Apparatus and method for generating channel state information to which machine learning-based codebook restriction is applied in wireless communication system

By employing machine learning techniques like auto-encoders and CNNs for CSI compression and codebook restriction transmission, the method addresses inefficiencies in CSI reporting and codebook restriction application, enhancing wireless communication system performance through reduced overhead and improved accuracy.

WO2024237669A1PCT designated stage expired Publication Date: 2024-11-21ELECTRONICS & TELECOMM RES INST
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
PCT/KR2024/006585
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2024-05-14
Filing Date
2024-05-14
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

Current wireless communication systems face inefficiencies in channel state information (CSI) reporting due to high overhead, especially in expressing precise channel changes for precoding, and struggle with applying machine learning-based codebook restrictions to limit specific CSI feedback effectively.

Method used

The proposed solution involves using machine learning-based methods, specifically auto-encoder neural networks and CNNs, to compress and restore wireless channel information, along with codebook restriction information transmission from the base station to the terminal, allowing for efficient CSI feedback generation and reporting that excludes unwanted beam directions, utilizing encoder and decoder models to support codebook restrictions.

Benefits of technology

This approach reduces the amount of information needed for CSI transmission, enhances the accuracy of channel state information reporting, and allows for effective application of codebook restrictions, improving system performance by minimizing overhead and clarifying precoding information.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure KR2024006585_21112024_PF_FP_ABST
    Figure KR2024006585_21112024_PF_FP_ABST
Patent Text Reader

Abstract

The present invention relates to an apparatus and a method for generating channel state information to which a machine learning-based codebook restriction is applied in a wireless communication system. An operation method of a base station (BS) for feeding back channel state information by using artificial intelligence / machine learning (AI / ML) in a wireless communication system comprises the steps of: transmitting codebook restriction information to a user equipment (UE), wherein the codebook restriction information is upper layer information; and receiving, from the UE, feedback on restricted channel state information generated on the basis of the codebook restriction information.
Need to check novelty before this filing date? Find Prior Art

Description

Device and method for generating channel state information using machine learning-based codebook restrictions in a wireless communication system

[0001] The present disclosure relates generally to wireless communication systems, and more particularly to a device and method for generating channel state information using machine learning-based codebook constraints in a wireless communication system.

[0002] The International Telecommunication Union (ITU) is developing an International Mobile Telecommunication (IMT) framework and standards, and is currently discussing 6th generation (6G) communications through a program called "IMT for 2030 and beyond."

[0003] Among the technologies for implementing 6G, one area that is receiving significant attention is artificial intelligence (AI).

[0004] 3GPP began research on AI / ML technologies for the air interface in Release 18. Key use cases for this research include (1) AI / ML for CSI feedback enhancement, (2) AI / ML for beam management, and (3) AI / ML for positioning performance enhancement.

[0005] The present disclosure is highly relevant to the first use case for improving performance for CSI feedback.

[0006] More specifically, in mobile communication networks, a transmitter performs data signal coding levels, power allocation, and beamforming using multiple transmit antennas to transmit data to a receiver. To this end, the transmitter must obtain information about the wireless channel between the transmitter and receiver antennas. However, since the channel from the transmitter to the receiver cannot be directly observed at the transmitter, a Channel State Information (CSI) reporting procedure is required, which is a procedure for reporting channel information measured by the receiver to the transmitter. CSI is information for scheduling data transmission from the transmitter to the receiver and includes rank, channel quality index, and precoding information.

[0007] To measure the channel status at the receiver, a reference signal such as a CSI-Reference Signal (CSI-RS) is designed. The transmitter transmits the CSI-RS periodically or aperiodically, and configures transmission-related information in advance so that the receiver can receive it. After the receiver receives the CSI-RS, a CSI reporting procedure is performed to generate CSI and transmit it back to the transmitter. At this time, in order to precisely express the channel information, the amount of information must be very large, which increases the occupation of wireless transmission resources and overhead, and becomes a factor that reduces system performance. In particular, precisely expressing the channel information for expressing the channel change for determining the precoding at the transmitter or the precoding information for recommending an appropriate precoding vector at the receiver causes a large overhead.

[0008] To address these issues, research has begun on a technology that can acquire channel state information from a transmitter with high accuracy while minimizing the amount of transmitted information using machine learning (ML) technology. Discussions are now underway to apply this technology to 5G and beyond mobile communication systems. To this end, an autoencoder-based neural network was proposed as a machine learning structure for transmitting channel information. Furthermore, a convolutional neural network (CNN)-based artificial neural network was proposed that inputs wireless channel information in image form, compresses it into a code vector in a low-dimensional latent space through an encoder network, and then reconstructs the original wireless channel information through a decoder network. This network demonstrated effective compression and decompression. However, the aforementioned technology transmits the entire channel information, resulting in a large amount of information to be transmitted. Furthermore, because the compressed low-dimensional code vector has real values, additional considerations such as quantization must be taken into account when transmitting information from the receiver to the transmitter in an actual system.

[0009] In addition, in the existing legacy channel state information reporting technique, there are cases where the base station restricts beam forming in a specific direction to prevent interference or coexist with other communication networks, and the operation for this exists as a codebook subset restriction technique.

