Method and apparatus for updating autoencoder for channel state information feedback in wireless communication system

Abstract

Disclosed is a method by which a base station (BS) updates an automatic encoder (AE) for channel state information (CSI) feedback in a wireless communication system, the method including transmitting a data collection instruction message to a user equipment (UE), the data collection instruction message indicating collection of pre-compression data corresponding to compressed data, receiving the pre-compression data from the UE, and transmitting the pre-compression data to the UE. And updating the AE based on the received data prior to compression.

Description

Technical Field

[0001] The present disclosure relates generally to wireless communication systems, and more particularly, to methods and apparatus for updating an artificial intelligence (AI) model included in an autoencoder (AE) for compressing channel state information (CSI) data. Background Art

[0002] In the development process of radio communication, technologies for services targeting people such as voice, multimedia and data have been developed. The exponentially increasing connected devices after the commercialization of the fifth generation (5G) communication system are expected to be connected to the communication network. As examples of things connected to the network, there may be cars, robots, drones, home appliances, displays, smart sensors installed in various infrastructures, construction machinery and factory equipment. It is expected that mobile devices will evolve to various form factors, such as augmented reality (AR) glasses, virtual reality (VR) headsets and hologram devices. In order to provide various services by connecting hundreds of billions of devices and things in the sixth generation (6G) era, efforts are being made to develop better 6G communication systems, which are called super 5G systems.

[0003] In the 6G communication system that is expected to become a reality by about 2030, the maximum transmission rate is 1 trillion bits per second (bps), that is, 1000 gigabits per second, and the maximum wireless delay is 100 microseconds (μsec). In other words, compared with the 5G communication system, the transmission rate becomes 50 times faster and the wireless delay is reduced to one tenth (1 / 10) in the 6G communication system.

[0004] In order to obtain these high data transmission rates and ultra-low latency, 6G communication systems are considered to be implemented in the terahertz (THz) band (e.g., ranging from 95 gigahertz (GHz) to 3THz). Since the path loss and atmospheric absorption phenomena in the THz band are more serious compared to the millimeter wave (mmWave) band introduced in the 5G system, it is expected that the importance of technologies for ensuring signal range (i.e., coverage) will increase. As the main technologies for ensuring coverage, it is necessary to develop antenna transmission technologies such as radio frequency (RF) elements, antennas, new waveforms that are superior to orthogonal frequency division multiplexing (OFDM) in terms of coverage, beamforming and massive multiple input multiple output (MIMO), full-dimensional MIMO (FFD-MIMO), array antennas, and massive antennas. New technologies for increasing the coverage of terahertz band signals are being discussed, such as metamaterial-based lenses and antennas, high-dimensional spatial multiplexing technologies using orbital angular momentum (OAM), and reconfigurable smart surfaces (RIS).

[0005] In order to enhance frequency efficiency and system network, the following technologies are being developed in 6G communication systems: full-duplex technology that uses the same frequency resources for uplink (UL) and downlink (DL) transmissions simultaneously, network technology that uses satellites and high altitude platform stations (HAPs) in combination, network structure innovation technology that supports mobile base stations (BSs) and allows optimization and automation of network operations, dynamic spectrum sharing technology through conflict avoidance based on spectrum usage prediction, AI-based communication technology that achieves system optimization by using AI from the design stage and internalizes end-to-end AI support functions, and next-generation distributed computing technology that enables services with complexity that exceeds the computing power limit of terminals by using ultra-high performance communication and computing resources (e.g., mobile edge computing (MEC) clouds). Attempts to strengthen connectivity between devices, further optimize networks, promote softwareization of network entities, and increase openness of wireless communications are continuing by designing new protocols to be used in 6G communication systems, developing mechanisms for achieving hardware-based secure environments and secure use of data, and developing technologies for protecting privacy.

[0006] With such research and development of 6G communication systems, it is expected that the next level of hyper-connected experience will be achieved through the hyper-connectivity of the 6G communication system, which includes not only connections between things but also connections between people and things. In particular, it is predicted that services such as truly immersive extended reality (XR), high-fidelity mobile holograms, and digital replicas can be provided. In addition, services such as remote surgery, industrial automation, and emergency response with enhanced safety and reliability can be provided through the 6G communication system for application in various fields such as industry, medical care, vehicles, and appliances.

[0007] However, in the prior art, from the perspective of tuning, the acquired data set is erroneous and the CSI data compression is excessively error-prone. Therefore, there is a need in the art for a method and apparatus for ensuring CSI data compression performance and reducing errors in CSI data compression when fine-tuning an AI model included in AE. Summary of the invention

[0008] Solution to the problem

[0009] Accordingly, the present disclosure provides embodiments designed to address at least the above-mentioned problems and / or disadvantages and to provide at least the advantages described below.

[0010] One aspect of the present disclosure is to provide a technology for obtaining a data set required for a BS to fine-tune an AI model included in an AE for compressing CSI data.

[0011] Another aspect of the present disclosure is to provide a technology that ensures CSI data compression performance when fine-tuning an AI model included in an AE by instructing a user equipment (UE) by the BS to use an encoder model with compressed data (z data) with more bits when fine-tuning the AI ​​model, or by reducing CSI data compression errors by instructing the UE to report codebook-based CSI without compressing the CSI data.

[0012] Another aspect of the present disclosure is to provide a method and apparatus for updating an AI model in an AE for CSI compression in a wireless communication system, whereby a BS can obtain a data set before compression from a UE required for fine-tuning the AI ​​model for CSI compression included in the AE.

[0013] According to one aspect of the present disclosure, a method by which a BS in a wireless communication system updates an AE for CSI feedback includes: sending a data collection instruction message that instructs a UE to collect data before compression corresponding to compressed data, receiving the data before compression from the UE, and updating the AE based on the data received before compression.

[0014] According to one aspect of the present disclosure, a BS for updating an AE for CSI feedback includes a transceiver and at least one processor. The at least one processor may be configured to: send a data collection instruction message to a UE through the transceiver, the data collection instruction message instructing to collect data before compression corresponding to the compressed data, receive the data before compression from the UE through the transceiver, and update the AE based on the data received before compression.

[0015] According to one aspect of the present disclosure, a method by which a UE communicates with a BS in a wireless communication system to update an AE for CSI feedback includes: receiving a data collection instruction message from the BS, the data collection instruction message instructing the UE to collect data before compression corresponding to compressed data, creating compressed data, sending the data before compression to the BS, and receiving an AE updated based on the data collected before compression from the BS.

