Method and apparatus for learning-based joint framework for channel state information (CSI) compression and prediction

Abstract

A method and apparatus for a learning-based joint framework for channel state information (CSI) compression and prediction are provided, the method comprising: an encoder of a user equipment (UE) compressing a channel vector of a channel state information (CSI) matrix to generate a corresponding compressed vector, and a processor of the UE generates a predicted channel vector by performing machine learning (ML)-based CSI prediction on the compression vector. The UE reports the predicted CSI to a base station (BS) based on the predicted channel vector.

Description

Technical Field

[0001] The present disclosure generally relates to channel state information (CSI) feedback in wireless communication systems. More specifically, the subject matter disclosed herein relates to improvements to machine learning (ML)-based CSI prediction techniques at user equipment (UE). Background Art

[0002] In massive multiple-input, multiple-output (MIMO) systems, the reception of real-time CSI at the base station (BS) plays an important role in leveraging the benefits of enhanced MIMO technology. However, existing challenges for CSI feedback can include large uplink bandwidth (i.e., feedback overhead), low reconstruction accuracy at the BS, and channel aging.

[0003] In medium and high UE mobility conditions, the channel response estimated by the UE and the actual physical downlink shared channel (PDSCH) transmission channel may be mismatched by the base station due to the UE's mobility, which can be referred to as channel aging. In high-speed UE scenarios, even when the downlink CSI is correctly fed back to the base station by the UE, channel aging can seriously affect downlink transmission performance.

[0004] CSI prediction can be used to mitigate channel aging issues. For example, the BS or UE can use previous channel observations (or CSI reference signal (RS) estimation results) to predict future CSI, thereby reducing the impact of processing and feedback delays.

[0005] Attempts have been made to perform CSI prediction using analytical techniques such as autoregression and polynomial extrapolation. However, such techniques rely on channel models and may not reflect the complexity of channel evolution.

[0006] To address this problem, ML-based CSI prediction has become a powerful solution.

[0007] One issue with the aforementioned approaches is that, for both analytical and ML-based solutions, the primary bottleneck for implementing CSI prediction in the UE is the complexity of the hardware (HW) required to store past observations. For example, a UE supporting Third Generation Partnership Project (3GPP) Release (Rel.) 17 maintains CSI-RS channel estimation results with dimensions of up to 273 resource blocks × 32 transmit ports × four receive antennas for a single timestamp. In 3GPP Rel. 18, the UE may need to maintain multiple CSI observations over time for CSI prediction, which would require a buffer size too large to be implemented in the UE due to limited HW resources.

[0008] To overcome these problems, this paper describes systems and methods for a joint ML-based framework for CSI compression and CSI prediction. The above methods improve upon previous approaches because they significantly reduce HW complexity while maintaining prediction performance. Summary of the Invention

[0009] In an embodiment, a method is provided, wherein: an encoder of a UE compresses a channel vector of a CSI matrix to generate a corresponding compressed vector, and a processor of the UE generates a predicted channel vector by performing ML-based CSI prediction on the compressed vector. The UE reports predicted CSI to a base station based on the predicted channel vector.

[0010] In an embodiment, a UE is provided, comprising: an encoder that compresses a channel vector of a CSI matrix to generate a corresponding compressed vector; and a processor that generates a predicted channel vector by performing ML-based CSI prediction on the compressed vector, and reports the predicted CSI to a BS based on the predicted channel vector.

[0011] In an embodiment, a UE is provided, comprising a processor and a non-transitory computer-readable storage medium storing instructions. When executed, the instructions cause the processor to compress a channel vector of a CSI matrix to generate a corresponding compressed vector, generate a predicted channel vector by performing ML-based CSI prediction on the compressed vector, and report predicted CSI to a base station (BS) based on the predicted channel vector. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In the following sections, various aspects of the subject matter disclosed herein will be described with reference to exemplary embodiments shown in the accompanying drawings, in which: Figure 1 is a diagram illustrating signal timing of a BS and a UE for CSI prediction; Figure 2 is a diagram illustrating ML-based CSI prediction without compression; Figure 3 is a diagram illustrating a joint framework for CSI compression and prediction according to an embodiment; Figure 4 is a diagram illustrating a joint framework with combined blocks for prediction and decoding according to an embodiment; Figure 5 is a flowchart illustrating a method for generating accurate CSI according to an embodiment; and Figure 6 is a block diagram of an electronic device in a network environment according to an embodiment. DETAILED DESCRIPTION

[0013] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, those skilled in the art will appreciate that the disclosed aspects may be practiced without these specific details. In other instances, well-known methods, processes, components, and circuits have not been described in detail in order to avoid obscuring the subject matter disclosed herein.