[0010] Furthermore, compared to conventional non-ML-based feedback techniques, general machine learning-based channel state information techniques have the characteristic that the precoding information (PMI) referred to by the channel state information feedback information is unclear. Therefore, it is difficult to apply a technique to restrict specific channel state information feedback information in a one-to-one correspondence in conventional non-ML-based feedback techniques. For example, the precoding information referred to by feedback with all bits set to 0 is expressed differently depending on the machine learning model or model pair. Therefore, it is difficult to derive the feedback information to be restricted at the base station, such as requiring a terminal-side encoder model to derive feedback information expressing the beam direction that the base station wants to restrict. Furthermore, when the feedback length is long, expressing the feedback to be restricted as bitmap information requires a bitmap with a length that is base 2 and an exponent, making it difficult to express and convey in reality.

[0011] Based on the discussion described above, the present disclosure provides a device and method for reporting channel state information by applying machine learning in a wireless communication system.

[0012] In addition, the present disclosure provides a device and method for generating channel state information by applying machine learning-based codebook restrictions in a wireless communication system.

[0013] In addition, the present disclosure provides a device and method for generating and reporting CSI feedback excluding a beam direction not preferred by a base station by using an encoder and decoder machine learning model on the terminal and base station sides in a wireless communication system.

[0014] According to various embodiments of the present disclosure, a method of operating a base station (BS) for feeding back channel state information using AI / ML (artificial intelligence / machine learning) in a wireless communication system includes the steps of transmitting codebook restriction information to a user equipment (UE), wherein the codebook restriction information is upper layer information, and receiving feedback on restricted channel state information generated based on the codebook restriction information from the UE.

[0015] According to various embodiments of the present disclosure, a method of operating a user equipment (UE) for feeding back channel state information using AI / ML (artificial intelligence / machine learning) in a wireless communication system includes the steps of receiving codebook restriction information from a base station (BS), wherein the codebook restriction information is upper layer information, and transmitting feedback on restricted channel state information generated based on the codebook restriction information to the base station.

[0016] According to various embodiments of the present disclosure, a base station (BS) for feeding back channel state information using AI / ML (artificial intelligence / machine learning) in a wireless communication system includes at least one transceiver and at least one control unit operably connected to the at least one transceiver, wherein the control unit transmits codebook restriction information to a user equipment (UE), the codebook restriction information is upper layer information, and can receive feedback on restricted channel state information generated based on the codebook restriction information from the UE.

[0017] According to various embodiments of the present disclosure, a user equipment (UE) for feeding back channel state information using AI / ML (artificial intelligence / machine learning) in a wireless communication system includes at least one transceiver, and at least one control unit operably connected to the at least one transceiver, wherein the at least one control unit receives codebook restriction information from a base station (BS), the codebook restriction information is upper layer information, and can transmit feedback on restricted channel state information generated based on the codebook restriction information to the base station.

[0018] Devices and methods according to various embodiments of the present disclosure enable a base station and a terminal to perform CSI feedback using a machine learning model, thereby additionally expressing whether the machine learning model supports codebook restriction information.

[0019] In addition, the device and method according to various embodiments of the present disclosure provide a method for learning a machine learning model to efficiently support codebook restrictions, thereby learning a machine learning model capable of efficiently supporting codebook restriction information to be applied, or additionally learning a learned machine learning model.

[0020] The effects that can be obtained from the present disclosure are not limited to the effects mentioned above, and other effects that are not mentioned can be clearly understood by a person having ordinary skill in the art to which the present disclosure belongs from the description below.

[0021] FIG. 1 illustrates an example of a vector group according to one embodiment of the present disclosure.

[0022] FIG. 2 illustrates an example of a beam limitation region in the horizontal and vertical directions, according to one embodiment of the present disclosure.

[0023] FIG. 3 illustrates an example of a limited channel state information feedback generation procedure according to one embodiment of the present disclosure.

[0024] FIG. 4 illustrates an example of a channel state information generation AI model that additionally receives codebook restriction information according to one embodiment of the present disclosure.

[0025] FIG. 5 illustrates an example of generating limited channel state information using preprocessing according to one embodiment of the present disclosure.

[0026] FIG. 6 illustrates an example of a support codebook restriction bitmap configuration according to one embodiment of the present disclosure.

[0027] FIG. 7 illustrates an example of a support codebook restriction proxy model configuration according to one embodiment of the present disclosure.

[0028] FIG. 8 illustrates an example of a codebook support information report according to one embodiment of the present disclosure.

[0029] FIG. 9 illustrates another example of a codebook support information report, according to one embodiment of the present disclosure.

[0030] FIG. 10 illustrates an example of a final channel state information derivation operation at a base station according to an embodiment of the present disclosure.

[0031] FIG. 11 illustrates an example of a final channel state information derivation operation based on AI model information of a terminal according to an embodiment of the present disclosure.

[0032] FIG. 12 illustrates an example for deriving final channel state information using a decoder AI model at a base station according to one embodiment of the present disclosure.

[0033] FIG. 13 illustrates an example for deriving final channel state information using post-processing at a base station according to one embodiment of the present disclosure.

[0034] FIG. 14 illustrates an example of an AI model learning method for efficient codebook limitation, according to one embodiment of the present disclosure.

[0035] FIG. 15 illustrates a learning method of a proxy model according to one embodiment of the present disclosure.

