Collaboration mechanism for encoder and decoder
Patent Information
- Application Number
- PCT/CN2023/095380
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-19
- Publication Date
- 2024-11-28
Smart Images

Figure CN2023095380_28112024_PF_FP_ABST
Abstract
Description
COLLABORATION MECHANISM FOR ENCODER AND DECODERFIELD
[0001] Embodiments of the present disclosure generally relate to the field of communication, and in particular, to a terminal device, a network device, methods, apparatuses and a computer readable storage medium for a collaboration mechanism for an encoder and a decoder.BACKGROUND
[0002] The accurate feedbacks of channel state information (CSI) are extremely important for efficient transmissions of frequency division duplex (FDD) multiple input multiple output (MIMO) , but the delivery of the original CSI to the base station consumes massive uplink resources. With the development of communications, artificial intelligence (AI) -based feature extraction and image compression has been introduced into the CSI feedback enhancement.
[0003] An overall structure of a AI-based CSI feedback scheme includes a terminal device side and a network device side. The terminal device side entity includes an encoder compressing the original CSI data into latent feature vector and a quantizer transforming the float-type latent vector into fixed-point codeword for feedback. The network device side entity conversely includes a dequantizer transforming the fixed-point codeword into the latent vector and a decoder reconstructing the original CSI data. However, there are some issues to be resolved between the encoder and decoder.SUMMARY
[0004] In general, example embodiments of the present disclosure provide a solution related to a collaboration mechanism for an encoder and a decoder.
[0005] In a first aspect, there is provided a terminal device. The terminal device comprises at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the terminal device at least to: transmit, to a network device, a first quantization type associated with an encoder at the terminal device; in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receive, from the network device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; determine an output size of the encoder based on the at least one input size of the at least one converter; and transmit, to the network device, the output size of the encoder.
[0006] In a second aspect, there is provided a network device. The network device comprises at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the network device at least to: receive, from a terminal device, a first quantization type associated with an encoder at the terminal device; based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmit, to the terminal device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; receive, from the terminal device, an output size of the encoder; and activate, based on the output size, a converter for the terminal device.
[0007] In a third aspect, there is provided a method. The method comprises transmitting, at a terminal device to a network device, a first quantization type associated with an encoder at the terminal device; in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receiving, from the network device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; determining an output size of the encoder based on the at least one input size of the at least one converter; and transmitting, to the network device, the output size of the encoder.
[0008] In a fourth aspect, there is provided a method. The method comprises receiving, at a network device from a terminal device, a first quantization type associated with an encoder at the terminal device; based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmitting, to the terminal device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; receiving, from the terminal device, an output size of the encoder; and activating, based on the output size, a converter for the terminal device.
[0009] In a fifth aspect, there is provided an apparatus. The apparatus comprises means for transmitting, at a terminal device to a network device, a first quantization type associated with an encoder at the terminal device; means for in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receiving, from the network device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; means for determining an output size of the encoder based on the at least one input size of the at least one converter; and means for transmitting, to the network device, the output size of the encoder.
[0010] In a sixth aspect, there is provided an apparatus. The apparatus comprises means for receiving, at a network device from a terminal device, a first quantization type associated with an encoder at the terminal device; means for based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmitting, to the terminal device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; means for receiving, from the terminal device, an output size of the encoder; and means for activating, based on the output size, a converter for the terminal device.
[0011] In a seventh aspect, there is provided a non-transitory computer readable medium comprising program instructions for causing an apparatus to perform at least the method according to any one of the above third to fourth aspect.
[0012] In an eighth aspect, there is provided a computer program comprising instructions, which, when executed by an apparatus, cause the apparatus at least to: transmit, to a network device, a first quantization type associated with an encoder at the terminal device; in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receive, from the network device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; determine an output size of the encoder based on the at least one input size of the at least one converter; and transmit, to the network device, the output size of the encoder.
[0013] In a ninth aspect, there is provided a computer program comprising instructions, which, when executed by an apparatus, cause the apparatus at least to: receive, from a terminal device, a first quantization type associated with an encoder at the terminal device; based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmit, to the terminal device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; receive, from the terminal device, an output size of the encoder; and activate, based on the output size, a converter for the terminal device.
[0014] In a tenth aspect, there is provided a terminal device. The terminal device comprises transmitting circuitry configured to transmit a first quantization type associated with an encoder at the terminal device to a network device; receiving circuitry configured to in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receive at least one input size of at least one converter at the network device from the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; determining circuitry configured to determine an output size of the encoder based on the at least one input size of the at least one converter; and transmitting circuitry configured to transmit the output size of the encoder to the network device.
[0015] In an eleventh aspect, there is provided a network device. The network device comprises receiving circuitry configured to receive a first quantization type associated with an encoder at the terminal device from a terminal device; transmitting circuitry configured to base on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmit at least one input size of at least one converter at the network device to the terminal device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; receiving circuitry configured to receive an output size of the encoder from the terminal device; and activating circuitry configured to activate a converter for the terminal device based on the output size.
[0016] It is to be understood that the summary section is not intended to identify key or essential features of embodiments of the present disclosure, nor is it intended to be used to limit the scope of the present disclosure. Other features of the present disclosure will become easily comprehensible through the following description.BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Some example embodiments will now be described with reference to the accompanying drawings, in which:
[0018] Fig. 1A illustrates an example network environment in which example embodiments of the present disclosure may be implemented;
[0019] Fig. 1B illustrates an AI-based CSI feedback scheme related to some embodiments of the present disclosure;
[0020] Fig. 1C illustrates an exemplary autoencoder framework with scalar quantization (SQ) related to some embodiments of the present disclosure;
[0021] Fig. 1D illustrates an exemplary autoencoder framework with vector quantization (VQ) related to some embodiments of the present disclosure;
[0022] Fig. 2 illustrates an example signaling chart illustrating an example process according to some embodiments of the present disclosure;
[0023] Fig. 3 illustrates an exemplary collaboration framework according to some embodiments of the present disclosure;
[0024] Fig. 4 illustrates an example process according to some embodiments of the present disclosure;
[0025] Fig. 5 illustrates a flowchart of a method implemented at a terminal device according to some embodiments of the present disclosure;
[0026] Fig. 6 illustrates a flowchart of a method implemented at a network device according to some embodiments of the present disclosure;
[0027] Fig. 7 illustrates a simplified block diagram of an apparatus that is suitable for implementing embodiments of the present disclosure; and
[0028] Fig. 8 illustrates a block diagram of an example computer readable medium in accordance with some embodiments of the present disclosure.
[0029] Throughout the drawings, the same or similar reference numerals represent the same or similar element.DETAILED DESCRIPTION
[0030] Principles of the present disclosure will now be described with reference to some example embodiments. It is to be understood that these embodiments are described only for the purpose of illustration and help those skilled in the art to understand and implement the present disclosure, without suggesting any limitation as to the scope of the disclosure. The disclosure described herein can be implemented in various manners other than the ones described below.