[0014] References throughout this specification to "one embodiment" or "an embodiment" mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment disclosed herein. Thus, the phrases "in one embodiment" or "in an embodiment" or "according to an embodiment" (or other phrases of similar meaning) that appear in various places throughout this specification may not necessarily all refer to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In this regard, as used herein, the word "exemplary" means "serving as an example, instance, or illustration." Any embodiment described herein as "exemplary" should not be construed as necessarily preferred or advantageous over other embodiments. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. Furthermore, depending on the context of the discussion herein, singular terms may include corresponding plural forms, and plural terms may include corresponding singular forms. Similarly, hyphenated terms (e.g., "two-dimensional," "predetermined," "pixel-specific," etc.) may occasionally be used interchangeably with corresponding non-hyphenated versions (e.g., "two-dimensional," "predetermined," "pixel-specific," etc.), and capitalized terms (e.g., "Counter Clock," "Row Select," "PIXOUT," etc.) may be used interchangeably with corresponding non-capitalized versions (e.g., "counter clock," "row select," "pixout," etc.). Such occasional interchangeable usage should not be considered inconsistent with one another.

[0015] Furthermore, depending on the context of the discussion herein, singular terms may include corresponding plural forms, and plural terms may include corresponding singular forms. It should also be noted that the various figures (including assembly diagrams) shown and discussed herein are for illustrative purposes only and are not drawn to scale. For example, the dimensions of some elements may be exaggerated relative to other elements for clarity. Furthermore, where considered appropriate, reference numerals have been repeated among the figures to indicate corresponding and / or similar elements.

[0016] The terms used herein are for the purpose of describing some example embodiments only and are not intended to limit the claimed subject matter. As used herein, unless the context clearly indicates otherwise, the singular forms "a," "an," and "the" are intended to include the plural forms as well. It will be further understood that when used in this specification, the terms "include" and / or "comprise" specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0017] It should be understood that when an element or layer is referred to as being on, “connected to,” or “coupled to” another element or layer, it can be directly on, connected to, or coupled to another element or layer, or there can be intermediate elements or layers. In contrast, when an element is referred to as being “directly on,” “directly connected to,” or “directly coupled to” another element or layer, there are no intermediate elements or layers. The same reference numerals always refer to the same elements. As used herein, the term “and / or” includes any and all combinations of one or more of the associated listed items.

[0018] As used herein, the terms "first," "second," and the like are used as labels for the nouns that follow them and do not imply any type of ordering (e.g., spatial, temporal, logical, etc.) unless explicitly defined as such. Furthermore, the same reference numerals may be used across two or more figures to refer to parts, components, blocks, circuits, units, or modules having the same or similar functions. However, this usage is for simplicity of illustration and ease of discussion only; it does not imply that the construction or architectural details of such components or units are the same in all embodiments, or that such commonly referenced components / modules are the only way to implement some example embodiments disclosed herein.

[0019] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this subject matter belongs. It will be further understood that terms (such as those defined in commonly used dictionaries) should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly defined as such herein.

[0020] As used herein, the term "module" refers to any combination of software, firmware, and / or hardware configured to provide the functionality described herein in conjunction with the module. For example, software may be implemented as a software package, code, and / or instruction set or instructions, and the term "hardware" as used in any embodiment described herein may include, for example, individually or in any combination, an assembly, a hardwired circuit, a programmable circuit, a state machine circuit, and / or firmware that stores instructions executed by a programmable circuit. Modules may be implemented collectively or individually as circuits that form part of a larger system (e.g., but not limited to, an integrated circuit (IC), a system on a chip (SoC), an assembly, etc.).

[0021] The embodiments described herein provide a method in which: a UE provides an improved trade-off between CSI prediction performance and HW complexity. More specifically, CSI prediction is performed in a latent space (e.g., a space of dimensionality reduction or data compression) to significantly reduce HW complexity in terms of the buffer size used to store past observations, and a joint ML block for CSI prediction and decompression and task-aware classification of the ML block significantly reduce HW complexity (i.e., storage complexity) in terms of the number of required ML model parameters.