[0036] FIG. 16 illustrates a configuration diagram of a base station in a wireless communication system according to various embodiments of the present disclosure.

[0037] FIG. 17 illustrates a configuration diagram of a terminal in a wireless communication system according to various embodiments of the present disclosure.

[0038] The terms used in this disclosure are used only to describe specific embodiments and may not be intended to limit the scope of other embodiments. The singular expression may include plural expressions unless the context clearly indicates otherwise. Terms used herein, including technical or scientific terms, may have the same meaning as commonly understood by those of ordinary skill in the art described in this disclosure. Terms defined in general dictionaries among the terms used in this disclosure may be interpreted as having the same or similar meaning in the context of the relevant technology, and shall not be interpreted in an idealized or overly formal sense unless explicitly defined in this disclosure. In some cases, even if a term is defined in this disclosure, it cannot be interpreted to exclude embodiments of the present disclosure.

[0039] The various embodiments of the present disclosure described below illustrate a hardware-based approach as an example. However, since the various embodiments of the present disclosure include techniques utilizing both hardware and software, the various embodiments of the present disclosure do not exclude a software-based approach.

[0040] Additionally, in the detailed description and claims of the present disclosure, “at least one of A, B, and C” can mean “only A,” “only B,” “only C,” or “any combination of A, B, and C.” Additionally, “at least one of A, B, or C” or “at least one of A, B, and / or C” can mean “at least one of A, B, and C.”

[0041] The present disclosure relates to a device and method for generating channel state information (CSI) using machine learning-based codebook constraints in a wireless communication system. Specifically, the present disclosure describes a technique for transmitting channel state information (CSI) from a receiver to a transmitter in a wireless communication system by applying machine learning techniques, thereby utilizing CSI for link adaptation and scheduling. More specifically, the present disclosure relates to an artificial neural network for generating and transmitting CSI using codebook constraints, as well as an operating method and device therefor.

[0042] An example of a receiver in the present disclosure may be a user equipment (UE), and an example of a transmitter may be a base station (BS). The transmitter and receiver may be configured in various devices depending on the network configuration.

[0043] [Method for Creating and Transmitting Restricted Channel Status Information]

[0044] [Composition 1]

[0045] The present disclosure proposes a method for transmitting channel state information (CSI) based on machine learning in a mobile communication system comprising a base station and a terminal, including a step of the base station transmitting codebook restriction information to the terminal in order to restrict channel state information generated and reported by the terminal, and the terminal generating restricted channel state information feedback based on the codebook restriction information and transmitting the same to the base station.

[0046] In one embodiment, an example of how to express codebook restriction information is as follows, and it is also possible to express it in any other arbitrary form below.

[0047] (1) Limited vector information based on vector groups (power limit information for specific vector components)

[0048] The restricted vector information based on vector groups is the same as that used to express the subset constraints of the Type2 and Enhanced Type2 codebooks of the existing 3GPP NR technology.

[0049] (2) Horizontal and vertical beam direction limitation information

[0050] FIG. 1 illustrates an example of a vector group according to one embodiment of the present disclosure.

[0051] Referring to Fig. 1, horizontal and vertical beam direction restriction information indicates a restricted area by displaying horizontal and vertical two-dimensional beam direction information as tiles. For example, as shown in Fig. 1, when a beam is restricted between 0 and 90 degrees in the horizontal direction and between 0 and 90 degrees in the vertical direction, this area is indicated as a restricted area.

[0052] FIG. 2 illustrates an example of a beam limitation region in the horizontal and vertical directions, according to one embodiment of the present disclosure.

[0053] Referring to Fig. 2, an example of a method for expressing beam restriction information in the horizontal and vertical directions may be bitmap information that divides a two-dimensional plane into grids, and expresses restricted areas of the entire grid as 1 and non-restricted areas as 0. For example, as in Fig. 2, if the horizontal and vertical directions are divided into four, the restricted areas can be expressed as a 16-bit bitmap.

[0054] FIG. 3 illustrates an example of a limited channel state information feedback generation procedure according to one embodiment of the present disclosure.

[0055] Referring to Figure 3, the base station can transmit codebook restriction information to the terminal. That is, the codebook restriction information can be transmitted to the terminal before performing a channel state information feedback request (301). For example, according to one embodiment, it can be transmitted to the terminal as upper layer information during the RRC connection phase.

[0056] In addition, according to another embodiment, when the codebook restriction information when performing channel state information feedback using machine learning is the same as the codebook restriction information when performing non-ML-based (legacy) channel state information feedback, the codebook restriction information in the non-ML-based channel state information feedback can be indicated through upper layer information. Through this, the n1-n2-codebookSubsetRestriction parameter for 3GPP NR Type 2 feedback or the n1-n2-codebookSubsetRestriction-r16 parameter for 3GPP NR Enhanced Type 2 feedback can be used as is to configure the codebook restriction information for machine learning.

[0057] In one embodiment of the present disclosure, the base station and the terminal operate by configuring CSI feedback using a specific machine learning model for the terminal. In addition, codebook restriction information may be set to prevent precoding in a specific beam direction at the base station. This may be conveyed through a higher layer message transmitted during the terminal's connection process. If the codebook restriction information is also set for a non-ML-based CSI feedback technique (e.g., Enhanced Type 2 feedback), this may be applied equally to machine learning-based feedback. A terminal configured with codebook restriction information in the above manner may, when performing channel state information feedback thereafter, generate constrained channel state information feedback information by referring to the information (303) and transmit the same to the base station (305).

