Channel state information feedback

Abstract

Embodiments of the present disclosure relate to channel state information feedback. In one aspect, a terminal device compresses channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option of a plurality of compression ratio options, where the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and sending the codeword and the compression ratio option to the network device. The embodiments of the present disclosure can process adaptive CSI compression and decompression under various bandwidth configurations.

Description

TECHNICAL FIELD

[0001] Various example embodiments relate to the field of telecommunications, and in particular to methods, devices, apparatuses, and computer-readable storage media for channel state information feedback. BACKGROUND

[0002] In the field of communications, there is an ongoing evolution in order to provide efficient and reliable solutions for utilizing wireless communication networks. Each new generation of technology has the technical challenge of handling different situations and procedures needed to connect and serve devices connected to wireless networks. In order to meet the demand for wireless data traffic that has increased since the deployment of fourth generation (4G) communication systems, efforts have been made to develop improved fifth generation (5G) or pre-5G communication systems. New communication systems can support various types of service applications of terminal devices.

[0003] In recent years, AI / ML based channel state information (CSI) feedback methods have been explored to improve reconstruction accuracy and reduce overhead. In some AI / ML based CSI feedback frameworks, an autoencoder architecture is employed, where the original CSI X is compressed by an encoder at the terminal device into a codeword Z , which is then transmitted to the network device. After receiving the CSI feedback codeword Z , a decoder at the network device can reconstruct the CSI . SUMMARY

[0004] Generally, example embodiments of the present disclosure provide a solution for channel state information feedback.

[0005] In a first aspect, a terminal device is provided. The terminal device comprises at least one processor and at least one memory including computer program code; the at least one memory and the computer program code configured to, with the at least one processor, cause the terminal device to: compress channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option among a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and transmit the codeword and the compression ratio option to a network device.

[0006] In a second aspect, a network device is provided. The network device includes at least one processor and at least one memory including computer program codes; the at least one memory and the computer program codes are configured to, with the at least one processor, cause the network device to: receive, from a terminal device, a codeword and a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and decompress the codeword based on the compression ratio option to obtain channel state information (CSI) data.

[0007] In a third aspect, a method is provided. The method includes: compressing, at a terminal device, channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and sending, to a network device, the codeword and the compression ratio option.

[0008] In a fourth aspect, a method is provided. The method includes: receiving, at a network device, from a terminal device, a codeword and a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and decompressing the codeword based on the compression ratio option to obtain channel state information (CSI) data.

[0009] In a fifth aspect, an apparatus is provided. The apparatus includes: means for compressing, at a terminal device, channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and means for sending, to a network device, the codeword and the compression ratio option.

[0010] In a sixth aspect, an apparatus is provided. The apparatus includes: means for receiving, at a network device, from a terminal device, a codeword and a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and means for decompressing the codeword based on the compression ratio option to obtain channel state information (CSI) data.

[0011] In a seventh aspect, a non-transitory computer readable medium is provided, the medium including program instructions for causing an apparatus to perform at least the method according to any one of the above third aspect to fourth aspect.

[0012] In an eighth aspect, there is provided a non-transitory computer readable medium comprising program instructions stored thereon for performing at least the method of any one of the third aspect to the fourth aspect.

[0013] In a ninth aspect, there is provided a computer program comprising instructions which, when executed by an apparatus, cause the apparatus to at least: compress channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and transmit the codeword and the compression ratio option to a network device.

[0014] In a tenth aspect, there is provided a computer program comprising instructions which, when executed by an apparatus, cause the apparatus to at least: receive, from a terminal device, a codeword and a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and decompress the codeword based on the compression ratio option to obtain channel state information (CSI) data.

[0015] In an eleventh aspect, there is provided a terminal device. The terminal network device comprises: compression circuitry configured to compress channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and transmission circuitry configured to transmit the codeword and the compression ratio option to a network device.

[0016] In a twelfth aspect, there is provided a network device. The network device comprises: reception circuitry configured to receive, from a terminal device, a codeword and a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and decompression circuitry configured to decompress the codeword based on the compression ratio option to obtain channel state information (CSI) data.

[0017] It is to be understood that the Summary is not intended to identify key or essential features of embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features, details, and advantages of the disclosure will become BRIEF DESCRIPTION OF DRAWINGS

[0018] Some example embodiments will now be described with reference to the drawings, in which:

[0019] Figure 1FIG. illustrates an example communication network in which embodiments of the present disclosure can be implemented;

[0020] Figure 2 FIG. illustrates a flow diagram of a communication procedure between a terminal device and a network device according to some embodiments of the present disclosure;

[0021] Figure 3 FIG. illustrates a schematic diagram of supporting CSI compression ratio option selection according to some embodiments of the present disclosure;

[0022] Figure 4 and Figure 5 collectively illustrate a schematic diagram of pre-processing operations of CSI input according to some embodiments of the present disclosure;

[0023] Figure 6 FIG. illustrates a schematic diagram of splitting feature vector based CSI data into patches according to some embodiments of the present disclosure;

[0024] Figure 7 FIG. illustrates a schematic diagram of CSI pre-processing according to some embodiments of the present disclosure;

[0025] Figure 8 FIG. illustrates a schematic diagram of a process architecture according to some embodiments of the present disclosure;

[0026] Figure 9 FIG. illustrates a schematic diagram of a general transformer enabled CSI feedback scheme according to some embodiments of the present disclosure;

[0027] Figure 10 FIG. illustrates a schematic diagram of a transformer enabled general encoder design according to some embodiments of the present disclosure;

[0028] Figure 11 FIG. illustrates a schematic diagram of an internal architecture of a transformer enabled general encoder according to some embodiments of the present disclosure;

[0029] Figure 12 FIG. illustrates a schematic diagram of an internal architecture of a transformer enabled general decoder according to some embodiments of the present disclosure;

[0030] Figure 13 FIG. illustrates a flow diagram of a communication procedure between a terminal device and a network device according to some other embodiments of the present disclosure;

[0031] Figure 14 FIG. illustrates a schematic diagram of CSI reconstruction accuracy (in GCS) of CSI data sets under various bandwidth configurations;

[0032] Figure 15 FIG. illustrates a schematic diagram of CSI reconstruction accuracy (in GCS) of baseline schemes and proposed schemes with different codeword lengths;

[0033] Figure 16 FIG. 1 illustrates a flow diagram of a method implemented at a terminal device, according to some embodiments of the present disclosure;

[0034] Figure 17 FIG. 1 illustrates a flow diagram of a method implemented at a network device, according to some embodiments of the present disclosure;

[0035] Figure 18 FIG. 1 illustrates a simplified block diagram of a device suitable for implementing embodiments of the present disclosure; and

[0036] Figure 19 FIG. 1 illustrates a block diagram of an example computer-readable medium, according to some embodiments of the present disclosure.

[0037] Throughout the drawings, identical or similar reference numerals can designate identical or similar elements throughout the drawings. DETAILED DESCRIPTION

[0038] The principles of the present disclosure will now be described with reference to some example embodiments. It should be understood that these embodiments are described for illustrative purposes only and help the skilled person to understand and implement the present disclosure, and do not represent any limitation on the scope of the present disclosure. The disclosure described herein can be implemented in various other ways than described below.

[0039] In the following description and claims, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

[0040] In the present disclosure, references to “one embodiment”, “an embodiment”, “example embodiments” etc. mean that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. The appearances of the phrase “in one embodiment” in various places in the specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0041] It can be understood that, although the terms “first” and “second” etc. can 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 the example embodiments. As used herein, the term “and / or” includes any and all combinations of one or more of the associated terms.

[0042] 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 " and "one or more of " along with similar phrases referring to elements using "a" or "an" refer to any one of these elements, or to any combination or subset of these elements.