[0022] In a massive MIMO orthogonal frequency division multiplexing (OFDM) system, a single user (i.e., UE) and a single BS can have data streams and data streams.

[0023] Figure 1 is a diagram showing signal timing of BS and UE for CSI prediction. BS 102 can K subcarriers (eg, access point (AP)-CSI-RS resources 106) to transmit a signal having OFDM transmission of data streams. k The received signal on the subcarriers can be expressed as shown in the following equation (1).

[0024] …(1) In equation (1), 、 、 、 Respectively in timestamp n The channel matrix in the frequency domain at the BS, the precoding matrix at the BS, the data symbols sent in the downlink, and the k The additive white Gaussian noise on the subcarriers. n In this paper, it represents the time slot index, but n Any suitable unit of time may be represented.

[0025] is a tensor representing the entire CSI stacked by the channel matrices over all subcarriers, while real(·) and imag(·) represent the real and imaginary parts of the input, respectively.

[0026] The resolution in the frequency dimension may vary according to the granularity of AI-based CSI feedback as defined by 3GPP, and in this document, resource blocks (RBs) are used.

[0027] The key parameters used in CSI prediction are defined in Table 1 below.

[0028] Table 1

[0029] The goal of UE 104 is to be based on having until time n of m The past of the offset AP-CSI-RS resource 106 p observations (i.e., ) to predict the future channel tensor. Specifically, Figure 1 , receiving AP-CSI-RS resources 106 at slot indices 1, 3, 5, and 7, where n =7. Figure 1 As shown, the UE 104 can predict and report CSI with multiple timestamps 108. The UE 104 can predict and report CSI with multiple timestamps 108. n Feedback delay from the last observation at n ref Then in the time slot index n' The predicted CSI 108 is reported to BS 102 at the offset. d =1 for time n + D (Right now, =1) predict a single future channel matrix 110 while leaving the extension to the general case.The method may then be repeated for the next burst 112 of AP-CSI-RS resources received starting at slot index 17.

[0030] UE can design prediction function To predict the future channel tensor, the future channel tensor can be expressed as the following equation (2).

[0031] …(2) In Equation (2), ρ represents the parameter used for prediction. However, due to the limited HW resources in UE, it is not possible to store all past p Therefore, a preprocessing function can be designed and post-processing functions , so that equation (3) is satisfied, as described below.

[0032] …(3) In equation (3), the prediction function is changed to . Preprocessing function Aims to The ratio will reduce the dimension of CSI, and UE stores Instead of H, this can significantly reduce storage complexity. can be jointly designed to minimize the prediction loss, which can be expressed by the following equation (4).

[0033] …(4) ML-based CSI prediction can be performed using the limited HW resources of the UE.

[0034] Figure 2 is a diagram illustrating ML-based CSI prediction without compression. More specifically, Figure 2 A naive ML based architecture for predicting future CSI is shown, where the prediction is performed in the channel domain.

[0035] The respective channel vectors 202, 204, and 206 may be stored in a buffer. Specifically, the CSI matrix ,in n is a timestamp (e.g., slot index in LTE / NR), which can be divided into a vector , so that each part has a single element in the BS antenna and UE antenna dimensions. Prediction may be performed at the ML-based CSI prediction block 208, which outputs the future channel vector or a predicted CSI of 210.

[0036] However, the past p CSI observations, , can have up to This is impractical to store in the UE due to limited HW resources.

[0037] In addition to the large buffer size, this architecture may require a large number of ML parameters.

[0038] Figure 3 is a diagram illustrating a joint framework of CSI compression and prediction according to an embodiment.

[0039] To provide prediction in the latent space, autoencoder (AE)-based CSI compression and ML-based CSI prediction can be jointly designed, which allows prediction in the compressed domain (ie, latent space) and maintains prediction performance.

[0040] Compression may be applied to each channel vector 302, channel vector 304, and channel vector 306 at encoders 312, 314, and 316, respectively, before prediction, and the resulting compressed latent vectors 318, 320, and 322 are stored in a cache, thereby allowing a significant reduction in the required buffer size. Specifically, the CSI matrix ,in n is a timestamp (e.g., slot index in LTE / NR), which can be divided into a vector , so that each part has a single element in the BS antenna and UE antenna dimensions. The divided vectors can be compressed by the encoder of AE into a latent vector , making the dimensionality of the latent vector much smaller than the original channel vector. The latent vector can be stored in a buffer at multiple timestamps to predict future channel information. Because the size of the latent vector is much smaller than the size of the original channel vector (i.e., it fits into 2), the UE can significantly save buffer storage complexity.

