Codebook for ai / ML-based CSI feedback

Abstract

A codebook for artificial intelligence (AI) / machine learning (ML)-based CSI feedback is described. A communication method of user equipment (UE) therefor is characterized by comprising: receiving AI / ML-related configuration information from a network; and transmitting channel status information (CSI) feedback to the network on the basis of the configuration information, wherein the CSI feedback is presented in one mode among a first mode in which channel information is presented in the form of a bitmap, a second mode in which the channel information is presented on the basis of a beam spacing and a start position, and a third mode in which the channel information is presented on the basis of a beam spacing, a start position, and the number of beams.

Description

Codebook for AI / ML-based CSI Feedback

[0001] The following description relates to a mobile communication system utilizing AI (Artificial Intelligence) / ML (Machine Learning), specifically to a method for determining a codebook for AI / ML-based beam management and / or CSI (Channel Status Information) compression and performing communication based thereon, and to an apparatus for the same.

[0002] Various technologies such as LTE, LTE-Advanced, and WiFi are used in wireless communication systems, and 5G is also included here.

[0003] Figure 1 shows the structure of a system for 5G communication.

[0004] Referring to FIG. 1, the NG-RAN (Next Generation - Radio Access Network) may include a base station (20) that provides user plane and control plane protocol termination to the UE (10). For example, the base station (20) may include a gNB (next generation-Node B) and / or an eNB (evolved-Node B). For example, the UE (10) may be fixed or mobile and may be referred to by other terms such as terminal, MS (Mobile Station), UT (User Terminal), SS (Subscriber Station), MT (Mobile Terminal), or Wireless Device. For example, the base station may be a fixed station communicating with the UE (10) and may be referred to by other terms such as BTS (Base Transceiver System) or Access Point.

[0005] The example in FIG. 1 illustrates a case including only gNB. Base stations (20) can be connected to each other via Xn interfaces. Base stations (20) can be connected to a 5th generation core network (5G Core Network: 5GC) via NG interfaces. More specifically, base stations (20) can be connected to an access and mobility management function (AMF) (30) via an NG-C interface and to a user plane function (UPF) (30) via an NG-U interface.

[0006]

[0007] Meanwhile, starting from 5G Release 19, communication methods incorporating AI / ML are being discussed as a Work Item (WI).

[0008] Specifically, regarding the physical layer, beam management, positioning improvements, and improvements related to CSI (Channel Status Information) are being discussed, and regarding the RAN (Radio Access Network) 2, matters related to LCM (Life Cycle Management) are being discussed, but there is still a lack of discussion regarding the codebook configuration and signaling methods for the UE's CSI feedback method.

[0009] In order to solve the problem described above, one aspect of the present invention proposes a method for determining a codebook for AI / ML-based beam management and / or CSI compression and performing communication based thereon, and an apparatus for such a method. Specifically, the invention proposes an extended codebook framework that is variably applicable according to the channel environment for AI / ML-based beam management and / or CSI compression, and a signaling protocol that supports the same.

[0010] Specifically, in one embodiment of the present invention, CSI feedback is proposed to be displayed in any one of the following ways to increase CSI feedback efficiency: a first mode in which channel information is displayed in the form of a bitmap, a second mode in which channel information is displayed based on beam spacing and start position, or a third mode in which channel information is displayed based on beam spacing, start position, and number of beams.

[0011] According to an embodiment, this can correspond to classifying codebook patterns into three types: irregular, uniform, and localized, and for regular patterns, it proposes significantly reducing feedback overhead by transmitting only structural parameters (spacing, starting point, number, etc.) instead of the entire bitmap.

[0012] Furthermore, we aim to implement practical AI-based CSI feedback in next-generation Massive MIMO systems by defining a procedure in which a UE or network (base station) determines a codebook structure optimized for a trained AI model and synchronizes it with one another through RRC (Radio Resource Control) reconfiguration messages.

[0013] The technology proposed below is assumed to be applicable not only to current 5G systems but also to 6G and subsequent mobile communication systems; therefore, although the term 'AI / ML model' used in 5G may be referred to by other terms such as 'Function' or 'Agent' that perform specific functions, for the convenience of the following explanation, 5G terminology will be used.

[0014] The problems to be solved by the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below.

[0015] In one aspect of the present invention for solving the problem described above, a communication method is proposed in which a user device (UE) in a mobile communication system communicates with a network, wherein the user device receives configuration information related to AI (Artificial Intelligence) / ML (Machine Learning) from the network; and, based on the configuration information, transmits Channel Status Information (CSI) feedback to the network, wherein the CSI feedback is represented in any one of a first mode in which the channel information is represented in the form of a bitmap, a second mode in which the channel information is represented based on beam spacing and a starting position, or a third mode in which the channel information is represented based on beam spacing, a starting position, and the number of beams.

[0016] According to an embodiment, the first mode may correspond to an irregular codebook-based operation mode, the second mode to a uniform codebook-based operation mode, and the third mode to a localized codebook-based operation mode.

[0017] According to an embodiment, the channel information may be based on either a first type that transmits the CSI feedback information based on a codebook or a second type that transmits the CSI feedback information without being based on a codebook.

[0018] In addition, the AI / ML related configuration information may be received through a Radio Resource Control (RRC) message representing one or more of the first mode, the second mode, or the third mode.

[0019] At this time, the CSI feedback information may be based on one or more of the first type method that determines any one of the first mode, the second mode, or the third mode based on a determination of the channel environment in the UE based on the information of the RRC message, or the second type method that specifies any one of the first mode, the second mode, or the third mode in the information of the RRC message.

[0020] Meanwhile, the above CSI feedback may be information for beam management of the network, and in this case, the RRC message may include information representing one or more of the first mode, the second mode, or the third mode as information for Set B representing a measurement target for input or learning of an AI / ML model.

[0021] According to an embodiment, the AI / ML related configuration information may include assignment information for different sets B for each of a plurality of UEs for MU-MIMO (Multi-User Multiple Input Multiple Output) communication, and the CSI feedback may include information for the set B assigned to the UE.