[0031] In the following description and claims, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skills in the art to which this disclosure belongs.
[0032] References in the present disclosure to “one embodiment, ” “an embodiment, ” “an example embodiment, ” and the like indicate that the embodiment described may include a particular feature, structure, or characteristic, but it is not necessary that every embodiment includes the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of one skilled in the art to affect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described.
[0033] It shall 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 only used to distinguish one element from another. For example, a first element could be termed a second element, and similarly, a second element could be termed a first element, without departing from the scope of example embodiments. As used herein, the term “and / or” includes any and all combinations of one or more of the listed terms.
[0034] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a” , “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises” , “comprising” , “has” , “having” , “includes” and / or “including” , when used herein, specify the presence of stated features, elements, and / or components etc., but do not preclude the presence or addition of one or more other features, elements, components and / or combinations thereof. As used herein, “at least one of the following: <a list of two or more elements>” and “at least one of <a list of two or more elements>” and similar wording, where the list of two or more elements are joined by “and” or “or” , mean at least any one of the elements, or at least any two or more of the elements, or at least all the elements.
[0035] As used in this application, the term “circuitry” may refer to one or more or all of the following:
[0036] (a) hardware-only circuit implementations (such as implementations in only analog and / or digital circuitry) and
[0037] (b) combinations of hardware circuits and software, such as (as applicable) :
[0038] (i) a combination of analog and / or digital hardware circuit (s) with software / firmware and
[0039] (ii) any portions of hardware processor (s) with software (including digital signal processor (s) ) , software, and memory (ies) that work together to cause an apparatus, such as a mobile phone or server, to perform various functions) and
[0040] (c) hardware circuit (s) and or processor (s) , such as a microprocessor (s) or a portion of a microprocessor (s) that requires software (e.g., firmware) for operation, but the software may not be present when it is not needed for operation.
[0041] This definition of circuitry applies to all uses of this term in this application, including in any claims. As a further example, as used in this application, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. The term circuitry also covers, for example and if applicable to the particular claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in server, a cellular network device, or other computing or network device.
[0042] As used herein, the term “communication network” refers to a network following any suitable communication standards, such as Long Term Evolution (LTE) , LTE-Advanced (LTE-A) , Wideband Code Division Multiple Access (WCDMA) , High-Speed Packet Access (HSPA) , Narrow Band Internet of Things (NB-IoT) and so on. Furthermore, the communications between a terminal device and a network device in the communication network may be performed according to any suitable generation communication protocols, including, but not limited to, the third generation (3G) , the fourth generation (4G) , 4.5G, the fifth generation (5G) communication protocols, and / or beyond. Embodiments of the present disclosure may be applied in various communication systems. Given the rapid development in communications, there will of course also be future type communication technologies and systems with which the present disclosure may be embodied. It should not be seen as limiting the scope of the present disclosure to only the aforementioned system.
[0043] As used herein, the term “network device” refers to a node in a communication network via which a terminal device accesses the network and receives services therefrom. The network device may refer to a base station (BS) or an access point (AP) , for example, a node B (NodeB or NB) , an evolved NodeB (eNodeB or eNB) , a NR NB (also referred to as a gNB) , a Remote Radio Unit (RRU) , a radio header (RH) , a remote radio head (RRH) , a relay, a low power node such as a femto, a pico, and so forth, depending on the applied terminology and technology.
[0044] The term “terminal device” refers to any end device that may be capable of wireless communication. By way of example rather than limitation, a terminal device may also be referred to as a communication device, user equipment (UE) , a Subscriber Station (SS) , a Portable Subscriber Station, a Mobile Station (MS) , or an Access Terminal (AT) . The terminal device may include, but not limited to, a mobile phone, a cellular phone, a smart phone, voice over IP (VoIP) phones, wireless local loop phones, a tablet, a wearable terminal device, a personal digital assistant (PDA) , portable computers, desktop computer, image capture terminal devices such as digital cameras, gaming terminal devices, music storage and playback appliances, vehicle-mounted wireless terminal devices, wireless endpoints, mobile stations, laptop-embedded equipment (LEE) , laptop-mounted equipment (LME) , USB dongles, smart devices, wireless customer-premises equipment (CPE) , an Internet of Things (loT) device, a watch or other wearable, a head-mounted display (HMD) , a vehicle, a drone, a medical device and applications (e.g., remote surgery) , an industrial device and applications (e.g., a robot and / or other wireless devices operating in an industrial and / or an automated processing chain contexts) , a consumer electronics device, a device operating on commercial and / or industrial wireless networks, and the like. In the following description, the terms “terminal device” , “communication device” , “terminal” , “user equipment” and “UE” may be used interchangeably.
[0045] Fig. 1A illustrates an example network environment 100 in which example embodiments of the present disclosure may be implemented. The environment 100, which may be a part of a communication network, comprises terminal devices and network devices.
[0046] As illustrated in Fig. 1A, the communication network 100 may comprise a terminal device 110 (hereinafter may also be referred to as user equipment 110 or a UE 110) . The communication network 100 may further comprise a network device 120. The network device 120 may manage a cell 101. The terminal device 110 and the network device 120 may communicate data and control information to each other in the coverage of the cell. A link from the network device 120 to the terminal device 110 is referred to as a downlink (DL) , while a link from the terminal device 110 to the network device 120 is referred to as an uplink (UL) .
[0047] It is to be understood that the number of network devices and terminal devices is only for the purpose of illustration without suggesting any limitations. The environment 100 may include any suitable number of network devices and terminal devices adapted for implementing embodiments of the present disclosure. Although not shown, it would be appreciated that one or more terminal devices may be located in the environment 100.
[0048] The communications in the communication network 100 may conform to any suitable standards including, but not limited to, Global System for Mobile Communications (GSM) , LTE, LTE-Evolution, LTE-Advanced (LTE-A) , Wideband Code Division Multiple Access (WCDMA) , Code Division Multiple Access (CDMA) , GSM EDGE Radio Access Network (GERAN) , Machine Type Communication (MTC) and the like. Furthermore, the communications may be performed according to any generation communication protocols either currently known or to be developed in the future. Examples of the communication protocols include, but not limited to, the first generation (1G) , the second generation (2G) , 2.5G, 2.75G, the third generation (3G) , the fourth generation (4G) , 4.5G, the fifth generation (5G) communication protocols.
[0049] In communication networks, the accurate feedbacks of CSI are extremely important for efficient transmissions of FDD MIMO, but the delivery of the original CSI to the base station consumes massive uplink resources. With the great success of AI-based feature extraction and image compression, great breakthroughs have been made in CSI feedback enhancement by using AI / ML methods. Thereupon, a new study item (SI) is approved in Release 18 to explore the benefits of augmenting the CSI feedback by AI / machine learning (ML) techniques.