[0041] Prediction may be performed in the latent space at the ML-based latent prediction block 308, which outputs a latent representation of the future CSI 324. Specifically, by utilizing p Past latent vectors 322, the ML-based potential prediction block can output the future potential vector ( + )324, of which D is the predicted length.

[0042] The decoder 326 of the AE can decompress the future potential vector 324 into the future channel vector Or predict CSI 310.

[0043] Therefore, in addition to the reduced buffer size, the above architecture can significantly reduce the number of ML parameters, which saves storage complexity and inference time. For example, a single (and identical) encoder can be applied to channel vectors with different timestamps for CSI compression. Furthermore, the dimensionality of the latent space can be significantly smaller than that of the original channel matrix. Since prediction is performed in the latent space, the number of ML parameters used for prediction can be significantly reduced.

[0044] For the joint framework of CSI compression and prediction, separate training and joint training can be considered. The training methods are shown in Table 3 below.

[0045] Table 3

[0046] Joint training can significantly improve the prediction performance of the proposed joint framework. For example, when the latent size is 16, the buffer size can be increased by The ratio of is reduced, and the performance degradation is small (for example, less than 0.25dB in terms of NMSE).

[0047] To further reduce HW complexity, ML prediction blocks and decoding in the latent space can be combined.

[0048] Figure 4 is a diagram illustrating a joint framework with combined blocks for prediction and decoding according to an embodiment.

[0049] Compression may be applied to each channel vector 402, 404, and 406 at encoders 412, 414, and 416, respectively, before prediction, and the resulting compressed latent vectors 418, 420, and 422 may be stored in a buffer, thereby achieving a significantly reduced required buffer size. Specifically, the CSI matrix ,in, n is a timestamp (e.g., slot index in LTE / NR), which can be divided into a vector 402, 404 and 406, so that each part has a single element in the BS antenna and UE antenna dimensions. The divided vectors can be compressed by the encoder of AE into a latent vector , making the dimensionality of the latent vector much smaller than the original channel vector. Latent vectors 418, 420, and 422 can be stored in a buffer at multiple timestamps to predict future channel information. Because the size of the latent vector is much smaller than the size of the original channel vector (i.e., it fits into 2), the UE can significantly save buffer storage complexity.

[0050] ML-based prediction and decoding may be performed in the latent space at ML-based latent prediction and decoder block 408, which outputs the future channel vector or predicted CSI 410.

[0051] like Figure 4 As shown, the number of ML parameters can be further reduced by combining prediction and decoding into a single block. Approximately 2k ML parameters can be saved for prediction and decoding by combining the blocks.

[0052] The NMSE performance of the new framework can be further evaluated using the combination block. The combination block not only reduces the number of ML parameters but also improves the prediction performance.

[0053] As described above, a new joint framework is provided for CSI compression and CSI prediction, which can significantly reduce the HW complexity while maintaining the prediction performance. Although the above embodiments are for the case where the number of predicted CSIs is one and the offset between CSI-RS measurements is fixed, the embodiments can be extended to generalized parameters. Multiple ML models are classified to support various channel characteristic ranges, Doppler ranges, and signal-to-noise ratio (SNR) ranges.

[0054] Depending on the embodiment, a different ML model can be trained for each channel characteristic. In this case, the number of ML models can scale linearly with the number of channel profiles to be supported. Categorizing different environments into groups can reduce the number of ML models while maintaining prediction performance. Compression and prediction models can be categorized using channel delay characteristics and Doppler, respectively. Using an architecture with joint compression and prediction, categorizing ML models can be beneficial to provide a trade-off between performance and hardware complexity.

[0055] Specifically, task-aware classification of ML blocks is designed to support diverse channel environments with limited hardware resources to store various ML models in the UE. For each block, the UE classifies the ML model using different classification methods based on the block's task. This feature saves hardware complexity to store multiple ML parameters.

[0056] To support various channel environments (e.g., SNR, delay characteristics, Doppler, etc.), the UE can support multiple ML models (e.g., encoder, decoder, prediction) for each block. For example, the UE can train and maintain three encoder models for short delay characteristics, medium delay characteristics, and long delay characteristics respectively. 、 and .