[0022] At this time, the AI / ML-related configuration information may additionally include information regarding Set A, which represents the AI / ML-based estimation result beam combination, and based on the inclusion relationship between Set A and Set B, it may be determined whether to omit information regarding either Set A or Set B.

[0023] On the other hand, the above CSI feedback may be information for CSI compression, and in this case, the AI / ML-related configuration information may include a Pairing ID that distinguishes the pairing of the UE-side model and the network-side model.

[0024] At this time, it may additionally include transmitting a message to the network that transmits UE Capability Info, which includes information about one or more UE-side models that the UE can support, and the AI / ML-related configuration information including the pairing ID may be received from the network via an RRC reconfiguration message based on the UE Capability Info.

[0025] In addition, the AI / ML related configuration information may additionally include parameter combination information that defines the output message of the UE-side model corresponding to the pairing ID, and the parameter combination information may include two or more parameter combination information among (i) the number of real number items (d) of the output message, (ii) the length of the segment (L), and (iii) a quantization coefficient (Q) corresponding to the number of bits for quantization of the segment.

[0026] In addition, the CSI feedback information can be represented in any one of the first to third modes through the output message defined according to the parameter combination information.

[0027] Meanwhile, in another aspect of the present invention, a communication method is proposed in which a network in a mobile communication system communicates with a user device (UE), wherein the network transmits configuration information related to AI (Artificial Intelligence) / ML (Machine Learning) to the UE; and receives Channel Status Information (CSI) feedback from the UE based on the configuration information, wherein the CSI feedback is expressed in any one of a first mode in which the channel information is expressed in a bitmap form, a second mode in which the channel information is expressed based on beam spacing and starting position, or a third mode in which the channel information is expressed based on beam spacing, starting position, and number of beams.

[0028] In addition, in another aspect of the present invention, a user device (UE) that communicates with a network in a mobile communication system comprises: at least one processor; and at least one computer memory that can be operably connected to the at least one processor and stores instructions that cause the at least one processor to perform operations when executed, wherein the operations include receiving AI (Artificial Intelligence) / ML (Machine Learning) related configuration information from the network; and transmitting CSI (Channel Status Information) feedback to the network based on the configuration information, wherein the CSI feedback is represented in any one of a first mode representing channel information in the form of a bitmap, a second mode representing channel information based on beam spacing and start position, or a third mode representing channel information based on beam spacing, start position, and number of beams.

[0029] In addition, in another aspect of the present invention, a network for communicating with a plurality of user devices (UEs) in a mobile communication system is proposed, comprising: at least one processor; and at least one computer memory that can be operably connected to the at least one processor and stores instructions that cause the at least one processor to perform operations when executed, wherein the operations include transmitting AI (Artificial Intelligence) / ML (Machine Learning) related configuration information to the UE; and receiving CSI (Channel Status Information) feedback from the UE based on the configuration information, wherein the CSI feedback is represented in any one of a first mode representing channel information in the form of a bitmap, a second mode representing channel information based on beam spacing and start position, or a third mode representing channel information based on beam spacing, start position, and number of beams.

[0030] According to the embodiments of the present invention as described above, communication environments can be responded to more flexibly by using an extended codebook framework that is variably applicable according to the channel environment for AI / ML-based beam management and / or CSI compression.

[0031] According to the embodiment, by classifying codebook patterns into three types—irregular, uniform, and localized—for regular patterns, only structural parameters (spacing, starting point, number, etc.) are transmitted instead of the entire bitmap, thereby drastically reducing feedback overhead.

[0032] Furthermore, by defining a procedure in which a UE or network (base station) determines a codebook structure optimized for a trained AI model and mutually synchronizes it through RRC reconstruction messages, practical AI-based CSI feedback can be implemented in next-generation Massive MIMO systems.

[0033] In addition, by presenting an extended codebook structure that includes existing non-AI / ML codebooks, technical compatibility is ensured, and a foundation can be provided for future expansion to various AI / ML-based wireless technologies such as 6G.

[0034] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below.

[0035] Figure 1 shows the structure of a system for 5G communication.

[0036] FIG. 2 is a diagram illustrating a method for a UE to transmit CSI feedback information for AI / ML-based communication to a network according to an embodiment of the present invention.

[0037] FIGS. 3 to 5 are drawings for explaining codebook forms according to embodiments of the present invention.

[0038] FIG. 6 is a diagram illustrating the concept of beam management according to one embodiment of the present invention.

[0039] FIG. 7 is a diagram illustrating a method using a user-specific localized codebook based on MU-MIMO according to an embodiment of the present invention.

[0040] FIGS. 8 to 11 are drawings for explaining the configuration method of set A and set B according to embodiments of the present invention.

[0041] FIG. 12 is a diagram illustrating a concept for CSI compression according to one embodiment of the present invention.

[0042] Figure 13 is a diagram illustrating the pairing procedure for CSI compression of Figure 12.

[0043] FIGS. 14 and FIGS. 15 are drawings for explaining a signaling method between a UE and a network according to embodiments of the present invention.

[0044] FIG. 16 illustrates a wireless device that can be applied to the present technology.

[0045] Hereinafter, embodiments of the present invention are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0046] Throughout the specification, when a part is described as "including" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0047]

[0048] Method for Configuring CSI Feedback in AI / ML-based Communication

[0049] FIG. 2 is a diagram illustrating a method for a UE to transmit CSI feedback information for AI / ML-based communication to a network according to an embodiment of the present invention.

[0050] A UE (110) according to one embodiment of the present invention receives AI / ML related configuration information from a network (120) (S110) and can transmit CSI feedback to the network (120) based thereon (S120).

[0051] In existing 5G (NR) systems, such CSI feedback was performed based on a fixed codebook. However, the fixed codebook definitions used in existing NR systems have fundamental limitations in simultaneously satisfying various channel environments and advanced performance goals (such as high precision and low latency).

[0052] AI / ML-based beam management and / or CSI compression technologies have been introduced to overcome these limitations, but to effectively support them, it is desirable to build a new framework that breaks away from existing rigid structures.

[0053] Accordingly, in one embodiment of the present invention, we propose an augmented codebook framework that can flexibly accommodate new technical requirements, such as ultra-large scale MIMO of 6G, while maintaining backward compatibility with existing NR systems.

