets of the first set of data, the subset comprising at least the associate ID, the model ID for generating the UE side encoder (121).
25. A method (900) of wireless communication, comprising:training (901), at a network entity (101), a network side encoder (111) and a network side decoder (113) using a first set of data associated with the network entity (101);transmitting (903), by the network entity (101), the network side encoder (111) to a user equipment (UE (103));receiving (905), at the UE ( 103), the network side encoder (111) from the network entity (101);compressing (907), at the UE ( 103), a channel state information (CSI) using the received network side encoder (111);transmitting (909), by the UE (103), the compressed CSI to the network entity (101); receiving (911), at the network entity (101), the compressed CSI from the UE (103); andreconstructing (913), at the network entity (101), the received compressed CSI using the network side decoder (113) to generate a reconstructed CSI.
26. A wireless communication system, comprising:a network entity (101) configured to:train at least a network side encoder (111) and a network side decoder (113) using a first set of data associated with the network entity (101); andtransmit the network side encoder (111) to a user equipment (UE (103)); and the UE (103) is configured to:receive the network side encoder (111) from the network entity (101); compress channel state information (CSI) using the received network side encoder (111); andtransmit the compressed CSI to the network entity (101),wherein the network entity (101) is configured to receive the compressed CSI from the UE (103), and reconstruct the received compressed CSI using the network side decoder (113) to generate a reconstructed CSI.
27. A method (1000) of wireless communication, comprising:transmitting (1001), by a network entity (101), an indication comprising a model identifier (ID) to a user equipment (UE (103)), the model ID comprising a pair of an encoder and a decoder between the network entity (101) and the UE (103);receiving (1003), at the UE (103), the indication comprising the model ID from the network entity (101);determining (1005), at the UE (103), an encoder corresponding to the received model ID;transmitting (1007), by the UE (103), compressed CSI to the network entity (101); receiving (1009), at the network entity (101), the compressed CSI from the UE (103); andreconstructing (1011), at the network entity (101), the received compressed CSI using a decoder corresponding to the model ID to generate a reconstructed CSI.
28. A wireless communication system, comprising:a network entity (101) configured to:transmit an indication comprising a model identifier (ID) to a user equipment (UE (103)), the model ID comprising a pair of an encoder (502) and a decoder (504) between the network entity (101) and the UE (103);the UE (103) is configured to:receive the indication comprising the model ID from the network entity (101); determine an encoder corresponding to the received model ID; and transmit compressed CSI to the network entity (101),wherein the network entity (101) is configured to receive the compressed CSI, and reconstruct the received compressed CSI using a decoder corresponding to the model ID to generate a reconstructed CSI.