[0057] Because each block has a different task (compression or prediction), different classification methods for multiple ML models can be applied to each block. Each task has different important characteristics. For example, delay characteristics are important for AEs (encoders and decoders) designed for compression and decompression, while Doppler is important for prediction blocks.

[0058] Assumptions and are the collections of ML model indices for delay characteristics and Doppler, respectively. By applying task-aware classification of ML models, the number of ML models can be obtained from Reduce to and ,in, 、 and They represent the ML model for encoder, the ML model for decoder, and the ML model for prediction, respectively.

[0059] Figure 5 is a flow chart illustrating a method for generating accurate CSI according to an embodiment. At 502, an encoder of a UE may compress a received channel vector of a CSI matrix to generate a compressed vector. The CSI matrix may be divided into channel vectors. The dimension of the compressed vector is smaller than the dimension of the channel vector. The encoder may be part of an AE.

[0060] At 504, the UE may store the compressed vector in a buffer of the UE at a plurality of timestamps.

[0061] At 506, the UE's processor may generate a predicted channel vector by performing ML-based CSI prediction on the compressed vector. The predicted channel vector may be generated in two steps: performing ML-based CSI prediction using the compressed vector to generate a latent representation of the predicted channel vector, and then decompressing the latent representation by the UE's decoder to generate the predicted channel vector. Alternatively, the predicted channel vector may be generated in a single step by performing ML-based CSI prediction and decoding on the compressed vector.

[0062] Compressing the channel vector, performing ML-based CSI prediction, and decompressing the latent representation may be performed using a task-dependent ML model. For example, compressing the channel vector and decompressing the latent representation may include selecting a first ML model from a first set of ML models for encoding and decoding, and performing ML-based CSI prediction may include selecting a second ML model from the first set of ML models for CSI prediction.

[0063] At 508, the UE may report the predicted CSI to the BS based on the predicted channel vector. Figure 1 As described, the UE reports predicted CSI to the BS after a feedback delay from the last observation. Using the predicted CSI, the BS can predict the future channel matrix.

[0064] Figure 6 is a block diagram of electronic devices in a network environment 600 according to an embodiment.

[0065] Reference Figure 6In network environment 600, electronic device 601 can communicate with electronic device 602 via a first network 698 (e.g., a short-range wireless communication network), or with electronic device 604 or server 608 via a second network 699 (e.g., a long-range wireless communication network). Electronic device 601 can communicate with electronic device 604 via server 608. Electronic device 601 may include a processor 620, a memory 630, an input device 650, an audio output device 655, a display device 660, an audio module 670, a sensor module 676, an interface 677, a haptic module 679, a camera module 680, a power management module 688, a battery 689, a communication module 690, a subscriber identification module (SIM card) 696, or an antenna module 697. In one embodiment, at least one of the components (e.g., display device 660 or camera module 680) may be omitted from electronic device 601, or one or more other components may be added to electronic device 601. Some of the components may be implemented as a single integrated circuit (IC). For example, the sensor module 676 (eg, a fingerprint sensor, an iris sensor, or an illumination sensor) may be embedded in the display device 660 (eg, a display).

[0066] The processor 620 may execute software (eg, program 640 ) to control at least one other component (eg, hardware or software component) of the electronic device 601 coupled to the processor 620 , and may perform various data processing or calculations.

[0067] As at least part of data processing or computing, the processor 620 may load commands or data received from another component (e.g., the sensor module 676 or the communication module 690) into the volatile memory 632, process the commands or data stored in the volatile memory 632, and store the resulting data in the non-volatile memory 634. The processor 620 may include a main processor 621 (e.g., a central processing unit (CPU) or an application processor (AP)) and an auxiliary processor 623 (e.g., a graphics processing unit (GPU), an image signal processor (ISP), a sensor hub processor, or a communication processor (CP)) that may operate independently of or in conjunction with the main processor 621. Additionally or alternatively, the auxiliary processor 623 may be adapted to consume less power than the main processor 621 or to perform specific functions. The auxiliary processor 623 may be implemented separately from the main processor 621 or as part of the main processor 621.