[0054]

[0055] For the practical commercialization of AI / ML-based CSI feedback, it is desirable to drastically reduce the signaling overhead required for feedback.

[0056] The conventional method of using the entire bitmap for codebook subset restriction (CBSR) in large-scale antenna systems causes a problem where overhead increases exponentially in proportion to the number of antennas.

[0057] To address this, technology is needed to efficiently utilize limited uplink control resources by minimizing the number of bits required for beam management and / or CSI reporting.

[0058] Furthermore, beyond simple data compression, it is important to simultaneously ensure the inference efficiency of AI / ML models and the accuracy of channel restoration.

[0059] Since the performance of AI / ML models varies depending on the UE's location or channel environment, a mechanism is required that enables the UE to actively optimize beam management and / or CSI compression structures to suit its channel characteristics.

[0060] User-customized precision feedback enables sophisticated beamforming of the network (base station), effectively supporting multi-user MIMO (Multiple Input Multiple Output) operation, thereby ensuring communication quality for individual users as well as improving the overall system yield.

[0061] To this end, in one embodiment of the present invention, such CSI feedback is proposed to be represented in any one of the following ways: a first mode representing channel information in the form of a bitmap, a second mode representing channel information based on beam spacing (e.g., μ1, μ2) and starting position (e.g., s1, s2), or a third mode representing channel information based on beam spacing (e.g., μ1, μ2), starting position (e.g., s1, s2) and number of beams (e.g., v1, v2).

[0062]

[0063] According to the embodiment, the CSI feedback information reported by the UE (110) to the network (120) may be based on either a type that transmits channel information based on a codebook for backward compatibility (hereinafter referred to as the 'first type') or a type that transmits the channel itself as CSI feedback information without being based on a codebook (hereinafter referred to as the 'second type').

[0064] For example, up to the NR system, CSI feedback information is defined in the form of Type I codebook, Type II codebook, and enhanced-Type II codebook, and it is assumed that the codebook representing the channel (e.g., W1 for a broadband beam and W2 for a narrowband beam) rather than the channel itself is fed back. In the first type method, it is proposed to maintain this codebook-based CSI feedback method while introducing a flexible, expandable codebook system to comply with AI / ML patents to perform CSI feedback.

[0065] At this time, the first mode illustrated in FIG. 2 may be referred to as an irregular codebook-based operation mode, the second mode as a uniform codebook-based operation mode, and the third mode as a localized codebook-based operation mode, and each codebook type will be explained in more detail below.

[0066] However, in the case of 6G communication methods, when prioritizing the efficiency of AI / ML-based communication over backward compatibility, the channel matrix itself can be transmitted in the form of CSI feedback rather than proposing a codebook format like this, and in this case, the form of CSI feedback transmitted by the UE (110) can be applied variably as in the first to third modes described above.

[0067]

[0068] FIGS. 3 to 5 are drawings for explaining codebook forms according to embodiments of the present invention.

[0069] Specifically, FIG. 3 shows an example of an unstructured codebook, FIG. 4 shows an example of a uniformly distributed codebook, and FIG. 5 shows an example of a locally dense codebook.

[0070] As described above, embodiments of the present invention propose utilizing a codebook divided into three codebook structures: an unstructured codebook, a uniformly distributed codebook, and a locally dense codebook. These three structures can be classified according to the patterns represented by a selected subset among the entire DFT code vectors aligned based on directionality.

[0071]

[0072] A. DFT Code Vector

[0073] 3GPP defines and uses beamforming vectors based on the Discrete Fourier Transform (DFT), where the code vector v for the l-th horizontal direction and the m-th vertical direction is l,m It can be defined as follows.

[0074] [Mathematical Formula 1]

[0075] v l,m = [u m , exp(j2πl / O1N1)*u m , ..., exp(j2π(N1-1) / O1N 1 * u m ] and

[0076] um = [1, exp (j2πl / O2N2), ..., exp(j2π(N2-1) / O2N2] if N2> 1, 1 if N2=1

[0077]

[0078] Here, N1 and N2 represent the number of antenna elements in the horizontal and vertical directions, respectively, and O1 and O2 represent the oversampling coefficients in the horizontal and vertical directions. The total code vector can be composed of Q=Q1Q2=N1N2Q1Q2, consisting of Q1=O1N2 horizontal code vectors and Q2=O2N2 vertical code vectors. Codebook C, which contains the total code vector, can be defined as follows.

[0079] [Mathematical Formula 2]

[0080] C={ v l,m│l= 0, 1, ..., N1O1-1, m = 0, 1, 000, N2O2-1}

[0081]

[0082] B. Mode 1_ Irregular Codebook

[0083] Since the code vectors consist of non-regular patterns and cannot be parameterized, it is advisable to use a bitmap indicating whether each code vector is included.

[0084] In other words, for a codebook with an irregular pattern across the entire area, since regularity in the pattern is absent, the codebook can be represented by utilizing a bitmap in the form of a bit sequence without separately defining parameters for intervals or starting positions.

[0085] A bit sequence having the same length as the total number (Q) of all code vectors is defined and used, and each bit of the bit sequence corresponds 1:1 to each code vector index within the entire grid, and the bit value (0 or 1) can indicate whether the code vector is included.

[0086] Through this, the index set S, which is an optimized subset of the codebook sub You can provide direct feedback.

[0087] That is, the codebook corresponding to mode 1 can be represented through bitmap information for the entire area instead of a separate combination of parameters.

[0088] FIG. 3 is a diagram showing an example of an unstructured codebook in (Q1, Q2) = (N1O1, N2O2) = (16, 16) defined according to (N1, N2) = (4, 4) and (O1, O2) = (4, 4). The total number of code vectors is Q = Q1Q2 = 256, and the set of code vector indices representing the unstructured codebook shown in FIG. 3 is as follows.

[0089] [Mathematical Formula 3]

[0090] Ssub = {(2, 1), (3, 9), (6, 9), (8, 5), (9, 12), (10, 4), (13, 1), (14, 9)}

[0091]

[0092] Each 2D index (l i , m i ) is i = Q1l i + m i The i-th bit of the bit sequence indicates whether it is included through the conversion formula, and the bitmap representing the unstructured codebook of FIG. 3 can be defined as follows.