[0050] In Release 18 AI / ML study item use cases, initial set of use cases includes: CSI feedback enhancement, e.g., overhead reduction, improved accuracy, prediction; beam management, e.g., beam prediction in time, and / or spatial domain for overhead and latency reduction, beam selection accuracy improvement; positioning accuracy enhancements for different scenarios including, e.g., those with heavy non-line of sight (NLOS) conditions.
[0051] Fig. 1B illustrates an AI-based CSI feedback scheme related to some embodiments of the present disclosure. As shown in Fig. 1B, the overall structure of the AI-based CSI feedback scheme 100B includes a terminal device side and a network device side. The terminal device side entity includes an encoder compressing the original CSI data into latent feature vector and a quantizer transforming the float-type latent vector into fixed-point codeword for feedback. The network (NW) side entity conversely includes a dequantizer transforming the fixed-point codeword into the latent vector and a decoder reconstructing the original CSI data.
[0052] As a necessary step to produce codeword for feedback after AI compression, envisioned two types of quantization schemes is envisioned, i.e., SQ and VQ. The SQ is designed to be simple and easy to implement, while the VQ exhibits superior performance in terms of encoding efficiency and robustness.
[0053] In CSI compression using two-sided model use case, evaluate and study quantization of CSI feedback, including at least the following aspects: quantization non-aware training; quantization-aware training; quantization methods including uniform vs non-uniform quantization, scalar versus vector quantization, and associated parameters, e.g., quantization resolution, etc.; how to use the quantization methods.
[0054] Fig. 1C illustrates an exemplary autoencoder framework with SQ related to some embodiments of the present disclosure. As shown in Fig. 1C, the framework 100C performs SQ-based CSI compression and reconstruction. The terminal device first creates and initializes an encoder, a scalar quantizer, and a decoder. The CSI data χ is taken as input to the encoder and is compressed into latent vector ξ with a dimension of Ls. Each element in ξ is quantized into a new scalar by a scalar quantizer. belongs to a finite set with 2b optional scalars, where b represents the number of bits carrying per latent element. After aggregating all quantized elements, the whole quantized latent vector qs is fed into the decoder to produce the reconstructed CSI
[0055] The input of UE-encoder is channel associated eigenvector χ, and the output of the UE-encoder after quantization is codeword z (χ) with a total number of (Ls·b) bits. The codeword z (χ) is transmitted over the air from UE to NW. The input of NW-decoder is qs (χ) with a size of Ls obtained from z (χ) . The output of the NW-decoder is reconstructed channel associated eigenvector
[0056] Fig. 1D illustrates an exemplary autoencoder framework with vector quantization (VQ) related to some embodiments of the present disclosure. As shown in Fig. 1D, the framework 100D performs VQ-codebook based CSI compression and reconstruction. UE firstly creates and initializes an encoder, a vector quantizer, and a decoder. Then the CSI data χ is taken as input to the encoder and is compressed into a latent vector ξ with a dimension of Lv. Since ξ is a long vector, it’s segmentized into W short pieces, each with a uniformed segment length of S. By looking up in the codebook (with a size of B, i.e., having B codevectors in the codebook) , latent segments ξ1, ξ2, ξW finds its closest codevectors in the codebook, e.g., the 3rd, Bth and B-1th, so it outputs codeword z of {3, B and B-1} or represented in a binary form. Per the codeword z, new latent segments are generated as and assembled into a vector qv. With qv as input to the decoder, a CSI is reconstructed.
[0057] The input of UE-encoder is channel associated eigenvector χ, and the output of UE-encoder-output after quantization is codeword z (χ) with a total number of W*log2 (B) bits. The codeword z (χ) is transmitted over the air from UE to NW. The input of NW-decoder is qv (χ) with a size of W*Sobtained from codeword z (χ) and preserved codebook. The output of NW-decoder is reconstructed channel associated eigenvector
[0058] When comparing among non-AI CSI feedback (Type-I / II) , SQ-based AI-CSI, VQ-based AI-CSI, VQ-AI-CSI is better than SQ-AI-CSI which is in turn better than non-AI-CSI in terms of performance. For example, VQ is more powerful than SQ simply for its capability of exploiting the correlations that exist among neighboring samples in an image or CSI, which scalar quantization does not. On the other hand, VQ-AI-CSI is greater than SQ-AI-CSI in terms of complexity. For instance, for complexity in training, VQ-AI-CSI is slightly different from SQ-AI-CSI. VQ-AI-CSI requires joint learning of encoder, decoder and codebook instead of mere the encoder and decoder in SQ-AI-CSI. For complexity in inference: in contrary to SQ-AI-CSI, VQ-AI-CSI must look up in the codebook for proper indices for both encoder and decoder.
[0059] The SQ and VQ are envisioned regarding the compressed CSI (i.e. latent vector) , there is an urgent need to devise a comprehensive NW-side decoder to simultaneously accommodate the SQ-based encoder and VQ-based encoder. Currently, all AI-CSI quantization scheme for encoder and decoder are homogeneous, that is, SQ-encode with SQ-decoder or VQ-encoder with VQ-decoder. There is no heterogeneous quantization between encoder and decoder, e.g., UE-side SQ-encoder with NW-side VQ-decoder, so latent elements of SQ-encoder are not compatible to VQ-decoder.
[0060] There is urgent request to facilitate NW-side VQ-decoder to both accommodate UE-side SQ-encoder and UE-side VQ-encoder at the same time. Therefore, the most fundamental problem to solve is the mismatch or conflict of learnt latent element between UE-side SQ-encoder output and NW-side VQ-decoder input. For example, size conflict is existed between UE-side SQ-encoder output and NW-side VQ-decoder input. For UE side, the output codeword by a UE-1 SQ-encoder is of 128 bits which consists of 64 latent elements each with 2-bit quantization. On the other hand, NW-side VQ-decoder which has been serving legacy UE-2 VQ-encoder is receiving 126 bits as input, where 126 consists of 21 vectors each is with 6 bits (using a codebook consisting of 2^6 codewords) . On top of mismatch / conflict of learnt latent element, there’s a constrain that the NW-side VQ-decoder must provide sustainable service to legacy UE-side VQ-encoder.
[0061] According to some embodiments of the present disclosure, a solution is provided for a collaboration mechanism for an encoder and a decoder. In one aspect of the solution of the present disclosure, a terminal device transmits a first quantization type associated with an encoder at the terminal device to a network device. In the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, the terminal device receives at least one input size of at least one converter at the network device from the network device. The at least one converter is configured to convert the first quantization type to the second quantization type. Based on the at least one input size of the at least one converter, the terminal device determines an output size of the encoder. Then the terminal device transmits the output size of the encoder to the network device. As such, a second quantization type of decoder at a network device side may serve both the first quantization type of encoder and the second quantization type of encoder at terminal device side simultaneously. Thereby the communication efficiency is improved.
[0062] Example embodiments of the present disclosure for data volume prediction will be described below with reference to Figs. 2-8.