[0068] When the main processor 621 is in an inactive state (e.g., a sleep state), the auxiliary processor 623 may control at least some of the functions or states related to at least one of the components of the electronic device 601 (e.g., the display device 660, the sensor module 676, or the communication module 690) instead of the main processor 621, or when the main processor 621 is in an active state (e.g., executing an application), the auxiliary processor 623 may control at least some of the functions or states related to at least one of the components of the electronic device 601 (e.g., the display device 660, the sensor module 676, or the communication module 690) together with the main processor 621. The auxiliary processor 623 (e.g., an image signal processor or a communication processor) may be implemented as part of another component (e.g., the camera module 680 or the communication module 690) that is functionally related to the auxiliary processor 623.

[0069] The memory 630 can store various data used by at least one component of the electronic device 601 (e.g., the processor 620 or the sensor module 676). The various data may include, for example, software (e.g., program 640) and input data or output data for commands associated therewith. The memory 630 may include a volatile memory 632 or a non-volatile memory 634. The non-volatile memory 634 may include an internal memory 636 and / or an external memory 638.

[0070] The program 640 may be stored as software in the memory 630 and may include, for example, an operating system (OS) 642 , middleware 644 , or applications 646 .

[0071] The input device 650 may receive a command or data to be used by another component (eg, processor 620) of the electronic device 601 from outside the electronic device 601 (eg, a user). The input device 650 may include, for example, a microphone, a mouse, or a keyboard.

[0072] The sound output device 655 can output sound signals to the outside of the electronic device 601. The sound output device 655 may include, for example, a speaker or a receiver. The speaker can be used for general purposes (such as playing multimedia or recording), and the receiver can be used to receive incoming calls. The receiver can be implemented separately from the speaker or as part of the speaker.

[0073] The display device 660 can visually provide information to an external portion of the electronic device 601 (e.g., a user). The display device 660 may include, for example, a display, a hologram device, or a projector, and a control circuit for controlling a corresponding one of the display, hologram device, and projector. The display device 660 may include a touch circuit adapted to detect a touch, or a sensor circuit adapted to measure the strength of a force caused by a touch (e.g., a pressure sensor).

[0074] The audio module 670 can convert sound into electrical signals and vice versa. The audio module 670 can obtain sound via the input device 650 or output sound via the sound output device 655 or an earphone of an external electronic device 602 directly (e.g., wired) or wirelessly coupled to the electronic device 601.

[0075] The sensor module 676 can detect the operating state of the electronic device 601 (e.g., power or temperature) or the environmental state outside the electronic device 601 (e.g., the state of the user), and then generate an electrical signal or data value corresponding to the detected state. The sensor module 676 may include, for example, a gesture sensor, a gyroscope sensor, an atmospheric pressure sensor, a magnetic sensor, an acceleration sensor, a grip sensor, a proximity sensor, a color sensor, an infrared (IR) sensor, a biometric sensor, a temperature sensor, a humidity sensor, or an illumination sensor.

[0076] The interface 677 may support one or more designated protocols for the electronic device 601 to be directly (e.g., wired) or wirelessly coupled to the external electronic device 602. The interface 677 may include, for example, a High-Definition Multimedia Interface (HDMI), a Universal Serial Bus (USB) interface, a Secure Digital (SD) card interface, or an audio interface.

[0077] The connection terminal 678 may include a connector through which the electronic device 601 can be physically connected to the external electronic device 602. The connection terminal 678 may include, for example, an HDMI connector, a USB connector, an SD card connector, or an audio connector (eg, a headphone connector).

[0078] The haptic module 679 may convert the electrical signal into mechanical stimulation (eg, vibration or movement) or electrical stimulation that can be recognized by the user through tactile sensation or kinesthetic sensation. The haptic module 679 may include, for example, a motor, a piezoelectric element, or an electrical stimulator.

[0079] The camera module 680 can capture still images or moving images. The camera module 680 may include one or more lenses, an image sensor, an image signal processor, or a flash. The power management module 688 can manage the power supplied to the electronic device 601. The power management module 688 can be implemented as at least a portion of a power management integrated circuit (PMIC), for example.

[0080] The battery 689 may provide power to at least one component of the electronic device 601. The battery 689 may include, for example, a non-rechargeable primary battery, a rechargeable secondary battery, or a fuel cell.