[0093] [Mathematical Formula 4]

[0094] b1= 1, if i ∈ {33, 57, 105, 133, 156, 164, 209, 233}, 0 otherwise

[0095]

[0096] C. Mode 2_ Uniform Codebook

[0097] The code vectors have a form in which they are distributed with uniform density across the entire reference codebook area, and to reduce CSI reporting overhead, the spacing between code vectors in the horizontal and vertical directions and the starting position of the corresponding pattern can be defined as parameters.

[0098] As exemplified in Fig. 4, in the case of a codebook having uniform density across the entire area, the codebook can be represented by defining and utilizing parameters for the code vector spacing and starting position within the entire grid.

[0099] In this case, (μ1, μ2) can be used by defining them as code vector numerology parameters for the horizontal and vertical directions, respectively.

[0100] It satisfies 0 ≤ μ1 ≤ log2Q1 and 0 ≤ μ2 ≤ log2Q2, and through this, the spacing between adjacent horizontal and vertical code vectors is d1 = 2 μ2 , d2= 2μ1 It can be defined as.

[0101] As the code vectors are distributed with a uniform density across the entire area, the number of horizontal and vertical code vectors is q1 = Q1 / 2 respectively, based on (μ1, μ2) without additional parameters. μ2 , q2 = Q2 / 2 μ1 It can be defined as.

[0102] The starting position within the entire code vector grid is determined using separate parameters (s1, s2), where s1 and s2 represent the horizontal and vertical starting positions of the code vectors, respectively, with 0 ≤ s1 ≤ 2 μ1 -1, 0 ≤ s2 ≤ 2 μ2 -1 can be satisfied.

[0103] That is, the codebook corresponding to mode 2 can be represented through the parameters (μ1, μ2, s1, s2).

[0104] Figure 4 is a figure showing an example of a uniform distribution codebook in (Q1, Q2) = (N1O1, N2O2) = (16, 16) defined according to (N1, N2) = (4, 4) and (O1, O2) = (4, 4).

[0105] The starting positions of the horizontal and vertical code vectors satisfy s1=1 and s2=2.

[0106] With the code vector numerology (μ1, μ2) = (1, 2) applied, the horizontal and vertical spacing between code vectors is d1 = 2, respectively. μ1 = 2, , d2= 2 μ2 It is defined as = 4, and accordingly, the number of horizontal and vertical code vectors is q1=Q1 / 2, respectively. μ1 = 8, q2=Q2 / 2 μ2 It can be defined as = 4.

[0107]

[0108] D. Mode 3_ Localized Codebook

[0109] It represents a form in which code vectors are concentrated in a limited area rather than the entire area, and similar to Mode 2, in addition to the spacing between code vectors in the horizontal and vertical directions and the starting position, it can be defined by including the number of code vectors in the horizontal and vertical directions in the dense area as parameters.

[0110] A codebook with a dense distribution in a limited area can be defined in a manner similar to mode 2, but since the code vectors are distributed in a limited area rather than the entire area, it is desirable to define and use additional parameters for the number of code vectors in addition to the code vector spacing and starting position within the entire grid.

[0111] Similar to Mode 2, the horizontal and vertical code vector intervals are set to d1=2 based on the numerology parameters (μ1, μ2) for the horizontal and vertical code vectors satisfying 0 ≤ μ1 ≤ log2Q1 and 0 ≤ μ2 ≤ log2Q2, respectively. μ1 , d2=2 μ2 It can be defined as.

[0112] s1 and s2 represent the horizontal and vertical starting positions of the code vectors within the entire grid, respectively, where 0 ≤ s1 ≤ 2 μ1 -1, 0 ≤ s2 ≤ 2 μ2 -1 can be satisfied.

[0113] Additionally, the number of beams is represented using separate parameters (v1, v2), where q1 = 2 for the number of horizontal and vertical code vectors, respectively. v1 and q2= 2 v2 It is defined as such, and can satisfy 0 ≤ v1 ≤ log2Q1 - μ1 and 0 ≤ v2 ≤ log2Q2 - μ2.

[0114] That is, the codebook corresponding to mode 3 can be represented through the parameters (μ1, μ2, s1, s2, v1, v2).

[0115] Figure 5 shows an example of a locally dense codebook in (Q1, Q2) = (N1O1, N2O2) = (16, 16) defined according to (N1, N2) = (4, 4) and (O1, O2) = (4, 4).

[0116] The starting positions of the horizontal and vertical code vectors satisfy s1=7 and s2=2.

[0117] With the code vector numerology (μ1, μ2) = (1, 2) applied, the horizontal and vertical spacing between code vectors is d1 = 2, respectively. μ1 = 2, , d2= 2 μ2 It is defined as = 4, and accordingly, the number of horizontal and vertical code vectors is q1=Q1 / 2, respectively. μ1 = 8, q2=Q2 / 2 μ2 It can be defined as = 4.

[0118] By applying the code vector count parameter (v1, v2) = (2, 1), the number of horizontal code vectors is q1=2 v1 =4, the number of vertical code vectors is q2=2 v2 = It can be 2 days.

[0119]

[0120] Application Example 1 - Beam Management

[0121] FIG. 6 is a diagram illustrating the concept of beam management according to one embodiment of the present invention.

[0122] The introduction of AI / ML into beam management (BM) is being discussed in 3GPP standards, and a method for selecting the optimal beam based on L1-RSRP (Layer 1 Reference Signal Received Power) measurements of some beams is being proposed.

[0123] In beam management, Set A represents the entire set of beams, and among these beams, the optimal beam for data transmission to a specific UE (110a) can be selected (S610). On the other hand, Set B is a set consisting of some beams to be used as input to an AI / ML model, and can be configured in various ways (S620).

[0124] Depending on the location of the AI / ML model, it is divided into NW-side and / or UE-side models, and the method of operation may differ accordingly.