[0063] Fig. 2 illustrates an example signaling chart illustrating an example process 200 according to some embodiments of the present disclosure. For the purpose of discussion, the process 200 will be described with reference to Fig. 1. The process 200 may involve the terminal device 110, the network device 120. It would be appreciated that although the process 200 has been described in the communication environment 100 of Fig. 1, this process 200 may be likewise applied to other communication scenarios with similar issues.
[0064] In the process 200, the terminal device 110 transmits 201 a first quantization type 203 to a network device 120. The first quantization type is associated with an encoder at the terminal device 110. In other words, the terminal device 110 reports its encoder type to the network device 120. The encoder type may comprise SQ and VQ.
[0065] After receiving 205 the first quantization type 203, based on determining 207 that the first quantization type is different from a second quantization type associated with a decoder at the network device, the network device 120 transmits 209 at least one input size 211 of at least one converter at the network device 120 to the terminal device 110. The at least one converter is configured to convert the first quantization type to the second quantization type. For example, the converter may be a SQ-VQ converter, the converter enables the collaboration between UE with SQ and NW with VQ. At the network device side, there may be one or more converter, and special supervised training procedure for converter and decoder is performed. In some embodiments, the first quantization type may comprise scalar quantization, and the second quantization type comprises vector quantization. In some embodiments, the first quantization type may comprise vector quantization, and the second quantization type comprises scalar quantization. It is to be understood that this solution is applicable for scenarios where UE and NW use different types of quantization schemes.
[0066] As the functionality of SQ-VQ converter is to transform the SQ-quantized latent vector to VQ-quantized latent vector, the input to the SQ-VQ converter is the SQ-quantized encoder output. As a result, the input size Ω of converter constrains the output size of the encoder at the terminal device 110, which in turn limits the choice of compression ratio of the encoder at the terminal device 110.
[0067] In some embodiments, prior to transmitting the at least one input size of at least one converter, the network device 120 may determine that the terminal device 110 and the network device 120 are to operate in a heterogeneous quantization mode. For example, the type associated with encoder at the terminal device 110 is SQ, and the type associated with decoder associated with the decoder at the network device 120 is VQ, then the network device 120 may determine to operate in a heterogeneous quantization mode.
[0068] Continuing with reference to Fig. 2, in the event that the first quantization type is different from the second quantization type associated with the decoder at the network device, the terminal device 110 receives 213 at least one input size of at least one converter from the network device 120. The terminal device 110 determines 215 an output size of the encoder based on the at least one input size of the at least one converter.
[0069] For a given input with size L and compression ratio (CR) , the terminal device 110 determines its output size (i.e. compressed CSI Ls) . For example, in the case of CSI with 32 ports and 13 subbands, the input to the encoder has a size of 32*13*2=832. The encoder with 3 optional compression ratios such as CR= {1 / 13, 1 / 32, 1 / 64} respectively leads to Ls belonging to Ψ= {64, 26, 13} . Once the terminal device 110 is aware of the optional input sizes Ω of the SQ-VQ converter, UE determines its compression ratio according to the uplink resources, QoS requirement and Ω.
[0070] In some embodiments, the network device 120 may further transmit an indication that the network device is configured with the at least one converter to the terminal device based on determining that the first quantization type is different from the second quantization type. In some embodiments, the terminal device 110 may receive an indication that the network device is configured with the at least one converter from the network device in the event that the first quantization type is different from the second quantization type. For example, the network device 120 informs the terminal device 110 of its SQ-VQ converter capability. The SQ-VQ converter capability includes whether the network device 120 can serve a terminal device with a SQ encoder. If yes, the network device 120 may indicate the acceptable input sizes Ω of the SQ-VQ converter such as Ω= {26, 64, 128} .
[0071] In some embodiments, in order to determine the output size, the terminal device 110 may determine a first compression ratio set compatible with the at least one converter based on the at least one input size of the at least one converter. Based on an intersection of the first compression ratio set and a second compression ratio set of the encoder, the terminal device 110 may determine at least one candidate compression ratio. Then the terminal device 110 may determine a compression ratio from the at least one candidate compression ratio. Based on the compression ratio, the terminal device 110 may determine the output size.
[0072] In some embodiments, the compression ratio may be determined from the at least one candidate compression ratio based on the at least one same compression ratio, a number of uplink resources, a quality of service (QoS) requirement, or any combination of two or more of the above-mentioned items.
[0073] For example, if Ω= {26, 64, 128} , the converter-compatible CR can be one of CR1= {1 / 32, 1 / 13, 2 / 13} with an original CSI size of 832. Then, combined with the UE-supported CR options CR, the actual CR can only be selected from CR∩ CR1= {1 / 32, 1 / 13} . When the uplink resources are sufficient, CR is determined as 1 / 13 to achieve good reconstruction performance. In contrary, if the uplink resource is quite limited, CR can be 1 / 32 to at least enable successful feedback of compressed CSI but with reconstruction performance degradation.
[0074] In some embodiments, the terminal device 110 may determine a quantization method of the first quantization type to be used by the terminal device after determining the output size of the encoder. The quantization method may indicate a mapping relationship between codewords and quantized latent vectors. In some embodiments, the terminal device 110 may transmit an indication of the quantization method to the network device.
[0075] For example, for the terminal device 110 indicated with a decided CR, the next step after determining the output size of the encoder is to select an appropriate quantization method, i.e. the mapping relationship between the codeword and quantized latent vector. Since there may be multiple choices in terms of the SQ implementation within one terminal device or across different terminal devices, the adopted SQ method should be reported to the network device 120 to remove the ambiguity of interpreting the UE-reported codeword to its quantized latent vector. For example, the mapping of uniform SQ can be from (00, 01, 10, 11) to (-0.75, -0.25, 0.25, 0.75) or from (00, 01, 10, 11) to (0.125, 0.375, 0.625, 0.875) . The terminal device may use non-uniform scalar quantization.
[0076] In some embodiments, the network device 120 may further receive an indication of a quantization method of the first quantization type to be used by the terminal device from the terminal device. The network device 120 may further perform dequantization of the codewords received from the terminal device based on the quantization method.
[0077] Then the terminal device 110 transmits 217 the output size of the encoder 219 to the network device 120. On the other side of the communication, the network device 120 receives 221 the output size of the encoder 219 from the terminal device 110. Based on the output size, the network device 120 activates 223 a converter for the terminal device 110.
[0078] In some embodiments, in order to activate the converter, the network device 120 may determine a model identity (ID) associated with the encoder based on the output size and the quantization method of the encoder. Based on the model ID, the network device 120 may choose the converter from the at least one converter. After that, the network device 120 may activate the converter. For example, the network device 120 may serve multiple users to achieve the alignment between the SQ encoder and the SQ-VQ converter, each user has potentially different supported CR and quantization schemes. The network device 120 may use the currently discussed model ID to differentiate between different UEs. The model ID needs to be fine enough to reveal the CR, original CSI size and the detailed quantization implementation. Typically, the model ID corresponds to a lookup table. With this model ID, the network device 120 may determine whether the CSI codeword comes from an SQ-UE or native VQ-UE and decide whether to activate a converter and which converter to activate.