[0016] According to one aspect of the present disclosure, a UE in a wireless communication system for communicating with a BS to update an AE for CSI feedback includes a transceiver and at least one processor, wherein the at least one processor is configured to: receive a data collection instruction message from the BS through the transceiver, the data collection instruction message instructing to collect data before compression corresponding to compressed data, create compressed data, send the data before compression to the BS through the transceiver, and receive an AE updated based on the data before compression from the BS through the transceiver. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other aspects, features and advantages of certain embodiments of the present disclosure will become more apparent through the following description in conjunction with the accompanying drawings, in which:

[0018] Figure 1 AE for CSI compression according to an embodiment is shown;

[0019] Figure 2 A method by which a BS updates AI of AE for CSI compression in a wireless communication system according to an embodiment is shown;

[0020] Figure 3 An operation of a BS requesting a UE to collect data before compression according to an embodiment is shown;

[0021] Figure 4 An operation of the UE transmitting the data before compression together with the compressed data to the BS according to the embodiment is shown;

[0022] Figure 5 An operation of a UE transmitting data before compression to a BS at an additional request from the BS according to an embodiment is shown;

[0023] Figure 6 An operation of requesting periodic collection of data before compression through radio resource control (RRC) signaling according to an embodiment is shown;

[0024] Figure 7 An operation of a BS requesting a UE to use another encoder AI model having more bits of compressed data during fine tuning of an AI model included in an AE according to an embodiment is shown;

[0025] Figure 8 An operation of a BS receiving data before compression from a UE in a wireless communication system according to an embodiment is shown;

[0026] Fig. 9 An operation of a UE transmitting data before compression and compressed data to a BS in a wireless communication system according to an embodiment is shown;

[0027] Fig.10 shows a BS according to an embodiment; and

[0028] Fig.11 A UE according to an embodiment is shown. DETAILED DESCRIPTION

[0029] Embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings.

[0030] For the sake of clarity and conciseness, descriptions of technical contents that are well known in the art or not directly related to the present disclosure will be omitted. The terms used herein are defined by considering the functions in the present disclosure, but may vary according to the practice or intention of the user or operator. Therefore, the terms should be defined based on the description throughout this specification.

[0031] For the same reason, some parts in the drawings may be exaggerated, omitted or schematically shown. The size of the corresponding element may not fully reflect its actual size. Throughout the drawings, the same reference numerals refer to the same elements.

[0032] The embodiments of the present disclosure may be implemented in various forms and should not be construed as being limited to the embodiments set forth herein; rather, these embodiments are provided to make the present disclosure thorough and complete and will fully convey the scope of the present disclosure to those of ordinary skill in the art. Throughout the specification, the same reference numerals represent the same elements.

[0033] Throughout this disclosure, the expression “at least one of a, b, or c” refers to only a, only b, only c, both a and b, both a and c, both b and c, all of a, b, and c, or variations thereof.

[0034] Throughout the specification, a layer may also be referred to as an entity.

[0035] The BS herein may refer to an entity for performing resource allocation for a UE, and may be at least one of a gNode B, an eNode B, a Node B (or an xNode B, where x represents any letter including 'g' and 'e'), a radio access unit, a BS controller, a satellite, an aircraft, or a node in a network. A UE may include a mobile station (MS), a vehicle, a satellite, an aircraft, a cellular phone, a smart phone, a computer, or a multimedia system with a communication function. The DL herein may refer to a radio transmission path for a signal sent from a BS to a UE, and the UL may refer to a radio transmission path for a signal sent from a UE to a BS. In addition, there may be a side link (SL), which refers to a radio transmission path for a signal sent from a UE to another UE.

[0036] Although the embodiments of this document will focus on long-term evolution (LTE), advanced LTE (LTE-A) or 5G systems as examples, they can be applied to other communication systems with similar technical backgrounds or channel types. For example, other communication systems may include 5G advanced, new radio (NR) advanced or 6G mobile communication technology developed after 5G mobile communication technology (or NR), and the term 5G includes existing LTE, LTE-A and other similar services. When considered by ordinary technicians in this field, the embodiments of the present disclosure will also be applied to different communication systems with some modifications, and the extent of these modifications does not significantly deviate from the scope of the present disclosure.

[0037] It should be understood that each box and combination of boxes of the flowchart can be performed by computer program instructions. The computer program instructions can be loaded on a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, and thus when executed by the processor of the computer or other programmable data processing device, they generate means for performing the functions described in the blocks of the flowchart. The computer program instructions can also be stored in a computer-usable or computer-readable memory for a computer or other programmable data processing device, so that a product containing instruction means for performing the functions described in the blocks of the flowchart can be manufactured. The computer program instructions can also be loaded on a computer or programmable data processing device, so that the instructions can generate a process executed by a computer or other programmable data processing device to provide steps for performing the functions described in the blocks of the flowchart.

[0038] In addition, each block may represent a module, segment or part of a code that includes one or more executable instructions for performing a specific logical function. Note that in some alternative embodiments, the functions described in the blocks may not occur in order. For example, two consecutive blocks may be executed substantially simultaneously or in reverse order, depending on the corresponding functions.

[0039] The term "module" (or sometimes "unit") used herein refers to a software or hardware component that performs some functions, such as a field programmable gate array (FPGA) or an application specific integrated circuit (ASIC). However, modules are not limited to software or hardware. A module can be configured to be stored in an addressable storage medium, or to execute one or more processors. For example, a module can include components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, procedures, subroutines, program code segments, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions served by components and modules can be combined into a smaller number of components and modules, or further divided into a larger number of components and modules. In addition, components and modules can be implemented as one or more central processing units (CPUs) in an execution device or a secure multimedia card. In an embodiment of the present disclosure, a module can include one or more processors.

[0040] In the following description, for ease of explanation, terms and definitions used in the latest standards among currently existing communication standards, that is, terms and definitions used in the LTE and NR standards defined in the 3rd Generation Partnership Project (3GPP) will be used. However, the present disclosure is not limited to the terms and definitions and can be equally applied to any system conforming to other standards.

[0041] Figure 1 AE for CSI compression according to an embodiment is shown;

[0042] The ultra-LTE communication system needs to support services that can freely reflect the various requirements of users and service providers. Services considered for ultra-5G communication systems may include enhanced mobile broadband (eMBB), massive machine type communication (mMTC), and ultra-reliability low latency communication (URLLC). Beam management or support for various frequency bands is required to meet various services. In this case, there may be different channel conditions for each frequency band or beam, and a lot of resources are consumed in the process of UE estimating the channel state and reporting the channel state to the BS. In order to reduce this resource consumption, CSI compression technology and methods for compressing CSI and sending compressed CSI are being discussed.