[0125] In the learning process, the user (UE(110a)) can measure the RSRP for set A and set B beams through beam sweeping, and then train the model by creating a dataset with the RSRP of set B beam as the input to the model and the RSRP of set A beam or the optimal beam of set A as the label.

[0126] In the inference process, after the user (UE(110b, 110c)) measures the RSRP for the set B beam through beam sweeping, the UE-side model can transmit the optimal beam to the network (120) as CSI feedback (S630) after the user (UE(110b, 110c)) selects the optimal beam through an AI / ML model.

[0127]

[0128] The NW-side model can receive feedback on RSRP from the UE (110b, 110c) and use it as input to select the optimal beam among the beams of the entire set A and inform the UE (110b, 110c) (S640).

[0129] As illustrated in FIG. 6, when the UE (110a, 110b, 110c) transmits CSI feedback for such beam management, it can transmit CSI feedback according to the first to third modes (600) described above.

[0130] That is, in the example of FIG. 6, the CSI feedback (S510) may be information for beam management of the network (120), and for this purpose, the RRC message (setting information) that the network (120) transmits to the UE (110a, 110b, 110c) may include information (600) representing one or more of the first mode, the second mode, or the third mode as information for set B representing a measurement target for input or learning of an AI / ML model.

[0131]

[0132] FIG. 7 is a diagram illustrating a method using a user-specific localized codebook based on MU-MIMO according to an embodiment of the present invention.

[0133] In one embodiment of the present invention, the AI / ML-related configuration information transmitted by the network (120) to the UE (110) is proposed to include assignment information for different sets B (710a, 710b) to each of the multiple UEs (110a, 110b) for MU-MIMO (Multi-User Multiple Input Multiple Output) communication. That is, different sets B are assigned to the first UE (110a) and the second UE (110b), and each UE performs CSI feedback including information for the set B (710a or 710b) assigned to it.

[0134] That is, as illustrated in FIG. 7, for each user (110a, 110b) in the cell, a parameter or bitmap (C) associated with the preferred set B codebook B (1) , C B (2) ) can be fed back to the network (120).

[0135] For example, if users (110a, 110b) who have selected a localized set B codebook corresponding to mode 3 are spatially sufficiently separated as shown in FIG. 7, they can operate to simultaneously transmit reference signals by utilizing the set B codebook selected by each user using the same time and frequency resources through multi-user MIMO-based multi-rank transmission.

[0136]

[0137] FIGS. 8 to 11 are drawings for explaining the configuration method of set A and set B according to embodiments of the present invention.

[0138] As described above, in beam management, Set B can be defined as a beam set for measurement, and Set A can be defined as a beam set for beam prediction using AI / ML.

[0139] Using the measurements from Set B as input to the model, information about Set A (e.g., Top-K beam indices in Set A, L1-RSRP through Set A beams, etc.) can be predicted.

[0140] Due to the relationship between Set B and Set A, Set B may be a subset of Set A, or Set B and Set A may be different. Reference numeral 810 in FIG. 8 illustrates an example where Set B is a subset of Set A, and reference numeral 820 in FIG. 8 illustrates an example where Set B is different from Set A.

[0141] As an example of a case where Set A and Set B are different, as shown in reference numeral 820 of FIG. 8, Set B may be a broadband beam (e.g., SSB (Synchronization Signal Block) beam) and Set A may be a narrowband beam (e.g., CSI-RS beam).

[0142] In addition, as shown in FIG. 9, if set B is a subset of set A, it can be expressed in various patterns depending on the ratio with set A. Also, even with a single ratio, it can be expressed in various patterns.

[0143] In addition, in one embodiment of the present invention, whether to omit information regarding either Set A or Set B may be determined based on the inclusion relationship between Set A and Set B.

[0144]

[0145] Meanwhile, Figures 10 and 11 illustrate the concept of predicting a prediction beam in the time domain to Set A based on the measurement results by Set B.

[0146] Specifically, FIGS. 10 and 11 illustrate three exemplary cases depending on the time domain arrangement of the measurement beam (set B) and the estimation beam (set A).

[0147] First, in drawing symbol 1010, time domain parameters can be defined as follows.

[0148] M t : Number of time instances to perform set beam measurements to be used as input for AI / ML

[0149] P t : Number of time instances predicting a set A beam as the output of AI / ML

[0150] Next, in reference numeral 1020, time domain parameters can be defined as follows.

[0151] X: Period [ms] during non-AI-based operation in which the UE performs set A beam measurements

[0152] Y: Period [ms] during AI-based operation in which the UE performs set B beam measurements

[0153] Next, in reference numeral 1030, time domain parameters can be defined as follows.

[0154] Y: Number of time instances of the cycle in which the UE performs Set B-beam measurements during AI-based operation

[0155] In these cases, T per represents the minimum unit representing the time instance interval.

[0156]

[0157] The aforementioned Set A and Set B can be distinguished in the form of an associated ID.

[0158] For example, according to one embodiment, the UE (110) may receive a CSI report configuration (CSI-ReportConfig) through upper-level signaling received from the network (120). In this case, the upper-level parameter reportQuantity within the CSI report configuration may be set to any one of 'none-bm', 'p-cri', 'p-cri-RSRP', 'p-ssb-index', or 'p-ssb-index-RSRP'.

[0159] Here, 'none-bm' represents a setting other than beam management, 'p-cri' represents the predicted CSI-RS resource indicator, 'p-cri-RSRP' represents the RSRP for p-cri, 'p-ssb-index' represents the predicted SSB index, and 'p-ssb-index-RSRP' represents the RSRP of the predicted SSB index.

[0160] In such a configuration environment, the network (120) may set one or two associated IDs within the CSI reporting configuration for the UE (110), and the specific operation is as follows.

[0161] First, if only the associated ID for Set A is set, the ID is associated with the resource sets of the second Resource Setting for Set A. In this case, the UE (110) can expect that all CSI-RS resources or SS / PBCH block resources existing in the resource set of the first Resource Setting associated with Set B are included in the CSI-RS resources and / or SS / PBCH block resources in the resource set of the second Resource Setting. That is, if only the associated ID for Set A is set, this may implicitly indicate that Set B is a subset of Set A.