[0081] The communication module 690 can support establishing a direct (e.g., wired) communication channel or a wireless communication channel between the electronic device 601 and an external electronic device (e.g., electronic device 602, electronic device 604, or server 608) and performing communication via the established communication channel. The communication module 690 may include one or more communication processors that can operate independently of the processor 620 (e.g., AP) and support direct (e.g., wired) communication or wireless communication. The communication module 690 may include a wireless communication module 692 (e.g., a cellular communication module, a short-range wireless communication module, or a global navigation satellite system (GNSS) communication module) or a wired communication module 694 (e.g., a local area network (LAN) communication module or a power line communication (PLC) module). A corresponding one of these communication modules may communicate via a first network 698 (e.g., a short-range communication network such as BLUETOOTH). TM , Wireless Fidelity (Wi-Fi) Direct, or Infrared Data Association (IrDA) standards) or a second network 699 (e.g., a long-distance communication network such as a cellular network, the Internet, or a computer network (e.g., a LAN or a wide area network (WAN))) to communicate with an external electronic device. These various types of communication modules may be implemented as a single component (e.g., a single IC), or may be implemented as multiple components (e.g., multiple ICs) separated from each other. The wireless communication module 692 may use user information (e.g., an International Mobile Subscriber Identity (IMSI)) stored in the user identification module 696 to identify and authenticate the electronic device 601 in a communication network (such as the first network 698 or the second network 699).

[0082] Antenna module 697 can transmit or receive signals or power to or from the outside of electronic device 601 (e.g., an external electronic device). Antenna module 697 may include one or more antennas, and at least one antenna suitable for the communication scheme used in a communication network (such as first network 698 or second network 699) may be selected by, for example, communication module 690 (e.g., wireless communication module 692). Signals or power can then be transmitted or received between communication module 690 and the external electronic device via the selected at least one antenna.

[0083] Commands or data can be sent or received between the electronic device 601 and the external electronic device 604 via the server 608 coupled to the second network 699. Each of the electronic device 602 and the electronic device 604 can be a device of the same type as the electronic device 601 or a different type. All or some of the operations to be performed at the electronic device 601 can be performed at one or more of the external electronic device 602, the external electronic device 604, or the server 608. For example, if the electronic device 601 should automatically perform a function or service, or perform the function or service in response to a request from a user or another device, the electronic device 601 can request the one or more external electronic devices to perform at least a portion of the function or service instead of performing the function or service, or the electronic device 601 can request the one or more external electronic devices to perform at least a portion of the function or service in addition to performing the function or service. The one or more external electronic devices that receive the request may execute at least a portion of the requested function or service, or execute additional functions or additional services related to the request, and transmit the execution results to the electronic device 601. The electronic device 601 may provide the results as at least a part of the reply to the request, with or without further processing the results. To this end, for example, cloud computing technology, distributed computing technology, or client-server computing technology may be used.

[0084] Embodiments of the subject matter and operations described in this specification may be implemented in digital electronic circuitry, or in computer software, firmware, or hardware (including the structures disclosed in this specification and their structural equivalents), or in a combination of one or more of these. Embodiments of the subject matter described in this specification may be implemented as one or more computer programs (i.e., one or more modules of computer program instructions) encoded on a computer storage medium for execution by, or to control the operation of, a data processing device. Alternatively or additionally, the program instructions may be encoded on an artificially generated propagated signal (e.g., a machine-generated electrical, optical, or electromagnetic signal) that is generated to encode information for transmission to a suitable receiver device for execution by the data processing device. A computer storage medium may be, or be included in, a computer-readable storage device, a computer-readable storage substrate, a random or serial access memory array or device, or a combination thereof. Furthermore, while a computer storage medium is not a propagated signal, a computer storage medium may be a source or destination of computer program instructions encoded in an artificially generated propagated signal. The computer storage medium may also be, or be included in, one or more separate physical components or media (e.g., multiple CDs, disks, or other storage devices). In addition, the operations described in this specification may be implemented as operations performed by a data processing apparatus on data stored on one or more computer-readable storage devices or on data received from other sources.

[0085] Although this specification may contain many specific implementation details, the implementation details should not be interpreted as limiting the scope of any claimed subject matter, but rather as descriptions of features specific to a particular embodiment. Specific features described in this specification in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented in multiple embodiments individually or in any suitable subcombination. Furthermore, although the above features may be described as functioning in a particular combination and even initially claimed as such, in some cases one or more features from the claimed combination may be removed from the combination, and the claimed combination may be directed to a subcombination or a variation of the subcombination.