[0079] In some embodiments, the terminal device 110 may further transmit CSI data encoded using the encoder based on the output size to the network device. For example, the message from the terminal device for the network side adaption includes 2 elements, CSI data (such as eigenvector or channel matrix) and CSI-associated codeword.
[0080] In some embodiments, the network device 120 may further convert first encoded channel state information (CSI) data received from the terminal device 110 into second encoded CSI data using the converter. Then the network device 120 may further decode the second encoded CSI data using the decoder.
[0081] In some embodiments, the terminal device 110 may be a first terminal device, the network device 120 may further receive the second quantization type associated with an encoder at a second terminal device from the second terminal device. Based on determining that the encoder at the second terminal device and the decoder at the network device are both associated with the second quantization type, the network device 120 may further determine that the second terminal device and the network device may be to operate in a homogeneous quantization mode. For example, the network device 120 may work with the first terminal device in SQ-VQ heterogeneous compatible mode and with SQ-VQ activated and meanwhile work with the second terminal device in VQ-VQ homogeneous mode with SQ-VQ deactivated.
[0082] In some embodiments, the network device 120 may further deactivate the at least one converter for the second terminal device, and decode encoded CSI data received from the second terminal device using the decoder. For example, the converter is deactivated for a native VQ-UE, while the converter with input size Ls is activated for SQ-UE with a encoder output size of Ls.
[0083] In this way, signaling and mutual exposure of UE and NW’s AI-model capability include at least SQ or VQ type of UE model and SQ or VQ type of NW model; at least NW side SQ-VQ converter capability and input / output size of SQ-VQ convertor; at least indication that UE and NW are operating at a homogeneous mode (SQ-SQ or VQ-VQ) or heterogeneous compatible mode (SQ-VQ or VQ-SQ) . Moreover, CSI compression on UE-side is configured per NW-side SQ-VQ converter request. According to previous signaling, UE configures a proper AI-model with corresponding CR that outputs desired codeword size that can fit NW-side SQ-VQ convertor. In addition, the SQ-VQ converter activated or deactivated based on model ID. According to previous signaling and configuration, NW-side SQ-VQ convertor is adaptively activated (for SQ-VQ heterogeneous compatible mode) or deactivated (for homogeneous mode) .
[0084] Fig. 3 illustrates an exemplary collaboration framework according to some embodiments of the present disclosure. As shown in Fig. 3, UE-1 is configured with a SQ encoder, and UE-2 is configured with a VQ encoder. GNB-1 is configured with a SQ decoder, and gNB-2 is configured with a VQ decoder and a SQ-VQ convertor. UE-1 with SQ encoder cooperates nicely with gNB-1 with SQ decoder, and meanwhile a UE-2 with a pre-trained VQ-encoder cooperates well with gNB-2 with a VQ-decoder. UE-1 may roam from gNB-1 to gNB-2, a SQ-VQ convertor is enabled to convert coderword-1 qs (x) to codeword-1’ qv (x) before sending to VQ-decoder. Meanwhile, the VQ-decoder is still directly serving codeword-2 from UE-2 for CSI-2 reconstruction. For VQ codeword, SQ-VQ convertor is skipped, and the VQ decoder is in common for both UE-1 and UE-2.
[0085] Fig. 4 illustrates an example process according to some embodiments of the present disclosure. The process 400 may involve a gNB1 401, a UE1 402, a gNB2 403, and a UE2 404. It is understood that the process 400 can be considered as a more specific example of the process 200 in Fig. 2. Thus, the a UE1 402 in Fig. 4 may be an example of the terminal device 110 in Fig. 1A or 2 and the gNB2 403 in Fig. 4 may be an example of the network device 120 in Fig. 1A or Fig. 2.
[0086] To enable a common decoder that can simultaneously serve two UEs with different types of quantization schemes (e.g. SQ-encoder and VQ-decoder collaboration) as shown in Fig 3. The gNB1 401 is configured with a SQ decoder, UE1 402 is configured with a SQ encoder. The gNB2 403 is configured with a VQ decoder and a SQ-VQ convertor, and the UE2 404 is configured with a VQ encoder.
[0087] In process 400, UE1 402 and gNB1 401 operate 405 in homogeneous operation mode (SQ-SQ) . The gNB2 403 and UE2 404 operate 410 in homogeneous operation mode (VQ-VQ) . When UE1 402 roams 415 into an area served by the gNB2 403, UE1 402 and gNB2 403 should exchange the information about their capabilities.
[0088] UE1 402 reports 420 its encoder type and capability to gNB2 403. The encoder type and capability includes at least the information about whether it’s SQ encoder or VQ encoder. The capability may be a AI-model’s capability. gNB2 403 determines 425 whether UE1 402 should work on homogeneous mode or heterogeneous compatible mode. In some embodiments, UE1 402 is with VQ encoder, then UE1 402 works on homogeneous mode. In some embodiments, UE1 402 is with SQ encoder, then UE1 402 works on heterogeneous compatible mode. Then gNB2 403 reports 430 decoder type and capability to UE1 402. The decoder type and capability includes requirement of UE1 402 output size for its SQ-VQ converter. After receiving decoder type and capability from gNB2 403, UE1 402 determines 435 UE-output size and a quantization method according to the NW SQ-VQ convertor requirement and QoS requirement. UE1 402 reports 440 UE-output size to gNB2 403. Alternatively, UE1 402 may report the quantization method to gNB2 403. gNB2 403 actives 445 SQ-VQ converter for UE1 402, and deactives 445 SQ-VQ converter for UE2 404. Then gNB2 403 performs 450 operations of heterogeneous SQ-VQ based AI-CSI with UE1 402.
[0089] In general, additional coordination procedures between the UE and the NW are introduced. These procedures include signaling and information exchange between the UE and NW entities; UE-side CSI compression configuration with SQ-VQ converter constraint; and SQ-VQ converter activation or deactivation based on model IDs.
[0090] The proposed solution of the present disclosure is verified by simulations. The simulation datasets are generated by the system level simulation (SLS) platform, which follows the channel model in 3GPP TR 38.901. The dataset containing 630K samples is drawn from a dense urban scenario (urban macro models) using the parameters shown in Table 1.
[0091] Table 1. SLS configurations for dataset generation
[0092] In the example simulations, transformer based neural network (NN) structure and SQ &VQ configuration are described as follows. For UE-side model, both the encoder and decoder on UE side employ the transformer-based architecture with 4 multi-head self-attention blocks (each block with 8 heads) , 128 embedding dimensions De and 3·De forward dimensions. As for the SQ on UE, the dimension of the encoder output (i.e. latent vector) is set as 63 and each latent vector element is quantized to 4 levels (i.e. 2 bits) , so the number of overhead bits is
[0093] For NW-side model, the NW also uses the transformer-based architecture for its autoencoder but with distinct configurations in terms of the number of attention layers and forward dimensions. Both encoder and decoder employ 6 multi-head self-attention blocks, 128 embedding dimensions De and 4·De forward dimensions. As for the VQ configuration, the number of code vectors in the codebook and the number of segments are 64 and 21 respectively, so the number of overhead bits is also
[0094] Comparisons based on the simulations are described as follows. Specifically, for the comparisons, the following NN training type were trained. Training type A is encoder-decoder jointly trained on UE with SQ, training type B is encoder-decoder jointly trained on NW with VQ; training type C is encoder-decoder independently trained on UE (with SQ) and NW (with VQ) respectively, i) without converter training and decoder fine-tuned; ii) with converter trained and decoder fine-tuned.