[0043] As used in this application, the term "circuitry" can refer to one or more or all of the following: (a) hardware-only circuit implementations (such as implementations in only analog and / or digital circuitry) and (b) combinations of hardware circuits and software, such as (as applicable): (i) a combination of analog and / or digital hardware circuit(s) with software / firmware and (ii) any portions of hardware processor(s) with software (including digital signal processors), software, and memory(ies) that work together to cause an apparatus, such as a mobile phone or server, to perform various functions and (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 need not necessarily have such software present.

[0044] 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 that includes both a hardware circuit and / or a processor (or multiple processors), and accompanying software or firmware that works together to cause an apparatus, such as a mobile phone or server, to perform various functions described herein.

[0045] As used herein, the term “communication network” refers to a network that follows any suitable communication standard, 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), etc. Furthermore, the communication between terminal devices and network devices in the communication network can be performed according to any suitable generation communication protocol, including 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 future fifth generation (5G) communication, and / or any other protocol that is currently known or will be developed in the future. Embodiments of the present disclosure can be applied in various communication systems. In consideration of the rapid development of communication, there will certainly be future types of communication technology and systems that can be used to embody the present disclosure. It should not be considered as limiting the scope of the present disclosure to only the above-mentioned systems.

[0046] 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 from the network. The network device can refer to a base station (BS) or an access point (AP), e.g., a NodeB (NB), an evolved NodeB (eNodeB or eNB), an NR NB (also known as gNB, 5G base station), a remote radio unit (RRU), a radio head (RH), a remote radio head (RRH), a relay, a low power node (such as femto, pico), etc., depending on the terminology used and technology.

[0047] The term "terminal device" refers to any terminal device capable of wireless communication. By way of example, and without limitation, a terminal device can 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). A terminal device can include, but is not limited to, a mobile phone, a cellular phone, a smart phone, a voice over IP (VoIP) phone, a wireless local loop phone, a tablet, a wearable terminal device, a personal digital assistant (PDA), a portable computer, a desktop computer, an image capture terminal device, such as a digital camera, a game terminal device, a music storage and playback appliance, a vehicle-mounted wireless terminal device, a wireless endpoint, a mobile station, a laptop-embedded equipment (LEE), a laptop-mounted equipment (LME), a USB dongle, a smart device, a wireless customer-premises equipment (CPE), an Internet of Things (loT) device, a watch or other wearable device, a head-mounted display (HMD), a vehicle, a drone, medical devices and applications (e.g., remote surgery), industrial devices and applications (e.g., robots and / or other wireless devices operating in industrial and / or

[0048] In practical wireless communication systems, in addition to the design of neural network (NN) architecture, the compression ratio (CR) for balancing the CSI (channel state information) reconstruction accuracy and feedback overhead is an important factor affecting the CSI feedback performance. Some researchers have studied the variable CR problem in DL-based CSI feedback. Some researchers have adopted an additional classification block that is used to determine the CR of the encoder according to the channel in a supervised manner.

[0049] From a standardization perspective, 3GPP Release 18 has approved to study the benefits of augmenting the air interface through the latest advances in the field of AI / ML, and CSI feedback enhancement is the first representative use case.

[0050]

[0051] For AI / ML models, 3GPP has studied the model lifecycle management (LCM) on the UE side and the network side, including model selection, activation, deactivation, switching, and fallback.

[0052]

[0053] For the AI / ML based CSI feedback enhancement use case, in the final report of 3GPP TSG RAN WG1 #110 v1.0.0, 3GPP evaluated the CSI feedback performance under different configurations, which requires detailed specification of the corresponding model LCM signaling under different configurations.

[0054]

[0055] For the evaluation of the AI / ML based CSI compression sub-use case, the precoding matrix is considered as the input and output of the AI / ML model in 3GPP (Final Report of 3GPP TSG RAN WG1 #110bis-e v1.0.0).

[0056]

[0057] For the DL based CSI feedback enhancement, the CSI feedback codebook length is a key factor to determine the CSI feedback accuracy and the final beamforming performance. The CSI feedback codebook length is mainly affected by two factors, i.e., the CSI feedback accuracy requirement of beamforming and the pre-determined signal inherent bandwidth (i.e., subband number, antenna port number, rank, etc.) configuration at the transmitter side. In practice, due to the flexible beamforming request and the wireless propagation environment, these two determinants are very dynamic, respectively.

[0058] However, most of the DL based CSI feedback solutions are dedicated to fixed RF configurations, i.e., specific subband number with fixed CR. In order to adapt to different CSI compression tasks with various RF configurations, multiple DL models need to be pre-stored and correctly selected. For example, in order to adapt to the CSI data with two different subband configurations and two different required CSI feedback codebook lengths of the feature vector format, the UE needs to store 2 4 CSI feedback DL models, which is impractical in terms of UE storage and feasibility. However, in the generalization of the DL based CSI feedback enhancement solution, few people consider to handle adaptive CSI compression under various bandwidth configurations.

[0059] In one scheme, researchers use a fully connected structured neural network to study 2x, 4x and 8x CR based on 1x CR using a serial or parallel framework. Although this scheme is limited to discrete CR of 2 power, the feasibility of jointly learning multiple compression schemes has been proven. The overall consideration of adaptive CR decision and management regarding the original input CSI size (e.g., subband or / and other RF configurations) and the CSI reconstruction accuracy / QoS request of beamforming has not been considered.

[0060] In another approach, researchers used low-complexity CNNs for CSI classification, which was initially aimed at providing multi-level compression ratios related to the complexity of the wireless propagation environment. This work did not address the above issues, and it revealed a key fact that CSI compression ratios need to be very flexible given the complex wireless environment.

[0061] There are two aspects of the problem: one is to adaptively select a suitable CSI CR for any CSI input, which can have various sizes corresponding to different configurations in the number of subbands; the other is how to construct a scheme that can flexibly provide general CSI compression and reconstruction functions. In view of the above, embodiments of the present disclosure provide a solution for feeding back CSI, in particular, an AI-based CSI feedback model adaptation is provided, which has a variable compression ratio and is suitable for scalable CSI size.

[0062] The principles and embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. First, refer to Figure 1 which illustrates an example communication system 100 in which embodiments of the present disclosure can be implemented. The system 100 includes terminal devices 110 and network devices 120. The network devices 120 provide services for corresponding areas (also referred to as cells) using different frequency bands in both DL and UL. In the communication system, UL refers to a link in the direction from the terminal devices 110 to the network devices 120, while DL refers to a link in the direction from the network devices 120 to the terminal devices 110. The number of network devices 120 and terminal devices 110 is for illustrative purposes only and does not represent any limitation. The system 100 can include any suitable number of network devices 120 and terminal devices 110 suitable for implementing embodiments of the present disclosure. In addition to communicating with the terminal devices 110, the network devices 120 can also communicate with each other, for example, via a backhaul link. In some embodiments, CSI data can be transmitted from the terminal devices 110 to the network devices 120. For example, in some embodiments, at the terminal device 110 side, the CSI data can be encoded and compressed, while at the network device 120 side, the CSI data encoded and compressed by the terminal device 110 can be decoded and decompressed.

[0063] Communications in the communication system 100 can be implemented according to any suitable communication protocol(s), including but not limited to first generation (1G), second generation (2G), third generation (3G), fourth generation (4G), fifth generation (5G), and wireless local area network communication protocols such as Institute of Electrical and Electronics Engineers (IEEE) 802.11, and / or any other protocol that is currently known or that may be developed in the future. Moreover, communications can utilize any suitable wireless communication techniques, including but not limited to code division multiple access (CDMA), frequency division multiple access (FDMA), time division multiple access (TDMA), frequency division duplexing (FDD), time division duplexing (TDD), multiple-input multiple-output (MIMO), orthogonal frequency division multiplexing (OFDM), discrete Fourier transform spread OFDM (DFT-s-OFDM), and / or any other techniques that are currently known or that may be developed in the future.