[0086] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that the operations be performed in the particular order shown, or in sequence, or that all illustrated operations be performed, in order to achieve the desired results. In certain circumstances, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems may generally be integrated together in a single software product or packaged into multiple software products.

[0087] Thus, particular embodiments of the subject matter have been described herein. Other embodiments are within the scope of the following claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve the desired results. In certain embodiments, multitasking and parallel processing may be advantageous.

[0088] As those skilled in the art will recognize, the innovative concepts described herein may be modified and varied over a wide range of applications. Accordingly, the scope of the claimed subject matter should not be limited to any of the specific exemplary teachings discussed above, but is instead defined by the appended claims.

Claims

1. A method for compressing and predicting channel state information (CSI), comprising: Compressing, by an encoder of a user equipment (UE), a channel vector of a channel state information (CSI) matrix to generate a corresponding compressed vector; Generating, by a processor of the UE, a predicted channel vector by performing CSI prediction based on machine learning (ML) on the compressed vector; and The UE reports predicted CSI to a base station BS based on the predicted channel vector.

2. The method according to claim 1, wherein The encoder comprises an autoencoder of the UE.

3. The method according to claim 1, wherein Each compression vector has fewer dimensions than the corresponding channel vector.

4. The method according to claim 1, wherein The CSI matrix is divided into the channel vectors.

5. The method according to claim 1, further comprising: The compressed vector is stored in a buffer of the UE.

6. The method according to claim 5, wherein: The compressed vectors are stored in the buffer at a plurality of time stamps.

7. The method according to claim 1, wherein Generating the predicted channel vector includes: performing, by the processor, ML-based CSI prediction using the compressed vector to generate a latent representation of the predicted channel vector; and The latent representation is decompressed by a decoder of the UE to generate the predicted channel vector.

8. The method according to claim 1, wherein Generating the predicted channel vector includes: The processor performs ML-based CSI prediction and decoding on the compressed vector to generate the predicted channel vector.

9. The method according to claim 8, wherein The compressing the channel vector includes: selecting a first ML model from a first group of ML models for encoding and decoding, and the performing ML-based CSI prediction includes: selecting a second ML model from a second group of ML models for CSI prediction.

10. The method according to claim 1, wherein The operations of compressing the channel vector and performing ML-based CSI prediction are performed using a task-dependent ML model.

11. A user equipment (UE), comprising: an encoder configured to compress a channel vector of a channel state information (CSI) matrix to generate a corresponding compressed vector; as well as The processor is configured to generate a predicted channel vector by performing CSI prediction based on machine learning (ML) on the compressed vector, and report the predicted CSI to the base station BS based on the predicted channel vector.

12. The UE according to claim 11, wherein: The encoder comprises an autoencoder of the UE.

13. The UE according to claim 11, wherein: Each compression vector has fewer dimensions than the corresponding channel vector.

14. The UE according to claim 11, further comprising: A buffer is configured to store the compressed vector.

15. The UE according to claim 14, wherein: The compressed vectors are stored in the buffer at a plurality of time stamps.

16. The UE according to claim 11, further comprising a decoder, wherein: When generating the predicted channel vector, the processor is configured to: perform ML-based CSI prediction using the compressed vector to generate a latent representation of the predicted channel vector; as well as When generating the predicted channel vector, the decoder is configured to decompress the latent representation to generate the predicted channel vector.

17. The UE according to claim 11, wherein: When generating the predicted channel vector, the processor is configured to: ML-based CSI prediction and decoding are performed on the compressed vector to generate the predicted channel vector.

18. The UE according to claim 17, wherein: When compressing the channel vector, the processor is configured to select a first ML model from a first group of ML models for encoding and decoding, and when performing ML-based CSI prediction, the processor is configured to select a second ML model from a second group of ML models for CSI prediction.

19. The UE according to claim 11, wherein: The operations of compressing the channel vector and performing ML-based CSI prediction are performed using a task-dependent ML model.

20. A user equipment (UE), comprising: processor; as well as A non-transitory computer-readable storage medium storing instructions that, when executed, cause the processor to perform the following operations: Compressing a channel vector of a channel state information CSI matrix to generate a corresponding compressed vector; generating a predicted channel vector by performing machine learning (ML)-based CSI prediction on the compressed vector; and The predicted CSI is reported to the base station BS based on the predicted channel vector.

Images