[0043] Due to the quantization problem of the codebook, the CSI reconstructed by the codebook-based CSI feedback may cause information loss in the process of transmitting the estimated channel H. For example, when the BS sends a reference signal (RS) to the UE, the UE can estimate the channel state based on the received RS to obtain the channel And in methods such as eigenvalue decomposition (EVD) or singular value decomposition, the relationship is obtained When the value of V is known from the obtained relationship, the channel Therefore, the UE sends a precoding matrix indicator (PMI), that is, an index of a codebook that is most similar to the value of V in the codebook-based CSI feedback, to the BS. In the case where the index of a codebook similar to the value of V is sent to the BS instead of the value of V, there may be an error due to the difference between the preset codebook and the actual value of V. When the codebook is more variously configured to reduce the information loss caused by the difference between the preset codebook and the actual value of V, the amount of data to be transmitted may increase accordingly.

[0044] In the case of CSI feedback based on CSI compression technology, the UE (transmitter) can compress CSI data (V data) through an encoder and send the compressed CSI data, and the BS can decode or reconstruct the received compressed CSI data (z data) through a decoder to obtain reconstructed CSI data ( data).

[0045] refer to Figure 1 , the AE for CSI compression includes an encoder 120 on the UE side and a decoder 140 on the BS side. The encoder 120 and the decoder 140 may each include an AI model. For example, the encoder AI model included in the encoder 120 may be used to compress the CSI data (V data) 110 to create compressed data (feedback data or z data) 130. The decoder AI model included in the decoder 140 may be used to decode or reconstruct the compressed data 130 received from the UE to create reconstructed CSI data ( Data)150.

[0046] When the characteristics of the task or input data change from the training data set and the environment, the AI ​​model has degraded performance. When the performance of the AI ​​model degrades, the AI ​​model can be replaced with another AI model suitable for the characteristics of the input data or the environment, the AI ​​model can be retrained, or fine-tuning can be performed on a certain layer of the AI ​​model.

[0047] Fine-tuning an AI model may refer to further retraining the model based on the characteristics of a data set or changes in the environment in which the AI ​​model is deployed. For example, fine-tuning may include further training the AI ​​model by using a minimum number of weights for downstream tasks in addition to all weights used for AI model pre-training.

[0048] In order to decode the compressed data 130 created by compressing (encoding) the data before compression (V data) 110, it is necessary to obtain the data before compression (V data). Therefore, the reconstructed CSI data ( The AI ​​model is fine-tuned by comparing the CSI data (V data) 150 with the CSI data (V data) 110 before compression.

[0049] A BS having the decoder 140 may have more computing resources than a UE. Therefore, when performed in a BS as a stage of the decoder 140, training or fine-tuning of an AI model may be more efficient.

[0050] At the same time, the data before compression (V data) 110 required in the operation of fine-tuning the AI ​​model is created on the UE side and is not stored in the stage of the decoder 140 on the BS side. The UE for sending compressed CSI data (z data) 130 to the BS can create compressed data (z data) 130 based on the CSI data (V data), and then logically delete unnecessary CSI data before compression (V data). Therefore, even when the BS determines that the AI ​​model needs to be fine-tuned and requests the UE to send the data before compression (V data), the BS can obtain only the latest data before compression (V data). In other words, since the BS needs to train the AI ​​model only with a finite value of V, the AI ​​model may overfit the finite value of V, or the features that the AI ​​model has been trained on may disappear before fine-tuning (catastrophic forgetting).

[0051] Therefore, the BS cannot create a data set representing improved characteristics of the coverage cell unless the UE creates compressed data (z data) and stores the corresponding data before compression (V data), in which case the BS may not be able to effectively train the AI ​​model when performing fine-tuning, and therefore cannot operate the environment adaptive CSI compression system.

[0052] In order to fine-tune the AI ​​model of AE for CSI compression, the BS may instruct the UE to collect data before compression (V data) corresponding to the compressed data (z data), request the UE to send the collected data before compression (V data), and receive the data before compression (V data) from the UE. The BS may update the AI ​​model on the decoding stage (referred to as the decoder AI model) based on the data before compression (V data) received from the UE. The BS may update the AI ​​model on the encoder stage (referred to as the encoder AI model). After updating the encoder AI model, the BS may send the updated encoder AI model to the UE.

[0053] Therefore, the BS may update the AI ​​model included in the AE by requesting the UE to collect data before compression (V data) representing better characteristics of the coverage cell served by the BS.

[0054] Figure 2 A method by which a BS updates an AI model of AE for CSI compression in a wireless communication system according to an embodiment is illustrated.

[0055] The BS may update the AE for CSI feedback. For example, the BS may update the decoder AI model for decoding compressed CSI data.

[0056] In step 210, the BS instructs the UE to collect data before compression (V data) corresponding to the compressed data (z data). For example, the BS sends a data collection instruction message to the UE, the data collection instruction message instructing to collect data before compression corresponding to the compressed data.

[0057] When it is determined based on the cell-specific information that there is a change in channel characteristics, a data collection instruction message may be sent to the UE.

[0058] The data collection instruction message may include an indicator including identification information of data before compression to be collected, and may be transmitted to the UE in a physical DL control channel (PDCCH).

[0059] The BS may determine the data collection interval based on the degree of change of the channel characteristic over time. The data collection instruction message may include data collection interval information and may be sent to the UE through RRC signaling.

[0060] The data collection instruction message may include a data transmission request message requesting the UE to send the collected pre-compression data immediately after collecting the pre-compression data.

[0061] The data collection instruction message may include a data storage instruction message instructing the UE to store the collected data before compression. Afterwards, the BS may send a data transmission request message requesting the UE to send at least one piece of data before compression.

[0062] The BS may request the UE to transmit the collected data before compression (Vdata) in step 220. However, operation 220 may be omitted.

[0063] In step 230, the BS receives data before compression (data-before-compression, Vdata) from the UE.

[0064] In step 240, the BS updates the AE based on the received data before compression. The AE may include a first encoder AI model for compressing the CSI data and a decoder AI model for reconstructing the compressed CSI data. The BS updating the AE may include fine-tuning at least one of the first encoder AI model and the decoder AI model.

[0065] The BS may send, to the UE, AE update start time information, updated duration information, and identification information of a second encoder AI model to be used during a period in which the AE update is performed.

[0066] The BS may transmit, to the UE, AE update start time information, duration information of an update operation, and identification information of a codebook to be used within a period in which the AE update is performed.

[0067] In this way, the BS can obtain data required to fine-tune at least one AI model included in the AE for CSI compression.