[0064] Reference is now made to Figure 2 which shows a communication procedure 200 between a terminal device 110 and a network device 120 according to some embodiments of the present disclosure. For the purpose of discussion, reference will be made to Figure 1 The procedure 200 will be described. The procedure 200 can involve the terminal device 110 and the network device 120 as shown in Figure 1 It should be understood that although the procedure 200 is described in the communication system 100 of Figure 1 , the procedure can equally be applied to other communication scenarios.

[0065] In the procedure 200, at the terminal device 110 side, the terminal device 110 can compress (210) channel state information (CSI) data to obtain a codeword 205 having a codeword length corresponding to a compression ratio option 215 among a plurality of compression ratio options, where the plurality of compression ratio options can correspond to a plurality of codeword lengths related to codeword bits having different priorities. The terminal device 110 can send (220) the codeword 205 and the compression ratio option 215 to the network device 120. At the network device 120 side, the network device 120 can receive (230) the codeword 205 and the compression ratio option 215 among the plurality of compression ratio options from the terminal device 110. The network device 120 can decompress (240) the codeword 205 based on the compression ratio option 215 to obtain the channel state information (CSI) data.

[0066] In this way, the terminal device can adaptively generate CSI codewords of various lengths according to the desired QoS requirement without any additional signaling to indicate the CSI configuration. According to the solution of embodiments, adaptive CSI compression and decompression under various bandwidth configurations can be implemented.

[0067] In some embodiments, to compress the CSI data, the terminal device 110 can perform a pre-processing operation on the CSI data to obtain a patch set of the CSI data, and can also generate a quantized bit set by inputting the patch set into an encoder, and then the terminal device 110 can generate a codeword 205 by inputting the quantized bit set into at least one subunit, wherein different numbers of the at least one subunit correspond to a plurality of compression ratio options.

[0068] In some embodiments, the subunit can be a sub-neural network (sub-NN). In some embodiments, the encoder can be a transformer-enabled encoder. In some other embodiments, the encoder can be a CNN (convolutional neural network), or an RNN (recurrent neural network), or other neural network capable of processing a scalable number of CSI patches, for example, a neural network taking scalable CSI patches as input.

[0069] As mentioned above, different numbers of the at least one subunit can correspond to a plurality of compression ratio options. According to some embodiments of the present disclosure, CSI compression ratio option selection can be supported, for example, different CSI compression ratios can be selected by selecting different numbers of subunits.

[0070] In some embodiments, the terminal device 110 can select at least one subunit among a plurality of subunits based on an association between the compression ratio options and the number of the at least one subunit.

[0071] Taking subunit as sub-NN as an example, Figure 3 A diagram supporting CSI compression ratio option selection according to some embodiments of the present disclosure is shown in FIG. 1, for example, there are three sub-NNs, namely subNN-A, subNN-B and subNN-C, at the transmitter side (for example, the terminal device 110 side, which can send the codeword), and correspondingly, there are also three sub-NNs, namely subNN-AA, subNN-BB and subNN-CC, at the receiver side (for example, the network device 120 side, which can receive the codeword). As shown in FIG. 1, ‘a / b / c’ are the corresponding codeword outputs of SubNN-A / B / C respectively. Figure 3 As shown in FIG. 1, ‘a / b / c’ are the corresponding codeword outputs of SubNN-A / B / C respectively. In this example, a, a+b and a+b+c represent three codeword lengths respectively, and correspond to three compression ratios respectively. Each codeword length corresponds to a compression ratio option. Therefore, different CSI compression ratio options can be selected by selecting different numbers of subunits, and at least one subunit can be selected among a plurality of subunits based on an association between the compression ratio options and the number of the at least one subunit.

[0072] As mentioned above, a plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits with different priorities, therefore, the codeword length is related to the codeword with priority. Referring to FIG. 1, Figure 2In some embodiments, the codeword length of the codeword 205 is related to at least one codeword bit having a highest priority among the plurality of priorities. In some other embodiments, the codeword length of the codeword 205 is related to a plurality of codeword bits corresponding to a priority and at least one priority higher than the priority.

[0073] With continued reference to Figure 3 , ‘a / b / c’ is the corresponding codeword output of SubNN-A / B / C, and a, b and c represent codeword bits having different priorities, specifically, a corresponds to the highest priority, b corresponds to a lower priority, and c corresponds to the lowest priority. In this example, a, a+b and a+b+c represent three codeword lengths, respectively. Based on this example, a corresponds to the highest priority among the plurality of priorities, a+b corresponds to the priority (i.e., the lower priority described above), and at least one priority higher than the priority (i.e., the highest priority described above), and a+b+c corresponds to the priority (i.e., the lowest priority described above) and at least one priority higher than the priority (i.e., the lower priority described above and the highest priority described above).

[0074] In some embodiments, a first codeword length among the plurality of codeword lengths is related to a set of most significant bits, a second codeword length among the plurality of codeword lengths is related to the set of most significant bits plus a set of less significant bits, and a third codeword length among the plurality of codeword lengths is related to the set of most significant bits plus the set of less significant bits plus a set of least significant bits. For example, as shown in Figure 3 , a corresponds to the most significant bits, b corresponds to the set of less significant bits, and c corresponds to the set of least significant bits. Thus, the first codeword length is related to a, the second codeword length is related to a+b, and the third codeword length is related to a+b+c.

[0075] In some embodiments, in a process of performing a pre-processing operation, the terminal device 110 can split CSI data into a set of original patches of the CSI data, where the CSI data is represented by a CSI matrix having a first dimension and a second dimension. The terminal device 110 can also concatenate a plurality of original patches in the set of original patches in a circular shift window in the first dimension to obtain a set of patches of the CSI data, where the circular shift window is shifted across the set of original patches.

[0076] In some embodiments, the first dimension can represent a first configuration of the CSI data, and the second dimension can represent a second configuration of the CSI data. In some embodiments, the first configuration or the second configuration can be a configuration among a plurality of configurations of the CSI data, and the plurality of configurations can include a number of subbands, a number of antenna ports, and a rank, etc.

[0077] In some embodiments, the first dimension can represent the number of subbands, and the second dimension represents the number of antenna ports.

[0078] Reference is made to Figure 4 and Figure 5 , the pre-processing operation can be further described. Figure 4 and Figure 5 collectively illustrate a schematic diagram of the pre-processing operation of CSI input according to some embodiments of the present disclosure. As Figure 4 and Figure 5 shown, the pre-processing operation of scalable CSI input can be summarized as two main steps: 1) CSI data splitting, 2) CSI patch data augmentation. In this example, the encoder can be a transformer-enabled encoder.

[0079] For step 1, first reference is made to Figure 4 , considering the scalability feature of the transformer backbone, the CSI data should be split into patches, which are then fed into the transformer network, which is similar to the operation in the visual transformer (ViT) in the field of computer vision, specifically, the patches are input into the transformer-enabled encoder. Therefore, in the first step of the pre-processing procedure of the present disclosure, the CSI data under various bandwidth configurations (e.g., 12 subband and 13 subband CSI data in Figure 4 ) are split into multiple patches (e.g., Figure 4 1 patch and N 2 patches in N ).

[0080] For detailed information on splitting the feature vector-based CSI data into patches, reference is made to Figure 6 . As Figure 6 shown, Figure 6 illustrates how to split the CSI data. It should be noted that considering the maximum number of words in a sentence in NLP (natural language processing) is 512, when splitting the feature vector-based CSI data into patches, the splitting should satisfy , where 2 represents the real and imaginary parts (not shown in the figure) of the CSI data, N sub is the number of subbands, N Tx is the number of Tx ports, ( p 1, p 2) represents the CSI patch size.