[0162] Second, if both the associated ID for Set A (associatedIDforSetA) and the associated ID for Set B (associatedIDforSetB) are set, they may be associated with the resource set of the second resource setting and the resource set of the first resource setting, respectively. Specifically, associatedIDforSetA may be configured to correspond to the resource set of the second resource setting, and associatedIDforSetB may be configured to correspond to the resource set of the first resource setting. That is, if both the associated ID for Set A and the associated ID for Set B are set, Set B may not necessarily be a subset of Set A, and various Set B settings may be possible as described above.

[0163]

[0164] Application Example 2 - CSI Compression

[0165] FIG. 12 is a diagram illustrating a concept for CSI compression according to one embodiment of the present invention.

[0166] The top of FIG. 12 illustrates an AI / ML-based CSI compression framework, showing the process in which an encoder (1210) of the UE compresses channel state information (Original CSI) into a latent vector (1230) and a decoder (1220) of the network (base station) generates channel state information (Reconstructed CSI) as input through this. The encoder (1210) of the UE represents a UE-side AI / ML model and is assumed to form a two-sided model that forms a pair with the decoder (1220) of the network.

[0167] At this time, the dotted arrow indicates that the codebook structure is referenced as a standard for the operation and data interpretation of the encoder (1210) and decoder (1220).

[0168] The bottom of FIG. 12 illustrates a concept representing codebook patterns and feedback, classifying codebook patterns into modes 1 to 3 as described above in relation to FIG. 2 to 5, and defining a feedback method suitable for each mode.

[0169]

[0170] Figure 13 is a diagram illustrating the pairing procedure for CSI compression of Figure 12.

[0171] As illustrated in FIG. 13, the UE (110) receives a UE Capability Enquiry from the network (120) (S1310), and then transmits AI model information that can be supported by the UE (110) side through the UE Capability Information (S1320).

[0172] Additionally, the network (120) can generate a pairing ID by combining an AI model ID available in the UE (110) and an AI model ID available in the network (120), and then transmit it to the UE (110) via an RRC reconstruction message (S1330).

[0173] When using the Applicable Functionality Report optionally, the following procedures may be performed.

[0174] When the network (120) transmits an RRC reconfiguration message to the UE (110), it may also transmit an Applicable Set Config along with specific configuration information (Other Config) (S1330).

[0175] The UE (110) may also transmit a list of currently available AI model IDs (encoder IDs) within the Applicable Set Config to the network (120) as an Applicable Functionality Report (S1350).

[0176] The network (120) can generate a pairing ID including the ID of the model to be used in the UE (110) and the ID of the model to be used in the network (120) and transmit it to the UE (110) (S1350).

[0177] Accordingly, activation / deactivation / inference / monitoring of the model can be performed based on the pairing ID set therein (S1360).

[0178]

[0179] That is, for CSI compression, the network (120) can transmit a pairing ID for model pairing with the UE (110) via an RRC reconstruction message (S1330, S1350), and at this time, the network may additionally include parameter combination information that defines the form of the potential vector (output message of the UE-side model) described above in relation to FIG. 12.

[0180] That is, in an environment utilizing a two-sided model, the CSI feedback of the UE (110) can feed back the output value of the UE-side encoder model (1210) to the network (120) in the form of a latent message (1230). At this time, the 'latent message' can be viewed as a message containing the output information of the encoder-side model (1210), unlike information that has a specific meaning as in the past, and the term can be referred to in various ways.

[0181] The potential message (1230) may have the following format for a specific rank v.

[0182] [Mathematical Formula 5]

[0183] z 1 =[z 1 0, z 1 1, ...., z 1 dz l,v -1 ]

[0184] In the above [Equation 5], l represents the layer index and can have l=1, ..., v.

[0185] Also, d z l,v ≡ represents the number of real entries in a potential message, L represents the length of a segment, and Q represents the number of bits for quantization of each segment. For example, L=1 represents SQ (Scalar Quantization), and L>1 represents VQ (Vector Quantization).

[0186] Accordingly, the payload of the corresponding layer can be calculated as follows.

[0187] [Mathematical Formula 6]

[0188] d z l,v / L * Q

[0189] Accordingly, the latent messages reported by the UE for a specific layer are {d zl,v It can be defined by a combination of parameters such as {, L, Q}. However, the performance improvement of such Vector Quantization (VQ) methods may be limited compared to Scalar Quantization (SQ) methods relative to complexity. Therefore, when using the SQ method, L is fixed at 1, and the latent message is {d z l,v It can be defined by a combination of parameters such as , Q}.

[0190]

[0191] In one embodiment of the present invention, parameter combination information as described above (e.g., {d z l,v , L, Q} or {d z l,v It is proposed to represent CSI feedback information in any one of the first to third modes described above through an output message (1230) defined according to , Q}).

[0192]

[0193] Signaling method

[0194] We will examine the signaling method between the UE (110) and the network (120) to utilize the CSI feedback as described above.

[0195] Basically, AI / ML-related configuration information received by the UE (110) from the network (120) can be received via a Radio Resource Control (RRC) message (e.g., an RRC Reconfiguration message) representing one or more of the first mode, the second mode, or the third mode.

[0196] Additionally, AI / ML-based beam management and / or CSI compression technology can be classified into (1) NW-side operations and (2) UE-side operations depending on the operator and decision-maker of the model, and different signaling procedures may be required accordingly.

[0197] The network-driven method (Type 1 method) performs feedback based on a predefined fixed codebook, and the UE-driven method (Type 2 method) allows individual UEs to directly determine and report beam management and / or CSI compression structures optimized for their own channel environment.

[0198] That is, in UE-driven operation (Type 2 method), any one of the first mode, the second mode, or the third mode can be determined based on the information of the RRC message transmitted by the network (120) and the determination of the channel environment in the UE (110). In contrast, in the network-driven method (Type 1 method), operation can be performed by specifying any one of the first mode, the second mode, or the third mode from the information of the RRC message transmitted by the network (120).

[0199] According to the above-described UE-driven operation (Type 2 method), this can increase the efficiency of CSI feedback and reduce signaling overhead, but separate parameter feedback may be required to synchronize the determined structure information with the network (base station).