[0068] Figure 3 An operation in which the BS 310 requests the UE 320 to collect data before compression (V data) according to an embodiment is shown.

[0069] BS 310 may request UE 320 to collect data before compression (V data) based on cell-specific information. For example, BS 310 may determine to update AE when determining that there is a change in channel characteristics based on cell-specific information, and request UE 320 to collect data before compression (V data) to update AE.

[0070] The cell-specific information may be preset and stored in the BS 310. For example, the cell-specific information may be a fixed value set to channel characteristic information determined by a field test when the BS 310 is installed. For example, the cell-specific information may refer to cell-specific channel characteristic information, which includes channel characteristic information corresponding to a specific cell, or channel characteristic information corresponding to a specific location in a cell. The cell-specific information may be a value updated in real time based on CSI data (V data) collected by a plurality of UEs 320 connected to the cell.

[0071] The specific time point or specific condition at which the BS 310 requests the UE 320 to collect the data before compression (Vdata) (or the BS 310 determines to update the AE) may be determined based on the cell-specific information in various methods.

[0072] Areas in a particular cell can be clustered based on channel characteristics or blocking characteristics. Figure 3 , the cell areas corresponding to the BS 310 may be clustered into a first cluster C1, a second cluster C2, and a third cluster C3. The UE 320 connected to the BS 310 is able to move freely in the cell areas, and when the UE 320 moves from one cluster to another, the BS 310 may request the UE 320 to collect data (V data) before compression for the CSI-RS over K time slots, where K is a natural number equal to or greater than 1.

[0073] For example, UE 320 may be located in the second cluster C2 and then may move to an area corresponding to the third cluster C3. In this case, BS 310 may request UE 320 to collect data (V data) before compression for CSI-RS over K time slots starting from the time point when UE 320 enters the third cluster C3.

[0074] BS 310 may request UE 320 to collect data before compression (V data) only when UE 320 enters a new cluster in which UE 320 is not yet located. For example, when UE 320 moves in the order of second cluster C2, third cluster C3, second cluster C2, and first cluster C1, when UE 320 moves from third cluster C3 to second cluster C2, BS 310 may not request UE 320 to collect data before compression (V data).

[0075] The BS 310 may determine a channel characteristic change interval by monitoring the state of the UE 320, such as the moving speed, frequency band, or RS received power (RSRP) of the UE 320, and request the UE 320 to collect the data before compression (V data) according to the determined channel characteristic change interval. For example, the BS 310 may determine that the channel characteristic changes approximately every 100 milliseconds (ms) according to the history of the state change of the UE 320, in which case the BS 310 may request the UE 320 to collect the data before compression (V data) every 100 ms.

[0076] When the BS 310 communicates with the UE 320 and the communication performance drastically deteriorates, the BS 310 may request the UE 320 to collect data before compression (V data) in various ways not limited to the aforementioned embodiments of the present disclosure.

[0077] Figure 4 An operation of the UE transmitting data before compression (V data) together with compressed data (z data) to the BS according to an embodiment is shown.

[0078] Once the UE collects the data, the UE can send the data before compression (V data) to the BS by creating compressed data (z data) based on the data before compression (V data), and then sending the created compressed data (z data) and the data before compression (V data) to the BS simultaneously without delay or sequentially. In this case, it is advantageous that the UE does not store the data before compression (V data) separately, and the BS can quickly use the received data before compression (V data). At the same time, when the BS does not immediately update the AE, the BS may bear the burden of identifying, storing and managing the data before compression (V data) received from each of the multiple UEs, and it may not be easy to perform flexible scheduling of resources for receiving the data before compression (V data).

[0079] refer to Figure 4 In step 410, the BS may request the UE to collect data before compression (Vdata) and also request the UE to immediately send the collected data before compression (Vdata).

[0080] In step 420, when receiving a request to collect data before compression (V data) and send the data without delay, the UE can create compressed data (z data) by compressing (encoding) the V data after obtaining the CSI data (V data), and send the z data and the V data before compression to the BS together or sequentially.

[0081] In step 430, the BS may request the UE to collect the data before compression (Vdata) again, and request the UE to immediately send the collected data before compression (Vdata). The BS may set a time interval for the UE to collect the data before compression (Vdata).

[0082] In steps 440 and 450, when receiving a request to collect data before compression (V data) within a specific time interval and send the data without delay, the UE can create compressed data (z data) by compressing (encoding) the CSI data obtained in the time interval, and send the z data and the corresponding V data before compression to the BS together or sequentially.

[0083] Figure 5 An operation of the UE transmitting data before compression to the BS at an additional request of the BS according to an embodiment is shown.

[0084] The UE may have stored the data before compression (V data) in a memory, and may then send the data to the BS at an additional request of the BS. For example, the UE may assemble the collected data before compression (V data) together in the form of a data set, and send the assembled data set to the BS. In this case, the BS may request the data before compression (V data) as needed, and does not bear the burden of identifying, storing, or managing the data before compression (V data) received from each UE. At the same time, the UE needs to store the data before compression (V data) until the data is sent to the BS, so the UE may bear the burden of allocating and managing the storage space.

[0085] refer to Figure 5 In step 510, the BS may request the UE to collect data before compression (Vdata), and also request the UE to store the collected data before compression (Vdata) until an additional request for data is received.

[0086] In step 520, when a request to collect and store data before compression (V data) is received, the UE may create compressed data (z data) by compressing (encoding) the V data after obtaining the CSI data (V data), sending the z data to the BS, and storing the V data before compression in a memory.

[0087] In step 530, the BS may request the UE to collect the data before compression (Vdata) again, and also request the UE to store the collected data before compression (Vdata) until an additional request is received. The BS may set a time interval for the UE to collect the data before compression (Vdata).

[0088] In step 540, when a request is received to collect and store data before compression (V data) in a certain time interval, the UE can create compressed data (z data) by compressing (encoding) the CSI data (V data) obtained in the time interval, sending the z data to the BS, and storing the V data before compression in a memory.

[0089] In step 550, the BS may request the UE to transmit data before compression (Vdata).

[0090] In step 560, when receiving a request to send data before compression (V data), the UE may send all stored data before compression to the BS together or sequentially. When data before compression (V data) has been stored and data is sent at an additional request of the BS, information and indexes of the data before compression (V data) may be shared between the BS and the UE in the form of metadata. For example, a V data set including at least one piece of data before compression may have a form as shown in Table 1 below.

[0091]

Table 1

[0092]

[0093] Referring to Table 1, the V data set may include various information such as the index, frequency and time stamp of the data, the RS type of the data before compression, which cell area (cluster number) the data corresponds to, and the position in the cell to which the data corresponds.