[0095] Regarding the performance metrics, squared generalized cosine similarity (SGCS) between the original CSI data and the reconstructed CSI data is used as the performance metric. The closer the value of SGCS is to 1, the better the performance of the model.
[0096] Some observations are provided. Table 2 presents the simulation results. The major observations are listed as follows: (1) For a) and b) , joint training on UE with SQ and joint training on NW with VQ respectively achieve good CSI reconstruction performances. (2) For c-i) , when the proposed converter is not included and the decoder is not fine-tuned, the NW-side decoder with VQ cannot reconstruct the CSI given the UE-reported codeword. (3) For c-ii) , with the SQ-VQ converter and VQ-decoder jointly trained on NW side, the NW-side VQ-decoder can reconstruct the CSI given the UE-reported codeword. There is a negligible 0.35%SGCS degradation compared with b) and slight 1.07%SGCS improvement compared with a) .
[0097] Table. 2 simulation evaluation results
[0098] Fig. 5 shows a flowchart of an example method 500 implemented at a terminal device in accordance with some embodiments of the present disclosure. For the purpose of discussion, the method 500 will be described from the perspective of the terminal device 110 with reference to Fig. 1.
[0099] At block 510, the terminal device 110 transmits a first quantization type associated with an encoder at the terminal device to a network device 120. At block 520, in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, the terminal device 110 receives at least one input size of at least one converter at the network device from the network device. The at least one converter is configured to convert the first quantization type to the second quantization type. At block 530, the terminal device 110 determines an output size of the encoder based on the at least one input size of the at least one converter. At block 540, the terminal device 110 transmits the output size of the encoder to the network device.
[0100] In some embodiments, the terminal device 110 may determine a quantization method of the first quantization type to be used by the terminal device after determining the output size of the encoder. The quantization method indicates a mapping relationship between codewords and quantized latent vectors.
[0101] In some embodiments, the terminal device 110 may transmit an indication of the quantization method to the network device. In some embodiments, the first quantization type may comprise scalar quantization, and the second quantization type comprises vector quantization.
[0102] In some embodiments, the terminal device 110 may receive an indication that the network device is configured with the at least one converter from the network device in the event that the first quantization type is different from the second quantization type.
[0103] In some embodiments, in order to determine the output size, the terminal device 110 may determine a first compression ratio set compatible with the at least one converter based on the at least one input size of the at least one converter. Based on an intersection of the first compression ratio set and a second compression ratio set of the encoder, the terminal device 110 may determine at least one candidate compression ratio. Then the terminal device 110 may determine a compression ratio from the at least one candidate compression ratio. Based on the compression ratio, the terminal device 110 may determine the output size.
[0104] In some embodiments, the compression ratio may be determined from the at least one candidate compression ratio based on the at least one same compression ratio, a number of uplink resources, a quality of service (QoS) requirement, or any combination of two or more of the above-mentioned items.
[0105] In some embodiments, the terminal device 110 may further transmit channel state information (CSI) data encoded using the encoder based on the output size to the network device.
[0106] Fig. 6 shows a flowchart of an example method 600 implemented at a network device in accordance with some embodiments of the present disclosure. For the purpose of discussion, the method 600 will be described from the perspective of the network device 120 with reference to Fig. 1.
[0107] At block 610, the network device 120 receives a first quantization type associated with an encoder at the terminal device 110 from a terminal device. At block 620, based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, the network device 120 transmits at least one input size of at least one converter at the network device to the terminal device. The at least one converter is configured to convert the first quantization type to the second quantization type. At block 630, network device 120 receives an output size of the encoder from the terminal device. At block 640, the network device 120 activates a converter for the terminal device based on the output size.
[0108] In some embodiments, the network device 120 may further receive an indication of a quantization method of the first quantization type to be used by the terminal device from the terminal device. The quantization method indicates a mapping relationship between codewords and quantized latent vectors. The network device 120 may further perform dequantization of the codewords received from the terminal device based on the quantization method.
[0109] In some embodiments, the first quantization type may comprise scalar quantization, and the second quantization type comprises vector quantization. In some embodiments, the network device may further transmit an indication that the network device is configured with the at least one converter to the terminal device based on determining that the first quantization type is different from the second quantization type.
[0110] In some embodiments, in order to activate the converter, the network device 120 may determine a model identity (ID) associated with the encoder based on the output size and the quantization method of the encoder. Based on the model ID, the network device 120 may choose the converter from the at least one converter. After that, the network device 120 may activate the converter.
[0111] In some embodiments, the network device 120 may further convert first encoded channel state information (CSI) data received from the terminal device into second encoded CSI data using the converter. Then the network device 120 may further decode the second encoded CSI data using the decoder.
[0112] In some embodiments, prior to transmitting the at least one input size of at least one converter, the network device 120 may determine that the terminal device and the network device are to operate in a heterogeneous quantization mode.
[0113] In some embodiments, the terminal device 110 may be a first terminal device, the network device 120 may further receive the second quantization type associated with an encoder at a second terminal device from the second terminal device. Based on determining that the encoder at the second terminal device and the decoder at the network device are both associated with the second quantization type, the network device 120 may further determine that the second terminal device and the network device may be to operate in a homogeneous quantization mode.
[0114] In some embodiments, the network device 120 may further deactivate the at least one converter for the second terminal device, and decode encoded CSI data received from the second terminal device using the decoder.
[0115] In some embodiments, an apparatus capable of performing any of the method 500 (for example, the terminal device 110) is provided. The apparatus may comprise means for performing the respective steps of the method 500. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module.
[0116] In some embodiments, the apparatus comprises means for transmitting, to a network device, a first quantization type associated with an encoder at the terminal device; means for in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receiving, from the network device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; means for determining an output size of the encoder based on the at least one input size of the at least one converter; and means for transmitting, to the network device, the output size of the encoder.
[0117] In some embodiments, the apparatus may comprise means for determining a quantization method of the first quantization type to be used by the terminal device after determining the output size of the encoder, wherein the quantization method indicates a mapping relationship between codewords and quantized latent vectors.
[0118] In some embodiments, the apparatus may comprise means for transmitting an indication of the quantization method to the network device. In some embodiments, the first quantization type may comprise scalar quantization, and the second quantization type may comprise vector quantization.
[0119] In some embodiments, the apparatus may comprise means for in the event that the first quantization type is different from the second quantization type, receiving, from the network device, an indication that the network device is configured with the at least one converter.