[0200] The additional parameters determined by the UE and reported to the network (base station) are as follows, and these can be transmitted via RRC reconfiguration requests, etc.

[0201] Bitmap defining an amorphous pattern (Mode 1)

[0202] Horizontal and vertical code vector interval parameters (Mode 2 and Mode 3)

[0203] Horizontal and vertical code vector start indices (Mode 2 and Mode 3)

[0204] Horizontal and vertical code vector count parameters (Mode 3)

[0205]

[0206] These UE-driven and network-driven operations will be explained below with reference to FIGS. 14 and FIGS. 15. For the convenience of explanation, the explanation will focus on examples of CSI compression, but it is not limited thereto and can be applied in the same way to examples of beam management described above.

[0207] FIGS. 14 and FIGS. 15 are drawings for explaining a signaling method between a UE and a network according to embodiments of the present invention.

[0208] Specifically, FIG. 14 is a flowchart illustrating a procedure for reporting and setting a UE-centric AI / ML-based CSI compression structure according to one embodiment of the present invention.

[0209] S1410. Initial Access & Configuration

[0210] The network (BS) can transmit configuration information regarding CSI reports and the entire codebook to the UE via RRC signals. This may include resource settings necessary for the UE to perform channel measurements and start model training.

[0211] S1420. Data Collection & Training

[0212] The UE can acquire channel information through CSI-RS (CSI Reference Resource) measurements. Based on this, the UE and the network (BS) train their respective AI / ML models, and if necessary, the UE can report training Target CSI or gradient information to the network (BS).

[0213]

[0214] S1430. Reporting Optimized CSI Compression Structure

[0215] After training is complete, the UE receives compression structure parameters optimized for the AI / ML model (e.g., latent message size d) via PUCCH (Physical Uplink Control Channel) or PUSCH (Physical Uplink Shared Channel). z , quantization bit count Q, structural patterns and required bitmaps or parameters for each pattern, etc.) can be proposed to the network (base station).

[0216] This process can be performed via an RRC Reconfiguration Request or related higher-level signaling.

[0217] S1440. Network Configuration & Synchronization

[0218] The network can evaluate the validity of the reported parameters to configure the CSI decomposer or select a compatible model. Subsequently, it can transmit the final confirmed structure (or model ID) information to the UE via an RRC reconstruction message to authorize its use.

[0219] S1450. Inference & Compressed CSI Generation

[0220] After approval, the UE can input the current channel measurement into the encoder to generate a latent vector, which is compressed feedback information according to the structure agreed upon in S1430 and S1440.

[0221] S1460. Final CSI Reporting

[0222] The UE can report compressed CSI feedback generated via PUCCH or PUSCH as CSI Part 2. In this case, essential higher-level information, such as Rank Indicators (RI) and Channel Quality Indicators (CQI), may be included in CSI Part 1 and transmitted together.

[0223] S1470. Channel Restoration and Data Transmission (Reconstruction & Data Transmission)

[0224] The network (base station) can restore the received compressed CSI into channel information through a decoder, calculate an optimized precoder based on this, and transmit data.

[0225]

[0226] Next, with reference to FIG. 15, a network-centric (NW-Centric) AI / ML-based CSI compression structure determination and instruction procedure according to one embodiment of the present invention will be described.

[0227] S1510. Initial Access & Configuration

[0228] The network (base station) can transmit configuration information regarding CSI reports and the entire codebook to the UE via RRC signals. This may include resource settings necessary for the UE to perform channel measurements and start model training.

[0229] S1520. Data Collection & Model Training Support

[0230] The UE can report channel information obtained through CSI-RS measurements to the network (base station) or assist in training the network (base station) model according to a predefined protocol.

[0231] Based on multiple collected channel data, the network (base station) determines the CSI compression structure and parameters most suitable for the current network (e.g., latent message size d). z Learning can be achieved by directly determining the number of quantization bits Q, structural patterns, and the necessary bitmaps or parameters for each pattern.

[0232] S1530. Determination and Configuration of Optimized CSI Compression Structure

[0233] The network (base station) may select an optimized compression structure on its own without a request from the UE. The network (base station) can instruct the UE to immediately apply the determined AI / ML model ID, compression structure parameters, and codebook configuration information by transmitting them via an RRC reconfiguration message (or MAC CE (Medium Access Control - Control Element)).

[0234] S1540 Model Synchronization and Application

[0235] The UE can perform synchronization by configuring the built-in encoder or loading the corresponding model according to the configuration information (Configuration) received from the network (base station).

[0236] After synchronization is complete, the UE can notify the network (base station) that it is ready by sending a 'RRC Reconfiguration Complete' message.

[0237] S1550. Inference and Compressed CSI Generation (Inference & Generation)

[0238] The UE can generate a latent vector by inputting the currently measured channel value into the encoder according to the structure (S1530 step setting) instructed by the network (base station).

[0239] S1560. Final CSI Reporting

[0240] The UE reports the generated compressed CSI feedback (CSI Part 2) via PUCCH or PUSCH, and it can be transmitted along with CSI Part 1 (RI / CQI).

[0241] S1570. Channel Restoration and Data Transmission (Reconstruction & Data Transmission)

[0242] The network (base station) can restore the received compressed CSI using the decoder structure it specified and transmit precoded data based on it.

[0243]

[0244] Among the two methods mentioned above, the concept of the UE-led (deterministic) codebook is explained in more detail.

[0245]

[0246] UE-led codebook

[0247] In 3GPP AI / ML-based CSI compression, it is desirable to adopt a method that generates an efficient compressed representation of channel state information using AI / ML models, rather than reporting the entire reference codebook.

[0248] The configuration method of the codebook-based structure for such compressed representation is generally defined by the base station settings and the user device's decision.

[0249] However, AI / ML model performance, such as compression efficiency and accuracy, may vary depending on the configuration of this compression structure, and the codebook structure that achieves optimal performance may differ depending on the user or channel environment.

[0250] Accordingly, the efficiency and accuracy of AI / ML-based CSI reporting can be improved by optimizing the CSI compression structure to suit the characteristics of the user's channel.