[0094] When receiving compressed data (z data), the BS may have index information of data before compression (V data) stored in the UE. When requesting the UE to send data before compression (V data), the BS may request the UE to send all stored V data sets or one or more pieces of data before compression (V data) corresponding to a specific index value or some index values. The data before compression (V data) that the BS requests the UE to send may be determined based on the priority of the corresponding V data in the V data set, which may indicate the level of importance of the V data to the AI ​​model update operation of the AE.

[0095] A weight may be set for a plurality of pieces of data (V data) before compression stored in the UE. The weight set for the data (V data) before compression may indicate the importance level of the V data for the AI ​​model update operation of the AE.

[0096] For example, the weight of the data before compression (V data) may be determined based on the timestamp. The most recently acquired CSI data may have a higher weight than the previously acquired CSI data. Referring to Table 1, the V data corresponding to index #4 may have the highest weight, and the V data corresponding to index #1 may have the lowest weight.

[0097] For example, the weight of the data before compression (V data) may also be determined based on the corresponding cell area (cluster). Compared with the CSI data obtained in the cell area with known channel information, the CSI data obtained in the cell area where the UE is not yet located may have a high weight. Referring to Table 1, the V data corresponding to index #2 may have a higher weight than the V data corresponding to other indexes.

[0098] Figure 6 An operation of requesting periodic collection of data before compression through RRC signaling according to an embodiment is shown.

[0099] The operation of the BS requesting the UE to collect the data before compression (V data) can be implemented not only by triggering the data collection through the aforementioned additional message, but also by setting the collection interval of the data before compression (V data) through RRC signaling. The method of setting the data collection interval through RRC signaling can reduce signal overhead because no additional message is sent compared to the triggering method.

[0100] The data collection interval may be determined according to the degree of change in the channel characteristics and may be reset by RRC reconfiguration.The change in channel characteristics may be made even when conditions in the cell change, such as a physical change in the environment in the cell.

[0101] refer to Figure 6In step 610, the UE may make initial access to the BS. In step 620, the BS may set a collection interval of the data before compression (V data) to 10 time slots through RRC signaling to the UE. Thereafter, when it is determined that the channel state change interval is significantly reduced, the BS may change the collection interval of the data before compression (V data) to 5 time slots through RRC signaling to the UE in step 630. The UE sends the collected data before compression (V data) to the BS. Figure 4 and Figure 5 The aforementioned method in is performed similarly.

[0102] although Figure 6 It involves setting the interval for the UE to collect the data before compression (V data) through RRC signaling, but the interval for collecting the data before compression (V data) can be set for the UE through a medium access control (MAC) control element (CE) (MAC CE) instead.

[0103] Due to the change in channel status, the data (V data) before partial compression collected and stored in the UE may no longer be used to update the AE. In this case, the BS may request the UE to delete the V data corresponding to the corresponding index. The BS may set a valid period (or timer) for the collected V data to delete the data at the end of the valid period.

[0104] Figure 7 An operation of a BS requesting a UE to use another encoder AI model having compressed data (z data) with more bits during fine tuning of an AI model included in an AE according to an embodiment is shown.

[0105] When the AI ​​model included in the AE needs to be updated due to degradation of the AE performance, a technology for ensuring the performance of the CSI feedback while the AI ​​model is updated (fine-tuned) is required.

[0106] In step 710, before the BS determines to update the AE, the UE may use a first encoder AI model that compresses CSI data (V data) into n bits to create compressed data (z data), and transmits the n-bit compressed data to the BS.

[0107] After the environment changes, the BS may determine to update the AE. Since it is difficult to ensure compression and reconstruction performance when the AE is updated, a 1-bit deactivation indicator for deactivating the existing model may be sent when the AE is updated. For example, when the deactivation indicator is 0, this indicates that the existing model in use is still to be used, and when the deactivation indicator is 1, this indicates that the existing model in use may no longer be used.

[0108] In step 720, when the BS determines to update the AE, the BS may indicate to the UE that the AE will be updated at time t 1 The update is performed within the UE, and the UE is requested to use a second encoder AI model different from the first encoder AI model to create compressed data (z data). The second encoder AI model can compress the CSI data (V data) into m bits, where m can be a number greater than n (m>n).

[0109] In step 730, the BS may request the UE to send the data before compression (V data) to update the AE. Operation 730 may correspond to the above description in conjunction with Figure 5 Operation 550 is described.

[0110] In step 740, when receiving a request to send the data before compression (V data), the UE may send all the stored data before compression to the BS together or sequentially. Operation 740 may correspond to the above in conjunction with Figure 5 Operation 560 is described.

[0111] In step 750, the UE may create compressed data (z data) using the second encoder AI model that compresses the CSI data (V data) into m bits at the request of the BS in step 720, and at the time t when the AE is updated 1 The m-bit compressed data is sent to the BS.

[0112] In step 760, when time t 1 When the UE receives an additional recovery request from the BS, the UE can reuse the first encoder AI model that compresses the CSI data (V data) into n bits to create compressed data (z data) and sends the n-bit compressed data to the BS. In this case, the first encoder AI model can be equal to the first encoder AI model in step 710, or the first encoder AI model updated by the BS during the process of fine-tuning AE.

[0113] In this way, when the AI ​​model included in the AE is being fine-tuned, the BS can instruct the UE to use an encoder model with more bits of compressed data (z data) to ensure CSI compression performance during the fine-tuning duration of the AI ​​model.

[0114] When the BS determines to update the AE, the BS may indicate to the UE that the BS will update the AE at time t 1 The UE may perform the update within t and request the UE to perform CSI feedback with the codebook method instead of the CSI compression method during the update. The codebook information may be stored in the BS and the UE to perform codebook-based CSI feedback. The UE may perform the update within t when the AE is being updated. 1The CSI is reported to the BS using the codebook method. 1 When the time has passed or the UE receives an additional recovery request from the BS, the UE may reuse the first encoder AI model that compresses the CSI data (V data) into n bits to create compressed data (z data) and sends the n-bit compressed data to the BS. In this case, the first encoder AI model may be equal to the first encoder AI model used before the update of the AE, or the first encoder AI model updated by the BS during the fine-tuning of the AE.

[0115] In this way, the BS instructs the UE to report CSI in a codebook method instead of compressing CSI, and at the same time fine-tunes the AI ​​model included in the AE, thereby reducing the CSI compression error.

[0116] Figure 8 An operation of a BS receiving data before compression from a UE in a wireless communication system according to an embodiment is illustrated.