[0081] For step 2, reference is made to Figure 5Unlike the conventional operation in transformer networks, in some embodiments of the present disclosure, the following operations can be performed on the CSI patch data augmentation, because if the split CSI data is directly fed into the transformer-enabled autoencoder platform, the CSI feedback reconstruction accuracy will be far from meeting the loss of relevant information in the CSI data. Here, as shown in Figure 5 each CSI patch can be augmented by concatenating the neighboring subbands in a circular shift window N of the patch, thereby amplifying the CSI patch size to (N∙ p 1, p 2). The circular shift window is shifted in all CSI patches.

[0082] In some embodiments, the terminal device 110 can concatenate at least one original patch in the bottom of the CSI matrix in the first dimension with at least one original patch on the top of the CSI-RS matrix in the first dimension. For example, for the CSI patch in the bottom of the CSI map, it will be concatenated with the patch in the upper part of the CSI map.

[0083] It should be noted that the design of the circular shift window can be adjusted according to the actual subband arrangement. After that, each CSI patch data is augmented by including more information from neighboring subbands, which is beneficial for CSI compression. After the above steps 1 and 2 in the preprocessing module, the original CSI data is split into enhanced patches, and the size of each patch is (N∙ p 1, p 2). Then, the CSI patch is fed into the transformer-enabled general encoder and decoder of the present disclosure.

[0084] The CSI preprocessing process can also be referred to Figure 7 . Figure 7 A schematic diagram of CSI preprocessing according to some embodiments of the present disclosure is illustrated. As shown in Figure 7 in some embodiments, the individual input CSI matrix is converted into a patch with a length of N1 (or N2), and since the sequence processing capability of the transformer, any Nx < Lmax (where Lmax is predefined by the encoder) can be supported. Therefore, the benefit is that the CSI input with different data sizes (e.g., the number of subbands or the number of antenna ports, rank) can be adapted by the general DL model (rather than a series of models corresponding to different CSI input sizes) without difference.

[0085] In some embodiments, after the network device 120 performs beamforming based on the reconstructed CSI data, the terminal device 110 can evaluate the quality of service (QoS) performance of the beamforming, wherein the reconstructed CSI data is reconstructed by the network device 120 based on the codebook 205 and the compression ratio option 215.

[0086] In some embodiments, the compression ratio option 215 is a first compression ratio option, the codeword 205 is a first codeword, the codeword length is a first codeword length, and the CSI data is first CSI data. The terminal device 110 can: based on determining that the QoS performance is lower than the threshold, select a second compression ratio option in the plurality of compression ratio options, the second compression ratio option corresponding to a second codeword length that is longer than the first codeword length. The terminal device 110 can further: compress second CSI data to obtain a second codeword having the second codeword length corresponding to the second compression ratio option, and send the second codeword and the second compression ratio option to the network device 120.

[0087] Reference Figure 3 And Table 1, for example, the first compression ratio option is 1, and the first codeword length corresponds to ‘a’. If the terminal device 110 determines that the QoS performance is lower than the threshold, the terminal device 110 can select a second compression ratio option 2, and the second codeword length is ‘a+b’, which is longer than the first codeword length.

[0088] Table 1

[0089] In some embodiments, based on determining that the QoS performance is lower than the threshold, and the first compression ratio option is the largest in the plurality of compression ratio options, the terminal device 110 can skip selecting a second compression ratio option in the plurality of compression ratio options.

[0090] In some embodiments, the first codeword having the first codeword length is obtained via a first number of subunits, and the second codeword having the second codeword length is obtained via a second number of subunits that is more than the first number of subunits.

[0091] On the network device 120 side, in some embodiments, the network device 120 can decompress the CSI data, and to decompress the CSI data, the network device 20 can generate a set of quantization bits by inputting the codeword 205 into at least one subunit, wherein different numbers of the at least one subunit correspond to the plurality of compression ratio options; and generate the CSI data by inputting the set of quantization bits into a decoder.

[0092] In some embodiments, the network device 120 can select the at least one subunit in the plurality of subunits based on an association between the compression ratio option and the number of the at least one subunit. Reference Figure 3, Table 2 and the related content introduced above, on the network device 120 side, for example, there can be three sub-NNs, namely sub-NN-AA, sub-NN-BB and sub-NN-CC, and a, a+b and a+b+c represent three code word lengths respectively, and correspond to three compression ratios respectively. If the network device 120 receives a compression ratio option, for example, the compression ratio option 1 in Table 2, the network device 120 can select the sub-NN AA. For each sub-NN at the terminal device 110, there is a corresponding sub-NN at the network device 120, for example, the sub-NN AA at the network device 120 corresponds to the sub-NN A at the terminal device 110.

[0093] Table 2

[0094] As described above, on the terminal device 110 side, based on determining that the QoS performance is lower than the threshold value, the terminal can select a second compression ratio option, and compress the second CSI data to obtain a second code word with a second code word length corresponding to the second compression ratio option, and then send the second code word and the second compression ratio option to the network device 120. Therefore, on the network device 120 side, the network device 120 can receive the second code word and the second compression ratio option from the terminal device 110, and decompress the second code word based on the second compression ratio to obtain the second CSI data.

[0095] In some embodiments, the first CSI data and the second CSI data can be the same or different.

[0096] In some embodiments, the first code word with the first code word length can be input into a first number of sub-units, and the second code word with the second code word length can be input into a second number of sub-units more than the first number of sub-units.

[0097] In some embodiments, corresponding to the sub-NN on the terminal device 110 side, on the network device 120 side, the sub-unit can be a sub-neural network (sub-NN). The decoder corresponds to the encoder, and in some embodiments, the decoder can be a transformer-enabled decoder. In some other embodiments, the decoder can be a CNN (convolutional neural network), or an RNN (recurrent neural network), or other neural networks capable of processing a scalable number of CSI patches, for example, by decoding the decompressed code word to output a scalable number of CSI patches.

[0098] Based on the above description, the embodiments described below will further introduce some other aspects of the scheme of the present disclosure.

[0099] Figure 8A schematic diagram illustrating a process architecture according to some embodiments of the present disclosure is shown. In some embodiments, to enhance the generalization of AI / ML models in CSI feedback enhancement applications, an exemplary process architecture is shown as Figure 8 The AI / ML based general-purpose CSI feedback scheme of the present disclosure can be embodied by the following: a pre-processing module of scalable CSI input; a transformer-enabled general-purpose encoder and decoder; sub-NNs for flexible codeword length.

[0100] Referring to the first part (1) of Figure 8 , the transmitter side performs CSI pre-processing before compression. For example, in some embodiments, the transmitter side can use a cyclic shift patching method for CSI pre-processing before compression (encoder).

[0101] Referring to the second part (2) of Figure 8 , the pre-processed CSI is indiscriminately adaptively compressed into codewords with required codeword length using a general-purpose DL model (e.g., preferred transformer) for feedback and over-the-air transmission. In some embodiments, the CSI compression function and corresponding codeword are pluggable, which consists of bits with different priorities, i.e., sending codewords with the highest significant bits (‘a’ in Figure 8 to save bandwidth, while sending lower significant bits (‘b’ or ‘c’ in Figure 8 ) is only to get better reconstruction performance as needed. At the receiver side, correspondingly, the codewords (which can have different lengths) are decompressed back to original CSI data with corresponding CSI size using a general-purpose DL model (preferred transformer).

[0102] Referring to the third part (3) of Figure 8 , according to the requirements of the receiver / gNB side (e.g., performance on beamforming based on CSI feedback), the controller can adaptively select how many SubNN-x (NN, i.e., neural network) to activate and the corresponding codeword a / b / c / … for feedback.