[0120] In some embodiments, means for determining the output size may comprise means for determining a first compression ratio set compatible with the at least one converter based on the at least one input size of the at least one converter; means for determining at least one candidate compression ratio based on an intersection of the first compression ratio set and a second compression ratio set of the encoder; means for determining a compression ratio from the at least one candidate compression ratio; and means for determining the output size based on the compression ratio.
[0121] In some embodiments, the compression ratio may be determined from the at least one candidate compression ratio based on the at least one same compression ratio, a number of uplink resources, a quality of service (QoS) requirement, or any combination of two or more of the above-mentioned items.
[0122] In some embodiments, the apparatus may further comprise means for transmitting, to the network device, channel state information (CSI) data encoded using the encoder based on the output size.
[0123] In some embodiments, the apparatus further comprises means for performing other steps in some embodiments of the method 500. In some embodiments, the means comprises at least one processor and at least one memory including computer program code, the at least one memory and computer program code configured to, with the at least one processor, cause the performance of the apparatus.
[0124] In some embodiments, an apparatus capable of performing any of the method 600 (for example, the network device 120) is provided. The apparatus may comprise means for performing the respective steps of the method 600. The means may be implemented in any suitable form. For example, the means may be implemented in a circuitry or software module.
[0125] In some embodiments, the apparatus comprises means for receiving, from a terminal device, a first quantization type associated with an encoder at the terminal device; means for based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmitting, to the terminal device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type; means for receiving, from the terminal device, an output size of the encoder; and means for activating, based on the output size, a converter for the terminal device.
[0126] In some embodiments, the apparatus may comprise means for receiving an indication of a quantization method of the first quantization type to be used by the terminal device from the terminal device, wherein the quantization method indicates a mapping relationship between codewords and quantized latent vectors; and means for performing dequantization of the codewords received from the terminal device based on the quantization method.
[0127] In some embodiments, the first quantization type may comprise scalar quantization, and the second quantization type may comprise vector quantization. In some embodiments, the apparatus may comprise means for transmitting an indication that the network device is configured with the at least one converter to the terminal device based on determining that the first quantization type is different from the second quantization type.
[0128] In some embodiments, means for activating the converter may comprise means for determining a model identity (ID) associated with the encoder based on the output size and the quantization method of the encoder; means for choosing the converter from the at least one converter based on the model ID; and means for activating the converter.
[0129] In some embodiments, the apparatus may comprise means for converting first encoded channel state information (CSI) data received from the terminal device into second encoded CSI data using the converter, and means for decoding the second encoded CSI data using the decoder.
[0130] In some embodiments, the apparatus may comprise means for prior to transmitting the at least one input size of at least one converter, determining that the terminal device and the network device are to operate in a heterogeneous quantization mode.
[0131] In some embodiments, the apparatus may comprise means for receiving the second quantization type associated with an encoder at the second terminal device from a second terminal device, means for determining that the second terminal device and the network device may be to operate in a homogeneous quantization mode based on determining that the encoder at the second terminal device and the decoder at the network device are both associated with the second quantization type.
[0132] In some embodiments, the apparatus may comprise means for deactivating the at least one converter for the second terminal device, and means for decoding encoded CSI data received from the second terminal device using the decoder.
[0133] In some embodiments, the device further comprises means for performing other steps in some embodiments of the method 600. In some embodiments, the means comprises at least one processor and at least one memory including computer program code, the at least one memory and computer program code configured to, with the at least one processor, cause the performance of the device.
[0134] Fig. 7 is a simplified block diagram of a device 700 that is suitable for implementing embodiments of the present disclosure. The device 700 may be provided to implement the communication device, for example the terminal device 110 and the network device 120 as shown in Fig. 1. As shown, the device 700 includes one or more processors 710, and one or more communication modules 740 coupled to the processor 710. The device 700 may further include one or more memories 720 coupled to the processor 710.
[0135] The communication modules 740 may be for bidirectional communications. The communication modules 740 has at least one antenna to facilitate communication. The communication interface may represent any interface that is necessary for communication with other network elements.
[0136] The processor 710 may be of any type suitable to the local technical network and may include one or more of the following: general purpose computers, special purpose computers, microprocessors, digital signal processors (DSPs) and processors based on multicore processor architecture, as non-limiting examples. The device 700 may have multiple processors, such as an application specific integrated circuit chip that is slaved in time to a clock which synchronizes the main processor.
[0137] The memory 720 may include one or more non-volatile memories and one or more volatile memories. Examples of the non-volatile memories include, but are not limited to, a Read Only Memory (ROM) 724, an electrically programmable read only memory (EPROM) , a flash memory, a hard disk, a compact disc (CD) , a digital video disk (DVD) , and other magnetic storage and / or optical storage. Examples of the volatile memories include, but are not limited to, a random access memory (RAM) 722 and other volatile memories that will not last in the power-down duration.
[0138] A computer program 730 includes computer executable instructions that are executed by the associated processor 710. The program 730 may be stored in the ROM 724. The processor 710 may perform any suitable actions and processing by loading the program 730 into the RAM 722.
[0139] The embodiments of the present disclosure may be implemented by means of the program 730 so that the device 700 may perform any process of the disclosure as discussed with reference to Figs. 2 to 6. The embodiments of the present disclosure may also be implemented by hardware or by a combination of software and hardware.
[0140] In some embodiments, the program 730 may be tangibly contained in a computer readable medium which may be included in the device 700 (such as in the memory 720) or other storage devices that are accessible by the device 700. The device 700 may load the program 730 from the computer readable medium to the RAM 722 for execution. The computer readable medium may include any types of tangible non-volatile storage, such as ROM, EPROM, a flash memory, a hard disk, CD, DVD, and the like. Fig. 8 shows an example of the computer readable medium 800 in form of CD or DVD. The computer readable medium has the program 730 stored thereon.
[0141] Generally, various embodiments of the present disclosure may be implemented in hardware or special purpose circuits, software, logic or any combination thereof. Some aspects may be implemented in hardware, while other aspects may be implemented in firmware or software which may be executed by a controller, microprocessor or other computing device. While various aspects of embodiments of the present disclosure are illustrated and described as block diagrams, flowcharts, or using some other pictorial representations, it is to be understood that the block, apparatus, system, technique or method described herein may be implemented in, as non-limiting examples, hardware, software, firmware, special purpose circuits or logic, general purpose hardware or controller or other computing devices, or some combination thereof.
[0142] The present disclosure also provides at least one computer program product tangibly stored on a non-transitory computer readable storage medium. The computer program product includes computer-executable instructions, such as those included in program modules, being executed in a device on a target real or virtual processor, to carry out the method 400 or 500 as described above with reference to Fig. 4-Fig. 5. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, or the like that perform particular tasks or implement particular abstract data types. The functionality of the program modules may be combined or split between program modules as desired in various embodiments. Machine-executable instructions for program modules may be executed within a local or distributed device. In a distributed device, program modules may be located in both local and remote storage media.