[0117] In step 810, the BS may determine whether it is a case where the UE is collecting CSI data (V data) before compression of a cell, for example, whether the AE needs to be updated.

[0118] In step 820, when it is determined that it is the case that the UE is collecting CSI data (V data) before compression, the BS may instruct the UE to store the V data. Figure 8 Operation 820 may correspond to Figure 2 Operation 210, Figure 4 Operation 410 or Figure 5 Operation 510 or 530. When it is determined that it is not the case that the UE is collecting CSI data (V data) before compression, the method ends.

[0119] The indicator (V collection indicator) with which the BS instructs the UE to store the V data may be 1 bit. For example, when the V collection indicator has a value of 1, the V collection indicator indicates that compressed data (z data) will be created based on the V data for the RS and the V data will be stored, and when the V collection indicator has a value of 0, this indicates that z data will be created based on the V data for the RS and that the V data does not need to be stored. When the V collection indicator is 1, a V collection number (V collection number) indicating that V data for a certain number of RSs is stored from now on may be further set, and the V collection number may be N bits. When all bits of the V collection number are 0, one piece of V data will be stored, and the number of consecutive pieces of V data to be stored may be determined based on other bit values.

[0120] The V collection indicator may also indicate that the UE stores some pieces of V data. For example, the partitioned V indicator may indicate that some pieces of V data formed of K×Nt complex numbers are to be collected, where Nt is the number of BS antennas and K is the number of subbands. The inactive partitioned V indicator indicates that the entire matrix formed of K×Nt complex numbers is collected, and the active partitioned V indicator indicates that a portion of the V data corresponding to a partial range of K and a partial range of Nt is collected.

[0121] In step 830, the BS may determine whether the UE is configured to send V2 data immediately, i.e., without delay. The UE may send the collected V data to the BS immediately after collecting the V data or at an additional request of the BS. How the UE sends the V data to the BS may be determined based on the parameter upload timing. When the upload timing is 0, this indicates a method of sending the V data immediately after collecting the V data, and when the upload timing is 1, this indicates a method of sending the V data at an additional request of the BS after the BS determines that the AE needs to be fine-tuned.

[0122] When the UE is configured to send V data to the BS without delay, the method proceeds to step 840, in which the BS may allocate RBs for the UE and store the V data received in the allocated resources. The case where the UE is configured to send V data to the BS without delay may correspond to the following example. Figure 4 The embodiments of the present disclosure are shown.

[0123] When the UE is not configured to send V data to the BS without delay in step 830, the method proceeds to step 850, in which the BS may allocate RB resources for the UE after determining that the AE needs to be updated, and request the UE to send V data. The case where the UE is configured to send V data to the BS at the request of the BS may correspond to the following example. Figure 5 The embodiments of the present disclosure are shown.

[0124] When the BS requests the UE to send V data through an additional message, this indicates whether all or part of the V data stored in the UE is to be sent in all V indicator parameters. When all V indicators are 0, this indicates that all V data stored in the UE is to be sent, and when all V indicators are 1, this indicates that part of the V data stored in the UE is to be sent. When part of the V data stored in the UE is to be sent, the consumption of frequency transmission resources can be reduced. At the same time, when the BS indicates the V data to be sent by using the index value corresponding to the V data, part of the V data can be sent even when all V indicators are 1.

[0125] In step 860, the BS may store the V data received from the UE and update the AE using the V data.

[0126] The BS may request the UE to collect V data periodically rather than via a triggered method. In the periodic method, information related to the interval may be exchanged during initial access, handover (HO), etc., and thereafter, the interval may be reset according to changes in the environment. The BS may indicate to the UE whether to use a periodic method utilizing RRC signaling or MAC CE or a triggered method. For example, when the parameter periodicity indicator is 0, this indicates the use of a triggered method, and when the parameter periodicity indicator is 1, this indicates the use of a periodic method. In the periodic method, a default interval may be set in time slots or in time units, and the UE may collect V data at the set interval without an additional request from the BS.

[0127] Due to the change in channel status, a portion of the data before compression (V data) collected and stored in the UE may no longer be used to update the AE. In this case, the BS may request the UE to delete the V data related to the corresponding index. For example, when the refresh indicator is 0, this indicates that the V data will not be deleted, and when the refresh indicator is 1, this indicates that a portion of the V data will be deleted. The BS may set a valid period (or timer) for the collected V data to delete the data when the valid period has passed.

[0128] Fig. 9 An operation of a UE transmitting data before compression (V data) and compressed data (z data) to a BS in a wireless communication system according to an embodiment is illustrated.

[0129] In step 910, the UE may determine whether there is an instruction from the BS to collect V data. The BS may determine whether it is a case where the UE collects CSI data (V data) before compression of a cell, for example, whether the AE needs to be updated, and instruct the UE to collect V data.

[0130] When the BS does not instruct the UE to collect V data, the method proceeds to step 920, where the UE may compress the V data to create z data, send the z data to the BS, and delete the V data before compression to ensure memory space.

[0131] When the BS instructs the UE to collect V data, the method proceeds to step 930, where the UE may compress the V data to create z data, send the z data to the BS, and store the V data before compression in a memory.

[0132] In step 940, the UE may determine whether the UE is configured to immediately send the V data to the BS without delay. The UE may send the collected V data to the BS immediately after collecting the V data or at an additional request of the BS. How the UE sends the V data to the BS may be determined based on the upload timing parameter. When the upload timing is 0, this indicates a method of sending the V data immediately after collecting the V data, and when the upload timing is 1, this indicates a method of sending the V data at an additional request of the BS after determining that the BS needs to fine-tune the AE.

[0133] When the UE is configured to send the V data to the BS without delay, the method proceeds to step 950, in which the UE may send the V data to the BS in the allocated RB. The case where the UE is configured to send the V data to the BS without delay may correspond to the following example. Figure 4 The embodiments of the present disclosure are shown.

[0134] When the UE is not configured to send the V data to the BS without delay, the method proceeds to step 960, in which the UE may send the V data to the BS at the request of the BS for sending the V data. The case where the UE is configured to send the V data to the BS at the request of the BS may correspond to the following example. Figure 5 The embodiments of the present disclosure are shown.

[0135] Fig.10 is a schematic block diagram of a BS 1000 according to an embodiment.

[0136] refer to Fig.10 , BS1000 may include a transceiver 1010, a processor 1020, and a memory 1030. The transceiver 1010, the processor 1020, and the memory 1030 of BS1000 may operate according to the aforementioned communication method of BS1000. However, the elements of BS1000 are not limited thereto. For example, BS1000 may include more or less elements than those described above. In addition, the transceiver 1010, the processor 1020, and the memory 1030 may be implemented in a single chip. The processor 1020 may include one or more processors.