[0058]

[0059] [Composition 1-1]

[0060] FIG. 4 illustrates an example of a channel state information generation AI model that additionally receives codebook restriction information according to one embodiment of the present disclosure.

[0061] Referring to FIG. 4, in order to generate constrained channel state information feedback based on machine learning, the present disclosure proposes to generate constrained channel state information feedback by further inputting configured codebook restriction information in addition to the original precoding vector (e.g., precoding vector) as input to the encoder machine learning model of a terminal among two-sided machine learning models. In addition, the original channel state information may be constrained channel state information generated by applying the preprocessing method according to [Configuration 1-2] of the present disclosure in advance.

[0062] In one embodiment of the present disclosure, the encoder machine learning model configured by the terminal may be a machine learning model trained to efficiently support specific codebook constraint information. For example, assuming that a feedback length of 100 bits is required to achieve a certain level of performance in an environment where codebook generation is not restricted, the model may be trained to achieve the same level of performance using a shorter feedback length (e.g., 90 bits) in an environment constrained by the codebook constraint information.

[0063] If efficient codebook constraint operation is possible, this information can be utilized in determining the feedback length at the base station and / or terminal. The terminal can input the desired channel state information to the machine learning model for generating channel state information and transmit the generated CSI feedback information to the base station. If the machine learning model requires preprocessed input for efficient codebook constraint, the raw channel state information can be input with preprocessed channel state information.

[0064]

[0065] [Composition 1-2]

[0066] FIG. 5 illustrates an example of generating limited channel state information using preprocessing according to one embodiment of the present disclosure.

[0067] Referring to FIG. 5, in order to generate constrained channel state information feedback based on machine learning, the present disclosure converts raw channel state information into constrained raw channel state information through preprocessing, and uses this as input to an encoder machine learning model to generate constrained channel state information feedback.

[0068] For example, if the codebook constraint information is defined in a form such as Type 2 or Enhanced Type 2 of 3GPP NR, the preprocessor may be a codebook generator in a form such as Type 2 or Enhanced Type 2 of 3GPP NR that can reflect this. In this case, the applied parameter (Parameter Combination) may use the highest level parameter defined in the standard specification (e.g., L=4 for Type 2, paramCombination=8 for enhanced Type 2) to enable high-resolution codebook generation, or may use a parameter at a higher level than that defined in the standard specification.

[0069] The machine learning model of the present disclosure can also use a model capable of performing efficient codebook restriction by additionally inputting codebook restriction information, as in [Configuration 1-1].

[0070]

[0071] [Composition 1-3]

[0072] In this disclosure, in order to generate constrained channel state information feedback based on machine learning, it is proposed to perform one of the following depending on the characteristics of the terminal's encoder machine learning model.

[0073] (1) Generating CSI feedback information using raw channel state information without preprocessing as input to a machine learning model.

[0074] (2) Generate CSI feedback information using preprocessed channel status information as input to a machine learning model.

[0075] (3) Generating CSI feedback information using raw channel state information and codebook restriction information without preprocessing as inputs for a machine learning model.

[0076] (4) Generating CSI feedback information using preprocessed channel status information and codebook restriction information as inputs to a machine learning model.

[0077]

[0078] The characteristics of an encoder machine learning model may consist of one or more of the following:

[0079] (1) Whether the input of codebook restriction information can be accepted

[0080] (2) Whether efficient channel status information feedback can be generated using specific codebook limitation information.

[0081] (3) Whether preprocessing is necessary for efficient channel status information feedback

[0082]

[0083] According to one embodiment of the present disclosure, in a base station and a terminal, when a specific machine learning model can generate efficient channel state information feedback using specific codebook restriction information, it may be a case where the machine learning model has performed learning or additional learning using the codebook restriction information.

[0084]

[0085] [Composition 1-4]

[0086] The encoder machine learning model in this disclosure can additionally provide information that can determine whether specific codebook restriction information is supported, and the terminal or base station can use this information to determine whether specific codebook restriction information is supported. Support for specific codebook restriction information can be provided using one of the following methods.

[0087] (1) Supported codebook restriction bitmap

[0088] (2) Supported codebook restriction proxy model

[0089] FIG. 6 illustrates an example of a support codebook restriction bitmap configuration according to one embodiment of the present disclosure.

[0090] Referring to FIG. 6, when expressing the supported codebook restriction information through a bitmap, when the codebook restriction information is expressed in 10 bits, the supported codebook restriction bitmap is a bitmap that expresses whether or not support for each configurable codebook restriction information is provided as 0 or 1, and in this case, the length of the entire bitmap can be 1024 (2^10).

[0091]

[0092] FIG. 7 illustrates an example of a support codebook restriction proxy model configuration according to one embodiment of the present disclosure.

[0093] Referring to Fig. 7, when expressing codebook restriction information through a proxy model, it refers to a model that provides specific codebook restriction information as input to the proxy model and outputs whether or not it is supported as binary information of 0 or 1, or real number information between 0 and 1. When expressing as real number information between 0 and 1, a model may output a larger real number as the efficiency level increases.