[0143] Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may execute entirely on a machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
[0144] In the context of the present disclosure, the computer program codes or related data may be carried by any suitable carrier to enable the device, apparatus or processor to perform various processes and operations as described above. Examples of the carrier include a signal, computer readable medium, and the like.
[0145] The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable medium may include but not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the computer readable storage medium would include an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM) , a read-only memory (ROM) , an erasable programmable read-only memory (EPROM or Flash memory) , an optical fiber, a portable compact disc read-only memory (CD-ROM) , an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. The term “non-transitory, ” as used herein, is a limitation of the medium itself (i.e., tangible, not a signal) as opposed to a limitation on data storage persistency (e.g., RAM vs. ROM) .
[0146] Further, while operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while several specific implementation details are contained in the above discussions, these should not be construed as limitations on the scope of the present disclosure, but rather as descriptions of features that may be specific to particular embodiments. Certain features that are described in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment may also be implemented in multiple embodiments separately or in any suitable sub-combination.
[0147] Although the present disclosure has been described in languages specific to structural features and / or methodological acts, it is to be understood that the present disclosure defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.
Claims
1.A terminal device comprising:at least one processor; andat least one memory storing instructions that, when executed by the at least one processor, cause the terminal device at least to:transmit, to a network device, a first quantization type associated with an encoder at the terminal device;in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receive, from the network device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type;determine an output size of the encoder based on the at least one input size of the at least one converter; andtransmit, to the network device, the output size of the encoder.2.The terminal device of claim 1, wherein the terminal device is further caused to:after determining the output size of the encoder, determine a quantization method of the first quantization type to be used by the terminal device, wherein the quantization method indicates a mapping relationship between codewords and quantized latent vectors.3.The terminal device of claim 2, wherein the terminal device is further caused to:transmit an indication of the quantization method to the network device.4.The terminal device of any of claims 1-3, wherein the first quantization type comprises scalar quantization, and the second quantization type comprises vector quantization.5.The terminal device of any of claims 1-4, wherein the terminal device is further caused to:in the event that the first quantization type is different from the second quantization type, receive, from the network device, an indication that the network device is configured with the at least one converter.6.The terminal device of any of claims 1-5, wherein the terminal device is caused to determine the output size by:determining, based on the at least one input size of the at least one converter, a first compression ratio set compatible with the at least one converter;determining at least one candidate compression ratio based on an intersection of the first compression ratio set and a second compression ratio set of the encoder;determining a compression ratio from the at least one candidate compression ratio; anddetermining the output size based on the compression ratio.7.The terminal device of claim 6, wherein the compression ratio is determined from the at least one candidate compression ratio based on at least one of the following:the at least one same compression ratio;a number of uplink resources; ora quality of service (QoS) requirement.8.The terminal device of any of claims 1-7, wherein the terminal device is further caused to:transmit, to the network device, channel state information (CSI) data encoded using the encoder based on the output size.9.A network device comprising:at least one processor; andat least one memory storing instructions that, when executed by the at least one processor, cause the network device at least to:receive, from a terminal device, a first quantization type associated with an encoder at the terminal device;based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmit, to the terminal device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type;receive, from the terminal device, an output size of the encoder; andactivate, based on the output size, a converter for the terminal device.10.The network device of claim 9, wherein the network device is further caused to:receive, from the terminal device, an indication of a quantization method of the first quantization type to be used by the terminal device, wherein the quantization method indicates a mapping relationship between codewords and quantized latent vectors; andperform, based on the quantization method, dequantization of the codewords received from the terminal device.11.The network device of claim 9 or 10, wherein the first quantization type comprises scalar quantization, and the second quantization type comprises vector quantization.12.The network device of any of claims 9-11, wherein the network device is further caused to:based on determining that the first quantization type is different from the second quantization type, transmit, to the terminal device, an indication that the network device is configured with the at least one converter.13.The network device of any of claims 9-12, wherein the network device is further caused to activate the converter by:determining a model identity (ID) associated with the encoder based on the output size and the quantization method of the encoder;based on the model ID, choosing the converter from the at least one converter; andactivating the converter.14.The network device of any of claims 9-13, wherein the network device is further caused to:convert, using the converter, first encoded channel state information (CSI) data received from the terminal device into second encoded CSI data; anddecode the second encoded CSI data using the decoder.15.The network device of any of claims 9-14, wherein the network device is further caused to:prior to transmitting the at least one input size of at least one converter, determine that the terminal device and the network device are to operate in a heterogeneous quantization mode.16.The network device of any of claims 9-15, wherein the terminal device is a first terminal device, and the network device is further caused to:receive, from a second terminal device, the second quantization type associated with an encoder at the second terminal device;based on determining that the encoder at the second terminal device and the decoder at the network device are both associated with the second quantization type, determine that the second terminal device and the network device are to operate in a homogeneous quantization mode.17.The network device of claim 16, wherein the network device is further caused to:deactivate the at least one converter for the second terminal device; anddecode, using the decoder, encoded CSI data received from the second terminal device.18.A method comprising:transmitting, at a terminal device to a network device, a first quantization type associated with an encoder at the terminal device;in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receiving, from the network device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type;determining an output size of the encoder based on the at least one input size of the at least one converter; andtransmitting, to the network device, the output size of the encoder.19.A method comprising:receiving, at a network device from a terminal device, a first quantization type associated with an encoder at the terminal device;based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmitting, to the terminal device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type;receiving, from the terminal device, an output size of the encoder; andactivating, based on the output size, a converter for the terminal device.20.An apparatus comprising:means for transmitting, at a terminal device to a network device, a first quantization type associated with an encoder at the terminal device;means for in the event that the first quantization type is different from a second quantization type associated with a decoder at the network device, receiving, from the network device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type;means for determining an output size of the encoder based on the at least one input size of the at least one converter; andmeans for transmitting, to the network device, the output size of the encoder.21.An apparatus comprising:means for receiving, at a network device from a terminal device, a first quantization type associated with an encoder at the terminal device;means for based on determining that the first quantization type is different from a second quantization type associated with a decoder at the network device, transmitting, to the terminal device, at least one input size of at least one converter at the network device, wherein the at least one converter is configured to convert the first quantization type to the second quantization type;means for receiving, from the terminal device, an output size of the encoder; andmeans for activating, based on the output size, a converter for the terminal device.22.A non-transitory computer readable medium comprising program instructions that, when executed by an apparatus, cause the apparatus to perform at least the method of claim 18 or 19.
Citation Information
Patent Citations
Efficient MIMO channel feedback method and device based on network aggregation strategy
CN113381950A
Channel quantitative feedback method and device, electronic equipment and storage medium
CN115001629A
Neural network-based channel state information feedback
CN115136505A
Method and device for performing sensing procedure in wireless LAN system
WO2023075386A1
Cited By
Method and apparatus for data sharing and model training for compressing and restoring CSI in wireless communication system
WO2026151337A1