[0103] As mentioned above Figure 3As shown, the general encoder performs general feature extraction (recommended to use a transformer to accommodate CSI input with different matrix size), followed by several parallel SubNN-A / B / Cs to perform dimension reduction. 'a / b / c' are the corresponding codeword outputs of SubNN-A / B / C, respectively, where 'a' denotes the most significant bits, and 'b / c' denote the lower / least significant bits. For bandwidth saving mode operation, only 'a' is transmitted over the air for feedback, and 'b' and 'c' are added to the codeword of 'a+b' or 'a+b+c' as needed. Note that the number of a / b / c is not limited to 3, and the length of a / b / c can be any integer, as shown herein. SubNN-A / B / C can be implemented by FCNN, CNN, or other options, which have very compact parameter size compared to the preceding general encoder.

[0104] Figure 9 Fig. 1 illustrates a schematic diagram of a general transformer-enabled CSI feedback scheme, according to some embodiments of the present disclosure. As shown, for example, a general transformer-enabled CSI feedback scheme is proposed to accommodate CSI data with scalable subband configuration. The proposed general scheme can be generalized to accommodate CSI input in the form of feature vectors with different data size (e.g., number of subbands, number of antenna ports, rank). Figure 9

[0105] In some embodiments, as shown, the input and output of the AI / ML-based CSI feedback scheme are precoding matrices of various sizes (feature vector-formatted CSI data), due to different choices of number of subbands, number of antenna ports, rank. In this case, the CSI data has the same 32 Tx port number, but under different subband configurations (i.e., 12 subbands and 13 subbands). Note that the real and imaginary dimensions of the CSI data are not depicted in this figure. Figure 9

[0106] Figure 10 Fig. 2 illustrates a schematic diagram of a transformer-enabled general encoder design, according to some embodiments of the present disclosure. As shown, after pre-processing, the CSI data of different subband configurations are split into different numbers of augmented patches, and fed into the transformer-enabled general encoder. Figure 10

[0107] Figure 11 Fig. 3 illustrates a schematic diagram of the internal architecture of the transformer-enabled general encoder, according to some embodiments of the present disclosure. The internal architecture of the transformer-enabled general encoder is as shown. In the transformer-enabled general encoder, the CSI patches are first processed by a transformer with embedding dimension of Figure 11 E dim ​​​​linear embedding layer processing and position encoding by adding a position vector. Then, the CSI data is processed by 6 basic blocks containing multi-head attention modules and fully connected (FC) blocks to extract features. Subsequently, the CSI data is processed by a compression module. Finally, the compressed CSI is quantized into bits and transmitted to the gNB.

[0108] Figure 12 Fig. illustrates a schematic diagram of the internal architecture of a transformer-enabled general decoder according to some embodiments of the present disclosure. For the general decoder, the feedback bitstream of CSI data is first dequantized and decompressed into a matrix of size N p , E dim . Then, the matrix is processed symmetrically as in the encoder to recover the CSI data.

[0109] In addition to the general encoder and decoder, sub-NNs for flexible feedback codeword length can be implemented in some schemes to prioritize codewords in model training, e.g., through serial / parallel multi-rate compression framework or extraction methods. Then, codewords of different bit numbers can be transmitted over the air in a pluggable manner (e.g., most / less / most significant bits). The relationship between the codeword length options, codeword length N bits and the selected sub-NNs are defined in Table 1 or Table 2 above.

[0110] Figure 13 Fig. illustrates a flowchart of a communication procedure 1300 between a terminal device 110 and a network device 120 according to some other embodiments of the present disclosure. For example, the terminal device 110 is a UE and the network device 120 is a gNB, and the procedure 1300 will be described based on this example. As shown in Figure 13 The procedure of UE with scalable CSI size adaptively selecting codeword length to meet QoS requirement is shown in the figure. It is assumed that there are M compression ratio options.

[0111] At 1302, the UE first selects an appropriate compression ratio option according to the QoS requirement. At 1304, the UE compresses CSI data of various sizes using the general encoder and obtains N bits codewords. At 1306, the UE sends the N bits codewords and the compression ratio option to the gNB. At 1308, the general decoder at the gNB decompresses the codewords and obtains reconstructed CSI of various sizes according to the compression ratio option.

[0112] At 1310, the gNB performs beamforming (BF) according to the reconstructed CSI. At 1312, the UE evaluates the QoS performance (throughput / BLER / ...) and if a certain performance cannot reach the threshold, the UE downgrades its compression ratio option to a lower level and then repeats 1304, 1306, 1308, 1310 and 1312 to reach the QoS threshold. At 1314, if the UE selects the Mth(maximum, e.g., 1 / 4) compression ratio option and the QoS requirement is still not satisfied, the UE will stop repeating 1304, 1306, 1308, 1310 and 1312 to reach the QoS threshold.

[0113] The proposed universal transformer enabled CSI feedback scheme for CSI data with different data sizes in this disclosure can be verified by simulation. The simulation dataset is generated according to 3GPP TR 38.901. The dataset configuration is shown in Table 3.

[0114] Table 3

[0115] Each sample in the above-mentioned CDLC300-48RB dataset and CDLC300-52RB dataset is a real number vector with size .

[0116] The CDLC300-48RB dataset and CDLC300-52RB dataset are tested on the proposed universal transformer enabled CSI feedback scheme (i.e., the scheme in this disclosure, or simply referred to as the proposed scheme) to verify the feasibility of the proposed scheme under various subband configurations. The baseline is tested with the traditional one-to-one transformer scheme (i.e., the traditional scheme). The CR is equal to 1 / 64 and the quantization bits are equal to 4, which means the feedback codeword is 48 bits for CDLC300 48RB and 52 bits for CDLC300 52RB.

[0117] The simulation results are shown in Figure 14 , specifically, Figure 14 Fig. 1 illustrates the CSI reconstruction accuracy (in terms of generalized cosine similarity, GCS) of the CSI dataset under various bandwidth configurations. The simulation results show that the proposed universal transformer scheme can achieve similar results as the traditional one-to-one transformer scheme. The simulation results verify that the proposed universal transformer enabled CSI feedback scheme can handle CSI under various bandwidth configurations without affecting the CSI reconstruction accuracy of the one-to-one scheme.

[0118] Then, the CDLC300-48RB dataset is tested, using the baseline scheme and the proposed flexible CR scheme to generate 48-bit codewords (CR = 1 / 64) and 96-bit codewords (CR = 1 / 32), respectively. The model in the baseline scheme generates feedback codewords with a fixed codeword length. The simulation results are shown in Figure 15 The simulation results show that the proposed flexible CR scheme can achieve similar results as the baseline scheme. The simulation results verify that the proposed scheme can generate CSI feedback codeword bits with priority while not affecting the accuracy of CSI reconstruction.

[0119] Figure 16 FIG. illustrates a flow chart of a method 1600 implemented at a terminal device 110, according to some embodiments of the present disclosure. For the purpose of discussion, the method 1600 will be described with reference to the terminal device 110, the network device 120, and the CSI feedback scheme 100 of FIGS. 1 and 2. Figure 1 The method 1600 is described from the perspective of the terminal device 110.

[0120] At block 1610, the terminal device 110 can compress channel state information (CSI) data to obtain a codeword with a codeword length corresponding to a compression ratio option of a plurality of compression ratio options, where the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits with different priorities. At block 1620, the terminal device 110 can send the codeword and the compression ratio option to the network device 120.

[0121] In some embodiments, to compress the CSI data, the terminal device 110 can: perform a pre-processing operation on the CSI data to obtain a patch set of the CSI data; generate a quantized bit set by inputting the patch set into an encoder; and generate the codeword by inputting the quantized bit set into at least one subunit, where different numbers of the at least one subunit correspond to the plurality of compression ratio options.

[0122] In some embodiments, the codeword length is related to at least one codeword bit with a highest priority of the plurality of priorities. In some other embodiments, the codeword length is related to a plurality of codeword bits corresponding to a priority and at least one priority higher than the priority.

[0123] In some embodiments, a first codeword length of the plurality of codeword lengths can be related to a most significant bit set, a second codeword length of the plurality of codeword lengths can be related to the most significant bit set plus a lower significant bit set, and a third codeword length of the plurality of codeword lengths can be related to the most significant bit set plus the lower significant bit set plus a least significant bit set.