[0094]

[0095] [Composition 1-5]

[0096] In the present disclosure, it is proposed that a base station obtains whether a terminal supports specific codebook restriction information of an encoder machine learning model configured by the terminal using one or more of the following methods.

[0097] (1) Query by codebook limitation information

[0098] (2) Transmitting information on the entire support codebook limitation

[0099] In the base station and terminal of the present disclosure, the base station may need to obtain information on whether the encoder machine learning model configured by the terminal supports specific codebook constraint information. Higher-layer signaling may be used as a method for this purpose.

[0100] FIG. 8 illustrates an example of a codebook support information report according to one embodiment of the present disclosure.

[0101] Referring to FIG. 8, a base station can request a terminal to report codebook restriction information including a list of configurable codebook restriction information (801).

[0102] In response, the terminal may report codebook limitation information supported by each AI model (803). In one embodiment, the reported codebook limitation information may be a subset of the codebook limitation information queried by the base station.

[0103] FIG. 9 illustrates another example of a codebook support information report, according to one embodiment of the present disclosure.

[0104] Referring to FIG. 9, a base station can request configurable codebook restriction information from a terminal (901).

[0105] In response, the terminal may report information about the supported codebook restriction bitmap or the supported codebook restriction proxy model (903). The reported information may vary by AI model, and if multiple AI models are configured, information about each configured model may be transmitted to the base station. The base station may use the received supported codebook restriction bitmap or supported codebook restriction proxy model to determine whether any codebook restriction information for the AI ​​models configured for the terminal is supported.

[0106]

[0107] [Composition 1-6]

[0108] The present disclosure proposes to perform one of the following according to the characteristics of a decoder machine learning model of a base station in order to recover constrained channel state information based on machine learning.

[0109] (1) Generate recovered channel state information using the received CSI feedback as input to a machine learning model.

[0110] (2) After generating recovered channel status information using the received CSI feedback as input to the machine learning model, the final channel status information is generated through a post-processing process.

[0111] (3) Generate recovered channel state information using the received CSI feedback and codebook limitation information as inputs to the machine learning model.

[0112] (4) After generating recovered channel state information using the received CSI feedback and codebook limitation information as inputs to the machine learning model, the final channel state information is generated through a post-processing process.

[0113]

[0114] The characteristics of a decoder machine learning model include one or more of the following:

[0115] (1) Whether the input of codebook restriction information can be accepted

[0116] (2) Whether efficient channel state information restoration is possible using specific codebook restriction information

[0117] (3) Whether post-processing is necessary for efficient channel status information feedback

[0118]

[0119] [Composition 1-7]

[0120] In this disclosure, we propose additionally transmitting to the base station whether the channel state information feedback generated by the terminal was generated using a machine learning model with additional codebook constraint information. This information can be included in the CSI feedback information and transmitted.

[0121] FIG. 10 illustrates an example of a final channel state information derivation operation at a base station according to an embodiment of the present disclosure.

[0122] Referring to FIG. 10, in the base station and terminal of the present disclosure, the base station can receive CSI feedback and constraint information from the terminal (1001). The base station can consider whether the constrained channel state information feedback generated by the terminal was generated through a machine learning model that additionally inputs codebook constraint information (1003), and can additionally input codebook constraint information as well as the channel state information feedback to the base station-side decoder machine learning model to derive the final channel state information (1005). If the channel state information was not generated through a machine learning model, the base station can derive the channel state information based on the machine learning model that does not add codebook constraint information (1007). This operation can be performed when the base station's decoder machine learning model can accept the input of codebook constraint information and can efficiently restore the channel state information using the configured codebook constraint information.

[0123]

[0124] [Composition 1-8]

[0125] FIG. 11 illustrates an example of a final channel state information derivation operation based on AI model information of a terminal according to an embodiment of the present disclosure.

[0126] Referring to FIG. 11, the present disclosure proposes that a base station can receive CSI feedback from a terminal (1101). The base station determines whether specific codebook restriction information of an encoder-side machine learning model used by the terminal is supported (1103), and uses this to perform an operation of inferring whether the received channel state information feedback was generated through a machine learning model with additional codebook restriction information input (1105), and performs a reception operation based on this. If the base station does not support the specific codebook restriction information of the encoder-side machine learning model used by the terminal, the base station can derive channel state information based on a machine learning model without additional codebook restriction information (1107). To this end, bitmap information or a proxy model for determining whether specific codebook restriction information is supported for an encoder machine learning model configured on the terminal side can be transmitted to the base station in advance.

[0127]

[0128] [Composition 1-9]

[0129] FIG. 12 illustrates an example for deriving final channel state information using a decoder AI model at a base station according to one embodiment of the present disclosure.

[0130] Referring to FIG. 12, in order to recover channel state information, the present disclosure proposes to derive final channel state information by inputting codebook restriction information in addition to the channel state information feedback received from the base station to the decoder machine learning model.

[0131]

[0132] [Composition 1-10]

[0133] FIG. 13 illustrates an example for deriving final channel state information using post-processing at a base station according to one embodiment of the present disclosure.

[0134] In this disclosure, in order to recover channel state information, it is proposed to post-process the final channel state information derived from the base station using codebook restriction information to derive restricted final channel state information.