[0137] The receiver and transmitter of BS1000 are collectively referred to as transceiver 1010, and transceiver 1010 can send signals to UE or network entity or receive signals from UE or network entity. The signal to be sent to UE or network entity or received from UE or network entity may include control information and data. To this end, transceiver 1010 may include an RF transmitter for up-converting the frequency of the signal to be sent and amplifying the signal and an RF receiver for low-noise amplification of the received signal and down-converting the received signal. However, this is only an example of transceiver 1010, and the elements of transceiver 1010 are not limited to RF transmitter and RF receiver.

[0138] The transceiver 1010 may perform a function for transmitting and receiving signals on a wireless channel. For example, the transceiver 1010 may receive a signal on a wireless channel, output the signal to the processor 1020, and transmit a signal output from the processor 1020 on a wireless channel.

[0139] The memory 1030 may store programs and data required for the operation of the BS 1000, may store control information or data included in a signal obtained by the BS, and may include a storage medium such as a read-only memory (ROM), a random access memory (RAM), a hard disk, a compact disk ROM (CD-ROM), and a digital versatile disk (DVD) or a combination of storage media. Alternatively, the memory 1030 may not exist separately but may be integrated into the processor 1020. The memory 1030 may include a volatile memory, a nonvolatile memory, or a combination of a volatile memory and a nonvolatile memory, and may provide stored data at the request of the processor 1020.

[0140] The processor 1020 may control a series of steps performed by the BS 1000. For example, the processor 1020 may receive a control signal and a data signal through the transceiver 1010, and process the received control signal and data signal. The processor 1020 may send the processed control signal and data signal through the transceiver 1010. The processor 1020 may record data to the memory 1030 or read data from the memory 1030. The processor 1020 may perform the function of the protocol stack requested by the communication standard, for which the processor 1020 may include at least one processor or microprocessor. The transceiver 1010 and a portion of the processor 1020 may be referred to as a communication processor (CP).

[0141] The processor 1020 may include one or more processors, including a general-purpose processor such as a CPU, an AP, a digital signal processor (DSP), a graphics processing unit (GPU), a visual processing unit (VPU), etc., or a dedicated AI processor such as a neural processing unit (NPU). For example, when one or more processors are dedicated AI processors, the dedicated AI processor may be designed in a hardware structure dedicated to processing a specific AI model.

[0142] The processor 1020 can instruct the UE to collect CSI data (V data) through the transceiver 1010, request the UE to send a CSI data set (V data set) through the transceiver 1010, receive the CSI data set (V data set) from the UE through the transceiver, and use the received CSI data set (V data set) to update the AE.

[0143] Fig.11 is a schematic block diagram of a UE 1100 according to an embodiment.

[0144] refer to Fig.11 , UE 1100 may include a transceiver 1110, a processor 1120, and a memory 1130. However, the elements of UE 1100 are not limited thereto. For example, UE 1100 may include more or fewer elements than those described above. Transceiver 1110, processor 1120, and memory 1130 may be implemented in a single chip.

[0145] Processor 1120 may include one or more processors. The one or more processors may include a general-purpose processor (such as a CPU, an AP, a digital signal processor (DSP), etc.), a GPU, a visual processing unit (VPU), etc., or a dedicated AI processor (such as a neural processing unit (NPU)). For example, when one or more processors are dedicated AI processors, the dedicated AI processor may be designed as a hardware structure dedicated to processing a specific AI model.

[0146] The processor 1120 may control a series of steps for the UE 1100 to perform. For example, the processor 1120 may receive a control signal and a data signal through the transceiver 1110, and process the received control signal and data signal. The processor 1120 may send the processed control signal and data signal through the transceiver 1110, may control the processing of input data derived from the received control signal and data signal according to a predefined operation rule or an AI model stored in the memory 1130, may record data to the memory 1130 or read data from the memory 1130, and may perform the functions of the protocol stack requested by the communication standard. The processor 1120 may include at least one processor, and a portion of the transceiver 1110 or the processor 1120 may be referred to as a CP.

[0147] The memory 1130 may store programs and data required for the operation of the UE 1100. In addition, the memory 1130 may store control information or data included in a signal obtained by the UE 1100, may store a predefined operation rule or an AI model used by the UE 1100, may include a storage medium such as a ROM, RAM, a hard disk, a CD-ROM, and a DVD, or a combination of storage media, may not exist separately but may be integrated into the processor 1120, may include a volatile memory, a nonvolatile memory, or a combination of a volatile memory and a nonvolatile memory, and may provide stored data at the request of the processor 1120.

[0148] The transceiver 1110 may refer to a transmitter and a receiver, and the transceiver 1110 of the UE 1100 may send a signal to a BS or a network entity or receive a signal from a BS or a network entity. The signal may include control information and data. To this end, the transceiver 1110 may include an RF transmitter for up-converting the frequency of the signal to be sent and amplifying the signal and an RF receiver for low-noise amplification of the received signal and down-converting the received signal. The elements of the transceiver 1110 are not limited to the RF transmitter and the RF receiver. In addition, the transceiver 1110 may receive a signal on a wireless channel, output the signal to the processor 1120, and send a signal output from the processor 1120 on a wireless channel.

[0149] According to the embodiments of the present disclosure, the AI-related functions are operated by a processor and a memory. One or more processors may exist. One or more processors may include a general-purpose processor (such as a CPU, AP, digital signal processor (DSP), etc.), a GPU, a visual processing unit (VPU), etc., or a dedicated AI processor (such as a neural processing unit (NPU)). One or more processors may control the processing of input data according to predefined operating rules or AI models stored in the memory. When one or more processors are dedicated AI processors, they can be designed as a hardware structure dedicated to processing a specific AI model.

[0150] The predefined operating rules or AI models can be made by learning, so that when the basic AI model (or deep learning model) is trained by a learning algorithm with a large amount of training data, the predefined operating rules or AI models established for executing the desired features (or objects) are made. This learning can be performed by the device in which the AI ​​is executed, or by a separate server and / or system. Examples of learning algorithms may include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning.