[0124] In some embodiments, the terminal device 110 can select the at least one subunit among the plurality of subunits based on an association between the compression ratio option and the number of the at least one subunit.

[0125] In some embodiments, the terminal device 110 can perform a pre-processing operation, specifically, the terminal device 100 can: split the CSI data into a set of original patches of the CSI data, wherein the CSI data is represented by a CSI matrix having a first dimension and a second dimension; and concatenate a plurality of original patches in the set of original patches in the first dimension in a circular shift window to obtain a set of patches of the CSI data, wherein the circular shift window is shifted throughout the set of original patches.

[0126] In some embodiments, the terminal device 110 can further perform the pre-processing operation by concatenating at least one original patch in a bottom of the CSI matrix in the first dimension with at least one original patch on a top of the CSI-RS matrix in the first dimension.

[0127] In some embodiments, the first dimension can represent a first configuration of the CSI data, and the second dimension can represent a second configuration of the CSI data.

[0128] In some embodiments, the first dimension can represent a number of subbands, and the second dimension represents a number of antenna ports.

[0129] In some embodiments, after the network device 120 performs the beamforming based on the reconstructed CSI data, the terminal device 110 can evaluate a quality of service (QoS) performance of the beamforming, wherein the reconstructed CSI data is reconstructed by the network device 120 based on the codeword and the compression ratio option.

[0130] In some embodiments, the compression ratio option is a first compression ratio option, the codeword is a first codeword, the codeword length is a first codeword length, the CSI data is a first CSI data, and the terminal device 110 can: based on determining that the QoS performance is lower than a threshold, select a second compression ratio option among the plurality of compression ratio options, the second compression ratio option corresponding to a second codeword length that is longer than the first codeword length; compress a second CSI data to obtain a second codeword having the second codeword length corresponding to the second compression ratio option; and then send the second codeword and the second compression ratio option to the network device 120.

[0131] In some embodiments, the first codeword having the first codeword length can be obtained via a first number of subunits, and the second codeword having the second codeword length can be obtained via a second number of subunits that is more than the first number of subunits.

[0132] In some embodiments, the compression ratio option is a first compression ratio option, and based on determining that the QoS performance is lower than a threshold, and the first compression ratio option is the largest among the plurality of compression ratio options, the terminal device 110 can skip selecting a second compression ratio option among the plurality of compression ratio options.

[0133] In some embodiments, the encoder can be a transformer-enabled encoder. In some embodiments, the subunit can be a sub-neural network (sub-NN).

[0134] Figure 17 A flowchart of a method 1700 implemented at the network device 120 according to some embodiments of the present disclosure is illustrated. For the purpose of discussion, reference will be made to Figure 1 The method 1700 is described from the perspective of the network device 120.

[0135] At block 1710, the network device 120 can receive, from the terminal device 110, a codeword and a compression ratio option of a plurality of compression ratio options, where the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, and where the codeword has a codeword length corresponding to the compression ratio option. At block 1720, the network device 120 can decompress the codeword based on the compression ratio option to obtain channel state information (CSI) data.

[0136] In some embodiments, to decompress the CSI data, the network device 120 can: generate a set of quantized bits by inputting the codeword into at least one subunit, where different numbers of the at least one subunit correspond to the plurality of compression ratio options; and generate the CSI data by inputting the set of quantized bits into a decoder.

[0137] In some embodiments, the codeword length can be related to at least one codeword bit having a highest priority of the plurality of priorities. In some embodiments, the codeword length can be related to a plurality of codeword bits corresponding to the priority and at least one priority higher than the priority.

[0138] In some embodiments, a first codeword length of the plurality of codeword lengths can be related to a set of most significant bits. A second codeword length of the plurality of codeword lengths can be related to the set of most significant bits plus a set of less significant bits. A third codeword length of the plurality of codeword lengths can be related to the set of most significant bits plus the set of less significant bits plus a set of least significant bits.

[0139] In some embodiments, the network device 120 can select the at least one subunit among the plurality of subunits based on an association between the compression ratio option and the number of the at least one subunit.

[0140] In some embodiments, the CSI data can be represented by a CSI matrix having a first dimension and a second dimension. The first dimension can represent a first configuration of the CSI data. The second dimension can represent a second configuration of the CSI data.

[0141] In some embodiments, the first dimension can represent a number of subbands. The second dimension can represent a number of antenna ports.

[0142] In some embodiments, the code word is a first code word, the compression ratio option is a first compression ratio option, and the CSI data is first CSI data. The network device 120 can: receive, from the terminal device 110, a second code word and a second compression ratio option of the plurality of compression ratio options; and decompress the second code word based on the second compression ratio option to obtain second CSI data.

[0143] In some embodiments, a first code word having a first code word length can be input into a first number of sub-units. A second code word having a second code word length can be input into a second number of sub-units that is more than the first number of sub-units.

[0144] In some embodiments, the decoder can be a transformer-enabled decoder. In some embodiments, the sub-unit can be a sub-neural network (sub-NN).

[0145] In some embodiments, an apparatus (e.g., the terminal device 110) capable of performing any one of the methods 1600 can include means for performing the corresponding steps of the methods 1600. This apparatus can be realized in any suitable form. For example, this apparatus can be realized in a circuitry or a software module.

[0146] In some embodiments, the apparatus includes: means for compressing channel state information (CSI) data to obtain a code word having a code word length corresponding to a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of code word lengths related to code word bits having different priorities; and means for sending the code word and the compression ratio option to the network device 120.

[0147] In some embodiments, the means for compressing the CSI data includes: means for performing a pre-processing operation on the CSI data to obtain a patch set of the CSI data; means for generating a quantization bit set by inputting the patch set into an encoder; and means for generating the code word by inputting the quantization bit set into at least one sub-unit, wherein different numbers of the at least one sub-unit correspond to the plurality of compression ratio options.

[0148] In some embodiments, the code word length is related to at least one code word bit having a highest priority of a plurality of priorities, or to a plurality of code word bits corresponding to a priority and at least one priority higher than the priority.

[0149] In some embodiments, a first codeword length of the plurality of codeword lengths is related to the most significant bit set, a second codeword length of the plurality of codeword lengths is related to the most significant bit set plus the less significant bit set, and a third codeword length of the plurality of codeword lengths is related to the most significant bit set plus the less significant bit set plus the least significant bit set.

[0150] In some embodiments, the apparatus further includes means for selecting the at least one subunit in the plurality of subunits based on an association between the compression ratio option and the number of at least one subunit.

[0151] In some embodiments, the means for performing the pre-processing operation includes means for splitting the CSI data into a set of original patches of the CSI data, where the CSI data is represented by a CSI matrix having a first dimension and a second dimension, and means for concatenating, in the first dimension, a plurality of original patches in the set of original patches in a circular shift window to obtain a set of patches of the CSI data, where the circular shift window is shifted across the set of original patches.

[0152] In some embodiments, the means for performing the pre-processing operation includes means for concatenating at least one original patch in a bottom of the CSI matrix in the first dimension with at least one original patch on a top of the CSI-RS matrix in the first dimension.

[0153] In some embodiments, the first dimension represents a first configuration of the CSI data, and the second dimension represents a second configuration of the CSI data.

[0154] In some embodiments, the first dimension represents a number of subbands, and the second dimension represents a number of antenna ports.

[0155] In some embodiments, the apparatus further includes means for evaluating a quality of service (QoS) performance of the beamforming after the network device 120 performs the beamforming based on the reconstructed CSI data, where the reconstructed CSI data is reconstructed by the network device 120 based on the codeword and the compression ratio option.