[0135] For example, if the codebook constraint information is defined in a form such as Type 2 or Enhanced Type 2 of 3GPP NR, the post-processor may be a codebook generator in a form such as Type 2 or Enhanced Type 2 of 3GPP NR that can reflect this. In this case, the applied parameter (Parameter Combination) may use the highest level parameter defined in the standard specification (e.g., L=4 for Type 2, paramCombination=8 for enhanced Type 2) to enable high-resolution codebook generation, or may use a parameter at a higher level than that defined in the standard specification.

[0136] Machine learning models can also use machine learning models that can additionally input codebook restriction information, as in [Configuration 1-9].

[0137]

[0138] [AI Model Training Methods for Efficient Codebook Restriction]

[0139] [Composition 2]

[0140] FIG. 14 illustrates an example of an AI model learning method for efficient codebook limitation, according to one embodiment of the present disclosure.

[0141] Referring to FIG. 14, the present disclosure proposes a method for performing training of an AI model for generating and transmitting channel state information for efficient codebook restriction by additionally inputting codebook restriction information into an encoder and / or decoder AI model, and simultaneously performing training using only input channel state information that satisfies this as a data set. A preprocessing technique may also be applied to generate a data set comprised solely of input channel state information that satisfies this.

[0142]

[0143] [Composition 2-1]

[0144] The present disclosure proposes that the encoder and / or decoder machine learning model further include bitmap information for representing codebook constraints on which learning has been performed. The bitmap information may represent the corresponding bit as 1 for the aforementioned learned specific codebook constraint, and represent the bit as 0 for the codebook constraint on which learning has not been performed.

[0145]

[0146] [Composition 2-2]

[0147] The present disclosure proposes that the encoder and / or decoder machine learning model further include a proxy model for representing the codebook constraints on which the learning was performed. The proxy model, when fed with the aforementioned learned specific codebook constraints, expresses the output as a binary number 1 between 0 and 1, or as a real number between 0 and 1, and can express larger numbers if efficiency is further enhanced through learning.

[0148]

[0149] [Composition 2-3]

[0150] FIG. 15 illustrates a learning method of a proxy model according to one embodiment of the present disclosure.

[0151] Referring to FIG. 15, the present disclosure proposes a method of performing learning by defining an additional loss function as the degree to which channel state information output by a decoder machine learning model does not satisfy specific codebook restriction information that is auxiliary input to an encoder and / or decoder AI model when performing learning of a machine learning model for generating and transmitting channel state information by applying an efficient codebook restriction.

[0152]

[0153] [Composition 2-4]

[0154] The present disclosure proposes performing additional learning on previously trained encoder or / and decoder machine learning models to efficiently support specific codebook restrictions at a base station.

[0155]

[0156] [Composition 2-5]

[0157] The present disclosure proposes a method for performing additional training of a machine learning model for generating and transmitting channel state information by applying efficient codebook constraints, by additionally inputting codebook constraint information into an encoder and / or decoder AI model, and simultaneously performing reverse learning using input channel state information that does not satisfy the constraints as a data set. Reverse learning refers to adjusting the values ​​of each artificial neural network node in the opposite direction of learning during the backpropagation process using a loss function.

[0158]

[0159] [Composition 2-6]

[0160] The present disclosure proposes to perform additional training on an encoder or / and decoder machine learning model, by requesting the terminal or terminal-side server to perform additional training, in order to efficiently support specific codebook restrictions at a base station. At this time, information about one or more specific codebook restrictions requiring support is transmitted to the terminal or terminal-side server.

[0161] FIG. 16 illustrates a configuration diagram of a base station in a wireless communication system according to various embodiments of the present disclosure. The configuration illustrated in FIG. 16 can be understood as a configuration of a base station. Terms such as "unit," "unit," etc., used hereinafter, refer to a unit that processes at least one function or operation, and may be implemented using hardware, software, or a combination of hardware and software.

[0162] Referring to FIG. 16, the base station may include a wireless communication unit (1610), a backhaul communication unit (1620), a storage unit (1630), and a control unit (1640).

[0163] The wireless communication unit (1610) can transmit and receive wireless signals via a wireless channel. For example, the wireless communication unit (1610) can perform a conversion function between baseband signals and bit streams according to the physical layer specifications of the system. Furthermore, the wireless communication unit (1610) can generate complex symbols by encoding and modulating the transmission bit stream when transmitting data. When receiving data, the wireless communication unit (1610) can restore the reception bit stream by demodulating and decoding the baseband signal.

[0164] The wireless communication unit (1610) can upconvert a baseband signal into an RF (radio frequency) band signal and transmit it through an antenna, and downconvert an RF band signal received through the antenna into a baseband signal. To this end, the wireless communication unit (1610) can include a transmission filter, a reception filter, an amplifier, a mixer, an oscillator, a digital to analog convertor (DAC), an analog to digital convertor (ADC), and the like.

[0165] The wireless communication unit (1610) may include a plurality of transmission and reception paths, and the wireless communication unit (1610) may include at least one antenna array composed of a plurality of antenna elements.

[0166] In terms of hardware, the wireless communication unit (1610) may include a digital unit and an analog unit, and the analog unit may include a plurality of sub-units depending on operating power, operating frequency, etc. The digital unit may be implemented with at least one processor (e.g., a digital signal processor (DSP)).