[0151] The AI ​​model (or deep learning model) may be composed of multiple neural network layers. Each of the multiple neural network layers may have multiple weight values, and the neural network operation may be performed by the operation between the operation result of the previous layer and the multiple weight values. The multiple weight values ​​owned by the multiple neural network layers may be optimized by the learning results of the AI ​​model. For example, multiple weight values ​​may be updated to reduce or minimize the loss value or cost value obtained by the AI ​​model during the training procedure. The artificial neural network may include a deep neural network (DNN), such as a convolutional neural network (CNN), a recursive neural network (RNN), a restricted Boltzmann machine (RBM), a deep belief network (DBN), a bidirectional recursive DNN (BRDNN), or a deep Q network, but is not limited thereto.

[0152] Various embodiments of the present disclosure can be implemented or supported by one or more computer programs, which are formed by computer-readable program code and can be embodied in computer-readable media. Throughout the specification, the terms "application" and "program" can refer to one or more computer programs, software components, instruction sets, processes, functions, objects, classes, examples, associated data or a part thereof that are properly implemented in computer-readable program code. Computer-readable program code can include various types of computer codes, including source code, object code and executable code. Computer-readable media can include various types of media that can be accessed by a computer, such as ROM, RAM, hard disk drive (HDD), compact disk (CD), digital video disk (DVD) or other various types of memory.

[0153] Computer-readable storage media may be provided in the form of non-transitory storage media. Non-transitory storage media are tangible devices that may exclude wired, wireless, optical, or other communication links for sending temporary electrical or other signals. Non-transitory storage media do not distinguish between when data is semi-permanently stored in the storage medium and when data is temporarily stored in the storage medium. For example, a non-transitory storage medium may include a buffer for temporarily storing data. Computer-readable media may be any available media that can be accessed by a computer, including volatile, non-volatile, removable, and non-removable media. Computer-readable media include media for permanent storage of data, as well as media for storing data that can be rewritten later, i.e., rewritable optical disks or erasable memory devices.

[0154] The methods disclosed herein may be provided in a computer program product, such as a commercial product that can be traded between a seller and a buyer. The computer program product may be distributed in the form of a storage medium, through an application store, directly between two user devices (e.g., smart phones), or online (e.g., download or upload). In online distribution, at least a portion of the computer program product (e.g., a downloadable app) may be at least temporarily stored or arbitrarily created in a storage medium that may be readable by a device such as a manufacturer's server, an application store's server, or a relay server.

[0155] Several embodiments have been described, but those of ordinary skill in the art will understand and appreciate that various modifications may be made without departing from the scope of the present disclosure. For example, the aforementioned methods may be performed in a different order, and / or the aforementioned systems, structures, devices, circuits, etc. may be combined in a different combination than described above, or replaced or substituted by other components or their equivalents to obtain appropriate results. Therefore, it is apparent to those of ordinary skill in the art that the present disclosure is not limited to the described embodiments, but may cover not only the attached claims but also equivalents. For example, an element described in a singular form may be implemented as a distributed, and an element described in a distributed form may be implemented as a combination.

[0156] While the disclosure has been particularly shown and described with reference to certain embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the subject matter defined by the appended claims and their equivalents.

Claims

1. A method for updating an auto encoder (AE) for channel state information (CSI) feedback performed by a base station (BS) in a wireless communication system, the method comprising: sending a data collection instruction message, wherein the data collection instruction message instructs a user equipment (UE) to collect data before compression corresponding to the compressed data; receiving data before compression from the UE; as well as An autoencoder (AE) is updated based on the received pre-compression data.

2. The method according to claim 1, in, The AE includes a first encoder artificial intelligence (AI) model configured to compress channel state information (CSI) data and a decoder AI model configured to reconstruct the compressed CSI data, and Wherein, updating AE includes fine-tuning at least one of the first encoder AI model and the decoder AI model.

3. The method according to claim 2, further comprising: AE update start time information, updated duration information, and identification information of a second encoder AI model to be used during a time period in which the update of the AE is performed are sent to the UE.

4. The method according to claim 2, further comprising: AE update start time information, updated duration information, and identification information of a codebook to be used during a time period in which the update of the AE is performed are transmitted to the UE.

5. The method according to claim 1, in, When it is determined based on the cell-specific information that there is a change in the channel characteristics, a data collection instruction message is sent to the UE.

6. The method according to claim 1, in, The data collection instruction message includes an indicator including identification information of the data before compression to be collected, The data collection instruction message is sent to the UE in a physical downlink control channel (PDCCH).

7. The method according to claim 1, further comprising: Determine the data collection interval based on the degree of change of channel characteristics over time, The data collection instruction message includes data collection interval information, and The data collection instruction message is sent to the UE via radio resource control (RRC) signaling.

8. The method according to claim 1, in, The data collection instruction message includes a data transmission request message requesting the UE to transmit the collected pre-compression data immediately after the pre-compression data is collected.

9. The method according to claim 1, wherein: The data collection instruction message includes a data storage instruction message, wherein the data storage instruction message instructs the UE to store the collected data before compression, and The method further includes sending a data transmission request message, wherein the data transmission request message requests the UE to send at least one piece of data before compression.

10. A base station (BS) for updating an auto encoder (AE) for channel state information (CSI) feedback in a wireless communication system, the BS comprising: Transceiver; and at least one processor, Wherein, the at least one processor is configured to: sending a data collection instruction message to a user equipment (UE) through a transceiver, wherein the data collection instruction message instructs the UE to collect data before compression corresponding to the compressed data, receiving the data before compression from the UE by the transceiver, and An autoencoder (AE) is updated based on the received pre-compression data.

11. A method for updating an auto encoder (AE) for channel state information (CSI) feedback performed by a user equipment (UE) communicating with a base station (BS) in a wireless communication system, the method comprising: receiving a data collection instruction message from a base station (BS), the data collection instruction message instructing the UE to collect data before compression corresponding to the compressed data; Create compressed data; sending the data before compression to the BS; and An auto encoder (AE) updated based on data before compression is received from the BS.

12. The method according to claim 11, in, The AE includes a first encoder artificial intelligence (AI) model configured to compress channel state information (CSI) data and a decoder AI model configured to reconstruct the compressed CSI data, and Wherein, the updated AE includes at least one of a fine-tuned first encoder AI model and a fine-tuned decoder AI model.

13. The method according to claim 11, in, When it is determined based on the cell-specific information that there is a channel characteristic change, a data collection instruction message is received from the BS.

14. The method according to claim 11, in, The data collection instruction message includes a data transmission request message requesting to transmit the collected data before compression to the BS immediately after the data before collection is collected.

15. The method according to claim 11, wherein: The data collection instruction message includes a data storage instruction message, wherein the data storage instruction message instructs storing the collected data before compression, and The method further includes receiving a data transmission request message from the BS, wherein the data transmission request message requests to send at least one piece of data before compression to the BS.

Images