[0156] In some embodiments, the compression ratio option is a first compression ratio option, the codeword is a first codeword, the codeword length is a first codeword length, the CSI data is first CSI data, and the apparatus further includes means for selecting, based on determining that the QoS performance is below a threshold, a second compression ratio option corresponding to a second codeword length that is longer than the first codeword length in a plurality of compression ratio options; means for compressing second CSI data to obtain a second codeword having the second codeword length corresponding to the second compression ratio option; and means for sending the second codeword and the second compression ratio option to the network device 120.

[0157] In some embodiments, a first codeword having a first codeword length is obtained via a first number of subunits, and a second codeword having a second codeword length is obtained via a second number of subunits that is more than the first number of subunits.

[0158] In some embodiments, the compression ratio option is a first compression ratio option, and the apparatus further includes means for skipping selecting a second compression ratio option among the plurality of compression ratio options based on determining that the QoS performance is below the threshold and the first compression ratio option is the largest among the plurality of compression ratio options.

[0159] In some embodiments, the encoder is a transformer-enabled encoder. In some embodiments, the subunit is a sub-neural network (sub-NN). In some embodiments, the encoder is a transformer-enabled encoder and the subunit is a sub-neural network (sub-NN).

[0160] In some embodiments, the apparatus further includes means for performing other steps of some embodiments of the method 1600. In some embodiments, the means includes at least one processor and at least one memory including computer program code, the at least one memory and the computer program code configured to, with the at least one processor, cause the performance of the apparatus.

[0161] In some embodiments, an apparatus (e.g., network device 120) capable of performing any of the methods 1700 can include means for performing the respective steps of the method 1700. The means can be implemented in any suitable form. For example, the means can be implemented in circuitry or software modules.

[0162] In some embodiments, the apparatus includes means for receiving, from the terminal device 110, a codeword and a compression ratio option among a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and means for decompressing the codeword based on the compression ratio option to obtain channel state information (CSI) data.

[0163] In some embodiments, the means for decompressing the CSI data includes means for generating a set of quantization bits by inputting the codeword into at least one subunit, wherein different numbers of the at least one subunit correspond to the plurality of compression ratio options; and means for generating the CSI data by inputting the set of quantization bits into a decoder.

[0164] In some embodiments, the codeword length is related to at least one codeword bit having a highest priority among the plurality of priorities, or to a plurality of codeword bits corresponding to a priority and at least one priority higher than the priority.

[0165] In some embodiments, a first codeword length of the plurality of codeword lengths is related to the set of most significant bits, a second codeword length of the plurality of codeword lengths is related to the set of most significant bits plus the set of less significant bits, and a third codeword length of the plurality of codeword lengths is related to the set of most significant bits plus the set of less significant bits plus the set of least significant bits.

[0166] In some embodiments, the apparatus further comprises means for selecting the at least one subunit in the plurality of subunits based on an association between the compression ratio option and the number of at least one subunit.

[0167] In some embodiments, the CSI data is represented by a CSI matrix having a first dimension and a second dimension, wherein the first dimension represents a first configuration of the CSI data and the second dimension represents a second configuration of the CSI data.

[0168] In some embodiments, the first dimension represents a number of subbands and the second dimension represents a number of antenna ports.

[0169] In some embodiments, the codeword is a first codeword, the compression ratio option is a first compression ratio option, the CSI data is first CSI data, and the apparatus further comprises means for receiving a second codeword and a second compression ratio option of the plurality of compression ratio options from the terminal device 110; and means for decompressing the second codeword based on the second compression ratio option to obtain second CSI data.

[0170] In some embodiments, a first codeword having a first codeword length is input into a first number of subunits; and a second codeword having a second codeword length is input into a second number of subunits that is more than the first number of subunits.

[0171] In some embodiments, the decoder is a transformer-enabled decoder, or the subunit is a sub-neural network (sub-NN), or the decoder is a transformer-enabled decoder and the subunit is a sub-neural network (sub-NN).

[0172] In some embodiments, the apparatus further comprises means for performing other steps of some embodiments of the method 1700. 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 the computer program code configured to, with the at least one processor, cause the performance of the apparatus.

[0173] Figure 18 is a simplified block diagram of a device 1800 suitable for implementing embodiments of the present disclosure. The device 1800 can be provided to implement a communication device, for example Figure 1The terminal device 110, network device 120 are shown. As shown, the device 1800 includes one or more processors 1810, one or more memories 1820 coupled to the processors 1810, and one or more communication modules 1840 coupled to the processors 1810.

[0174] The communication module 1840 is for bidirectional communication. The communication module 1840 has at least one antenna to facilitate communication. The communication interface can represent any interface needed for communication with other network elements.

[0175] The processor 1810 can be of any type suitable to the local technical network and can include, by way of non-limiting example, one or more of general purpose computers, special purpose computers, microprocessors, digital signal processors (DSPs) and processors based on multi-core processor architectures, as non-limiting examples. The device 1800 can have multiple processors such as a dedicated integrated circuit chip that is time-slaved to a clock that is synchronized with a master processor.

[0176] The memory 1820 can include one or more non-transitory memories and one or more transitory memories. Examples of non-transitory memories include, but are not limited to, read-only memories (ROMs) 1824, electrically programmable read-only memories (EPROMs), flash memories, hard disks, compact disks (CDs), digital video disks (DVDs), and other magnetic and / or optical storage devices. Examples of transitory memories include, but are not limited to, random access memories (RAMs) 1822 and other volatile memories that do not persist at power cycles.

[0177] The computer program 1830 includes computer executable instructions that are executed by the associated processor 1810. The program 1830 can be stored in the ROM 1824. The processor 1810 can perform any suitable action and processing by loading the program 1830 into the RAM 1822.

[0178] Embodiments of the present disclosure can be implemented by the program 1830 so that the device 1800 can perform any process of the present disclosure discussed Figures 1 to 17 Embodiments of the present disclosure can also be implemented by hardware or by a combination of software and hardware.

[0179] In some embodiments, the program 1830 can be tangibly contained on a computer readable medium, which can be included in the device 1800 (such as in the memory 1820) or in another storage device accessible to the device 1800. The device 1800 can load the program 1830 from the computer readable medium into the RAM 1822 for execution. The computer readable medium can include any type of tangible non-transitory memory, such as ROM, EPROM, flash memory, a hard disk, a CD, a DVD, and the like.Figure 19 An example of a computer readable medium 1900 in the form of a CD or DVD is shown. The computer readable medium has stored thereon the program 1830.

[0180] In general, the various embodiments of the disclosure can be implemented using hardware or special-purpose circuits, software, logic or any combination thereof. Some aspects can be implemented using hardware, while other aspects can be implemented using software or firmware that is executed by a controller, microprocessor or other computing device. Although various aspects of the disclosure are illustrated and described as block diagrams, flow charts, or using some other pictorial representation, it is well understood that these blocks, apparatus, systems, techniques or methods described herein can be implemented using hardware, software, firmware, special-purpose circuits or logic, general purpose hardware or controller or other computing devices, or some combination thereof.

[0181] The disclosure also provides at least one computer program product which is 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, executed by devices on a target real or virtual processor to perform the methods described above with respect to Figures 2-17 The program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that implement particular tasks or implement particular abstract data types. In various embodiments, the functionality of the program modules can be combined or split between program modules as desired. Machine executable instructions for a program module can be executed within a local or distributed device. In a distributed device, program modules can be located in both local and remote storage media.

[0182] Program code used to implement the methods of the present disclosure can be written in any combination of one or more programming languages. This program code can 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 code, when executed by the processor or controller, causes the functions / operations specified in the flow charts and / or block diagrams to be implemented. The program code can be entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine or entirely on a remote machine or server.

[0183] In the context of the present disclosure, computer program code or related data can be carried by any suitable carrier and means to enable the device, apparatus or processor to perform the various processes and operations as described above. Examples of carriers include signals, computer readable media, etc.