[0167] The wireless communication unit (1610) can transmit and receive wireless signals as described above. Accordingly, all or part of the wireless communication unit (1610) may be referred to as a "transmitter," a "receiver," or a "transceiver or transceiver." Furthermore, in the following description, transmission and reception performed via a wireless channel may include processing performed by the wireless communication unit (1610) as described above.

[0168] The backhaul communication unit (1620) may provide an interface for communicating with other nodes within the network. That is, the backhaul communication unit (1620) may convert a bit string transmitted from a base station to other nodes, such as other access nodes, other base stations, upper nodes, and core networks, into a physical signal, and may convert a physical signal received from other nodes into a bit string.

[0169] The storage unit (1630) can store data such as basic programs, application programs, and setting information for the operation of the base station. The storage unit (1630) can be composed of volatile memory, non-volatile memory, or a combination of volatile and non-volatile memory. In addition, the storage unit (1630) can provide stored data upon request from the control unit (1640).

[0170] The control unit (1640) can control the overall operations of the base station. For example, the control unit (1640) can transmit and receive signals through the wireless communication unit (1610) or the backhaul communication unit (1620). In addition, the control unit (1640) can record and read data in the storage unit (1630). In addition, the control unit (1640) can perform the functions of the protocol stack required by the communication standard.

[0171] For this purpose, the control unit (1640) may include at least one processor.

[0172] According to various embodiments of the present disclosure, the control unit (1640) can control to perform operations according to various embodiments of the base station of FIGS. 2 to 15 described below.

[0173] FIG. 17 illustrates a configuration diagram of a terminal in a wireless communication system according to various embodiments of the present disclosure. The configuration illustrated in FIG. 17 can be understood as a configuration of a terminal. Terms such as "... unit," "... unit," etc., used hereinafter, refer to a unit that processes at least one function or operation, which can be implemented using hardware, software, or a combination of hardware and software.

[0174] Referring to FIG. 17, the terminal may include a communication unit (1710), a storage unit (1720), and a control unit (1730).

[0175] The communication unit (1710) may perform functions for transmitting and receiving signals via a wireless channel. For example, the communication unit (1710) may perform a conversion function between a baseband signal and a bit stream according to the physical layer specifications of the system. For example, when transmitting data, the communication unit (1710) may generate complex symbols by encoding and modulating a transmission bit stream. When receiving data, the communication unit (1710) may restore a reception bit stream by demodulating and decoding the baseband signal. In addition, the communication unit (1710) may upconvert a baseband signal to an RF band signal and transmit it through an antenna, and downconvert an RF band signal received through the antenna to a baseband signal. For example, the communication unit (1710) may include a transmission filter, a reception filter, an amplifier, a mixer, an oscillator, a DAC, an ADC, etc.

[0176] Additionally, the communication unit (1710) may include multiple transmit / receive paths. Furthermore, the communication unit (1710) may include at least one antenna array composed of multiple antenna elements. In terms of hardware, the communication unit (1710) may be composed of digital circuits and analog circuits (e.g., a radio frequency integrated circuit (RFIC)). Here, the digital circuits and analog circuits may be implemented in a single package. Additionally, the communication unit (1710) may include multiple RF chains. Furthermore, the communication unit (1710) may perform beamforming.

[0177] The communication unit (1710) transmits and receives signals as described above. Accordingly, all or part of the communication unit (1710) may be referred to as a "transmitter," a "receiver," or a "transceiver." Furthermore, in the following description, transmission and reception performed via a wireless channel may be used to mean that the communication unit (1710) performs the processing described above.

[0178] The storage unit (1720) can store data such as basic programs, application programs, and setting information for the operation of the terminal. The storage unit (1720) can be composed of volatile memory, non-volatile memory, or a combination of volatile and non-volatile memory. In addition, the storage unit (1720) can provide stored data upon request from the control unit (1730).

[0179] The control unit (1730) can control the overall operations of the terminal. For example, the control unit (1730) can transmit and receive signals through the communication unit (1710). In addition, the control unit (1730) can record and read data in the storage unit (1720). The control unit (1730) can perform the functions of the protocol stack required by the communication standard. To this end, the control unit (1730) may include at least one processor or microprocessor, or may be a part of a processor. In addition, a part of the communication unit (1710) and the control unit (1730) may be referred to as a CP (communication processor).

[0180] According to various embodiments, the control unit (1730) can control the terminals of FIGS. 2 to 15 described above to perform operations according to various embodiments.

[0181] The terms used in the following description, including terms referring to signals, channels, control information, network entities, and device components, are provided for convenience of explanation. Therefore, the present disclosure is not limited to the terms described below, and other terms with equivalent technical meanings may be used.

[0182] Additionally, while this disclosure describes various embodiments using terminology used in certain communication standards (e.g., 3rd Generation Partnership Project (3GPP)), these are merely illustrative examples. The various embodiments of this disclosure can be easily modified and applied to other communication systems.

[0183] The methods according to the embodiments described in the claims or specification of the present disclosure may be implemented in the form of hardware, software, or a combination of hardware and software.

[0184] When implemented in software, a computer-readable storage medium storing one or more programs (software modules) may be provided. The one or more programs stored in the computer-readable storage medium are configured for execution by one or more processors within an electronic device. The one or more programs include instructions that cause the electronic device to execute methods according to the embodiments described in the claims or specification of the present disclosure.

[0185] These programs (software modules, software) may be stored in random access memory, non-volatile memory including flash memory, read only memory (ROM), electrically erasable programmable read only memory (EEPROM), magnetic disc storage devices, compact disc-ROM (CD-ROM), digital versatile discs (DVDs) or other forms of optical storage devices, magnetic cassettes, or may be stored in memories formed by a combination of some or all of these. In addition, each configuration memory may include multiple copies.

[0186] Additionally, the program may be stored on an attachable storage device that is accessible via a communication network, such as the Internet, an intranet, a local area network (LAN), a wide area network (WAN), a storage area network (SAN), or a combination thereof. Such a storage device may be connected to a device performing an embodiment of the present disclosure via an external port. Additionally, a separate storage device on the communication network may be connected to a device performing an embodiment of the present disclosure.

[0187] In the specific embodiments of the present disclosure described above, components included in the disclosure are expressed in the singular or plural form, depending on the specific embodiment presented. However, the singular or plural expressions are selected to suit the presented situation for convenience of explanation, and the present disclosure is not limited to singular or plural components. Components expressed in the plural form may be composed of singular elements, or components expressed in the singular form may be composed of plural elements.

[0188] While the detailed description of this disclosure has described specific embodiments, it should be understood that various modifications are possible without departing from the scope of this disclosure. Therefore, the scope of this disclosure should not be limited to the described embodiments, but should be defined not only by the scope of the claims described below, but also by equivalents thereof.

Claims

1. In a method for operating a base station (BS) for receiving channel status information using AI / ML (artificial intelligence / machine learning) in a wireless communication system, The process of transmitting codebook restriction information to a user equipment (UE), A method including a process of receiving feedback on restricted channel state information generated based on the codebook restriction information from the terminal.

2. In claim 1, A method further comprising a process of generating final channel state information using the decoder machine learning model of the base station, using the codebook restriction information and the decoder machine learning model of the base station, from feedback on the restricted channel state information.

3. In claim 1, A process of generating final channel state information using the decoder machine learning model of the base station as feedback on the above-mentioned restricted channel state information, A method further comprising a process of generating restricted final channel state information by post-processing the final channel state information based on the codebook restriction information.

4. In claim 1, A method further comprising the step of transmitting a codebook restriction information report request including a list of configurable codebook restriction information to the terminal.

5. In claim 4, A method further comprising a process of receiving codebook limitation information supported by each AI model from the terminal in response to the above codebook limitation information report request.

6. In claim 4, A method further comprising, in response to the above codebook restriction information report request, receiving information about a supportable codebook restriction bitmap or a supportable codebook restriction proxy model from the terminal.

7. In claim 3, if the codebook restriction information is defined in a form such as Type 2 of NR (new radio) or Enhanced Type 2, the postprocessor for the postprocessing is a codebook generator in a form such as Type 2 or Enhanced Type 2.

8. In a wireless communication system, a method for operating a user equipment (UE) to feed back channel status information using AI / ML (artificial intelligence / machine learning), The process of receiving codebook restriction information from a base station (BS), A method comprising a process of transmitting feedback on restricted channel state information generated based on the codebook restriction information to the base station.

9. In claim 8, The process of transmitting feedback on the constrained channel state information generated based on the codebook restriction information to the base station is as follows: A process of converting raw channel state information into restricted raw channel state information by preprocessing based on the above codebook restriction information, A method including a process of generating the constrained channel state information by inputting the constrained raw channel state information into an encoder machine learning model of the terminal.

10. In claim 8, The process of transmitting feedback on the constrained channel state information generated based on the codebook restriction information to the base station is as follows: A method including a process of generating the restricted channel state information by inputting the raw channel state information and the codebook restriction information into the encoder machine learning model of the terminal.

11. In claim 9, A method further comprising the step of receiving a codebook restriction information report request including a list of configurable codebook restriction information from the base station.

12. In claim 11, A method further comprising a process of transmitting codebook limitation information supported by each AI model to the base station in response to the above codebook limitation information report request.

13. In claim 11, A method further comprising, in response to the above codebook restriction information report request, transmitting information about a supportable codebook restriction bitmap or a supportable codebook restriction proxy model to the base station.

14. In claim 9, if the codebook restriction information is defined in a form such as Type 2 of NR (new radio) or Enhanced Type 2, the preprocessor for preprocessing is a codebook generator in a form such as Type 2 or Enhanced Type 2.

15. In a base station (BS) for receiving channel status information using AI / ML (artificial intelligence / machine learning) in a wireless communication system, At least one transceiver, At least one control unit operably connected to at least one transceiver, At least one control unit, Transmit codebook restriction information to the user equipment (UE), A device for receiving feedback on restricted channel status information generated based on the codebook restriction information from the terminal.

Citation Information

Patent Citations

  • Codebook subset restriction for enhanced type ii channel state information reporting

    KR102459803B1

  • Method and apparatus for codebook subset restriction for CSI reporting in advanced wireless communication systems

    US20200136702A1

  • Provision of precoder selection policy for a multi-antenna transmitter

    WO2021049984A1

  • Method and device for transmitting or receiving channel state information in wireless communication system

    WO2023287095A1

  • KR20220142442A