[0184] The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. It can include, but is not limited to, 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 one or more of 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 to modify the medium does not encompass a transitory signal per se, as the term is of no meaning in quantum mechanics. It is to be understood that the terms "data store," "data storage," "database," and / or the like as used herein can refer to any device or combination of devices capable of storing data. That is, the terms "data store," "data storage," "database," and / or the like can refer to a memory, a collection of memories, a cache, a database, and / or the like.

[0185] Also, although operations are described as being in a particular order, this should not be understood as requiring that particular order or that all illustrated operations be performed in order to achieve desirable results. In certain circumstances, multitasking and parallel processing can be advantageous. Likewise, while a number of specific implementation details have been included for the purpose of providing a thorough description of certain embodiments, it will be understood that these are by way of example only. Certain features that are described in the context of separate embodiments can also be implemented in combination with each other. Conversely, various features that are described in the context of a single embodiment can also be implemented or provided in the context of multiple embodiments. Furthermore, the term "comprising" as used in this description should not be interpreted as being restricted to members of a set or range of values; but rather, should be interpreted as including members of a set or range of values.

[0186] Although the present disclosure has been described in language 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. Rather, the specific features and acts described are disclosed as example forms of implementing the claims.

Claims

1. A terminal device, comprising: 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: compress channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option among a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and send the codeword and the compression ratio option to a network device. 2.The terminal device of claim 1, wherein the terminal device is caused to compress the CSI data by: performing a pre-processing operation on the CSI data to obtain a patch set of the CSI data; generating a set of quantization bits by inputting the patch set into an encoder; and generating the codeword by inputting the set of quantization bits into at least one subunit, wherein different numbers of the at least one subunit correspond to the plurality of compression ratio options. 3.The terminal device of claim 1 or 2, wherein the codeword length is related to one of: at least one codeword bit having a highest priority among a plurality of priorities; or a plurality of codeword bits corresponding to a priority and at least one priority higher than the priority. 4.The terminal device of any one of claims 1 to 3, wherein: a first codeword length among the plurality of codeword lengths is related to a set of most significant bits; a second codeword length among the plurality of codeword lengths is related to the set of most significant bits plus a set of less significant bits; and a third codeword length among the plurality of codeword lengths is related to the set of most significant bits plus the set of less significant bits plus a set of least significant bits. 5.The terminal device of claim 2, wherein the terminal device is further caused to: select the at least one subunit among a plurality of subunits based on an association between the compression ratio option and the number of the at least one subunit. 6.The terminal device of claim 2, wherein the terminal device is caused to perform the pre-processing operation by: splitting the CSI data into an original patch set of the CSI data, wherein the CSI data is represented by a CSI matrix having a first dimension and a second dimension; and concatenating, in the first dimension, a plurality of original patches among the original patch set in a circular shift window to obtain the patch set of the CSI data, wherein the circular shift window is shifted throughout the original patch set. 7.The terminal device of claim 6, wherein the terminal device is further caused to perform the pre-processing operation by: concatenating at least one original patch in a bottom of the CSI matrix in the first dimension with at least one original patch on a top of the CSI-RS matrix in the first dimension. ​ ​ 8. The terminal device of claim 6 or 7, wherein the first dimension represents a first configuration of the CSI data, and the second dimension represents a second configuration of the CSI data.

9. The terminal device of any one of claims 6 to 8, wherein the first dimension represents a number of subbands, and the second dimension represents a number of antenna ports.

10. The terminal device of any one of claims 1 to 9, wherein the terminal device is further caused to: evaluate a quality of service (QoS) performance of a beamforming performed by the network device based on reconstructed CSI data, wherein the reconstructed CSI data is reconstructed by the network device based on the codeword and the compression ratio option.

11. The terminal device of claim 10, wherein the compression ratio option is a first compression ratio option, the codeword is a first codeword, the codeword length is a first codeword length, the CSI data is first CSI data, and wherein the terminal device is further caused to: based on determining that the QoS performance is below a threshold, select a second compression ratio option among the plurality of compression ratio options, the second compression ratio option corresponding to a second codeword length that is longer than the first codeword length; compress second CSI data to obtain a second codeword having the second codeword length corresponding to the second compression ratio option; and send the second codeword and the second compression ratio option to the network device.

12. The terminal device of claim 11, wherein: the first codeword having the first codeword length is obtained via a first number of subunits; and the second codeword having the second codeword length is obtained via a second number of subunits that is more than the first number of subunits.

13. The terminal device of claim 10, wherein the compression ratio option is a first compression ratio option, and wherein the terminal device is further caused to: based on determining that the QoS performance is below a threshold, and the first compression ratio option is the largest among the plurality of compression ratio options, skip selecting a second compression ratio option among the plurality of compression ratio options.

14. The terminal device of claim 2, wherein at least one of: the encoder is a transformer-enabled encoder; or the subunit is a sub-neural network (sub-NN).

15. A network device comprising: 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 codeword, and a compression ratio option among a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and decompress the codeword based on the compression ratio option to obtain channel state information (CSI) data. ​ ​ 16. The network device of claim 15, wherein the network device is caused to decompress the CSI data by: generating a set of quantization bits by inputting the codeword into at least one subunit, wherein different numbers of the at least one subunit correspond to the plurality of compression ratio options; and generating the CSI data by inputting the set of quantization bits into a decoder.

17. The network device of claim 15 or 16, wherein the codeword length is related to one of: at least one codeword bit having a highest priority among a plurality of priorities; or a plurality of codeword bits corresponding to a priority and at least one priority higher than the priority.

18. The network device of any one of claims 15 to 17, wherein: a first codeword length of the plurality of codeword lengths is related to a set of most significant bits; a second codeword length of the plurality of codeword lengths is related to the set of most significant bits plus a set of less significant bits; and a third codeword length of the plurality of codeword lengths is related to the set of most significant bits plus the set of less significant bits plus a set of least significant bits.

19. The network device of claim 16, wherein the network device is further caused to: select the at least one subunit among a plurality of subunits based on an association between the compression ratio options and the number of the at least one subunit.

20. The network device of any one of claims 15 to 19, wherein the CSI data is represented by a CSI matrix having a first dimension and a second dimension, and wherein the first dimension represents a first configuration of the CSI data and the second dimension represents a second configuration of the CSI data.

21. The network device of claim 20, wherein the first dimension represents a number of subbands and the second dimension represents a number of antenna ports.

22. The network device of claim 16, wherein the codeword is a first codeword, the compression ratio option is a first compression ratio option, the CSI data is first CSI data, and the network device is further caused to: receive, from the terminal device, a second codeword and a second compression ratio option of the plurality of compression ratio options; and decompress the second codeword based on the second compression ratio option to obtain second CSI data.

23. The network device of claim 22, wherein: the first codeword having the first codeword length is input into a first number of subunits; and the second codeword having the second codeword length is input into a second number of subunits more than the first number of subunits.

24. The network device of claim 16, wherein at least one of: the decoder is a transformer-enabled decoder; or the subunit is a sub-neural network (sub-NN).

25. A method comprising: ​ At a terminal device, compress channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and send the codeword and the compression ratio option to a network device.

26. A method comprising: At a network device, receive, from a terminal device, a codeword and a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and decompress the codeword based on the compression ratio option to obtain channel state information (CSI) data.

27. An apparatus comprising: means for compressing, at a terminal device, channel state information (CSI) data to obtain a codeword having a codeword length corresponding to a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities; and means for sending the codeword and the compression ratio option to a network device.

28. An apparatus comprising: means for receiving, at a network device, from a terminal device, a codeword and a compression ratio option of a plurality of compression ratio options, wherein the plurality of compression ratio options correspond to a plurality of codeword lengths related to codeword bits having different priorities, wherein the codeword has a codeword length corresponding to the compression ratio option; and means for decompressing the codeword based on the compression ratio option to obtain channel state information (CSI) data.

29. 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 any one of claims 25-26. ​

Images