[0251] In one embodiment of the present invention, a mechanism for determining a CSI compression structure optimized for the channel environment of a UE is provided, and structural patterns capable of configuring such structure (unstructured codebook, uniformly distributed codebook, locally dense codebook), parameters for each pattern, and a method for signaling the corresponding parameters to a network (base station) are proposed.

[0252]

[0253] For AI / ML-based CSI compression, efficient channel representation structures can be defined in various forms, and a user-defined codebook structure C based on a subset of the entire code vector set, Codebook C. sub When defined as such, this is the set of indices S of C. sub It can be represented by selecting.

[0254] [Mathematical Formula 7]

[0255] C sub = {v l,m │(l, m) ⊆ S sub} ⊆ C where S sub = {(l0, m0), (l1, m1), ..., (l q-1 , m q-1 )}

[0256] When the user determines such a structure, the structure can be communicated to the base station by feeding back a set of selected code vector indices.

[0257] In this case, the subset can represent whether a code vector within the entire reference codebook is included (on / off) in the form of a bitmap composed of bit sequences.

[0258] In a bit sequence with a length equal to the size of the entire codebook Q=N1O1N2O2, each bit can be mapped to a code vector within the entire codebook, and defined as 1 if the corresponding code vector is included in the user-defined codebook structure and 0 if it is not included.

[0259] That is, bit sequence b Q-1 , b Q-2, … , b0 can be defined as follows.

[0260] [Mathematical Formula 8]

[0261] b lN2O2+m = 1 if v l,m is a subset of C sub , 0 if v l,m is not a subset of C sub

[0262]

[0263] Here, b0 and b Q-1 represents the least significant bit (LSB) and the most significant bit (MSB), respectively.

[0264] If a user-defined codebook structure has an irregular pattern within an overall reference codebook aligned based on directionality, the structure can be shared between the user and the base station via a bitmap as shown above.

[0265] However, if this structure has a regular pattern, the number of bits required for CSI reporting can be reduced by defining a parameter representing the pattern of that structure within the entire reference codebook and sharing only that parameter instead of a bitmap.

[0266] In other words, to reduce overhead, the type of data to be shared (bitmap or parameter) must be applied differently depending on the regularity of the codebook structure.

[0267] In one embodiment of the present invention, as described above, the user-defined codebook structure is classified into cases having regular patterns and cases having irregular patterns, and the regular patterns are further divided into three types: cases where they are uniformly distributed over the entire area and cases where they are densely distributed in a specific direction area.

[0268]

[0269] Device configuration

[0270] FIG. 16 illustrates a wireless device that can be applied to the present technology.

[0271] Referring to FIG. 16, the first wireless device (100) and the second wireless device (200) can transmit and receive wireless signals through various wireless access technologies (e.g., LTE, NR). Here, the first wireless device (100) and the second wireless device (200) can correspond to the UE (110) and network (120) of FIG. 2, respectively.

[0272] The first wireless device (100) includes one or more processors (102) and one or more memories (104), and may additionally include one or more transceivers (106) and / or one or more antennas (108). The processor (102) controls the memory (104) and / or transceivers (106) and may be configured to implement the descriptions, functions, procedures, proposals, methods and / or flowcharts of operation disclosed in this document. For example, the processor (102) may process information within the memory (104) to generate a first information / signal and then transmit a wireless signal containing the first information / signal through the transceiver (106). Additionally, the processor (102) may receive a wireless signal containing a second information / signal through the transceiver (106) and then store information obtained from the signal processing of the second information / signal in the memory (104). Memory (104) may be connected to the processor (102) and may store various information related to the operation of the processor (102). For example, memory (104) may store software code containing instructions for performing some or all of the processes controlled by the processor (102) or for performing the descriptions, functions, procedures, proposals, methods, and / or operation sequence diagrams disclosed in this document. Here, the processor (102) and memory (104) may be part of a communication modem / circuit / chip designed to implement wireless communication technology (e.g., LTE E-UTRA, 5G NR). A transceiver (106) may be connected to the processor (102) and may transmit and / or receive wireless signals through one or more antennas (108). The transceiver (106) may include a transmitter and / or receiver. The transceiver (106) may be combined with an RF (Radio Frequency) unit. In the present invention, the wireless device may refer to a communication modem / circuit / chip.

[0273] The second wireless device (200) includes one or more processors (202) and one or more memories (204), and may additionally include one or more transceivers (206) and / or one or more antennas (208). The processor (202) controls the memory (204) and / or transceivers (206) and may be configured to implement the descriptions, functions, procedures, proposals, methods and / or operation sequences disclosed in this document. For example, the processor (202) may process information within the memory (204) to generate a third information / signal and then transmit a wireless signal containing the third information / signal through the transceiver (206). Additionally, the processor (202) may receive a wireless signal containing a fourth information / signal through the transceiver (206) and then store information obtained from the signal processing of the fourth information / signal in the memory (204). The memory (204) may be connected to the processor (202) and may store various information related to the operation of the processor (202). For example, the memory (204) may store software code containing instructions for performing some or all of the processes controlled by the processor (202) or for performing the descriptions, functions, procedures, proposals, methods, and / or operation sequence diagrams disclosed in this document. Here, the processor (202) and the memory (204) may be part of a communication modem / circuit / chip designed to implement wireless communication technology (e.g., LTE E-UTRA, 5G NR). The transceiver (206) may be connected to the processor (202) and may transmit and / or receive wireless signals through one or more antennas (208). The transceiver (206) may include a transmitter and / or receiver. The transceiver (206) may be used in combination with an RF unit. In the present invention, the wireless device may refer to a communication modem / circuit / chip.

[0274] Hereinafter, hardware elements of the wireless device (100, 200) will be described in more detail. Although not limited thereto, one or more protocol layers may be implemented by one or more processors (102, 202). For example, one or more processors (102, 202) may implement one or more layers (e.g., functional layers such as PHY, MAC, RLC, PDCP, RRC, SDAP). One or more processors (102, 202) may generate one or more Protocol Data Units (PDUs) and / or Service Data Units (SDUs) according to the descriptions, functions, procedures, proposals, methods, and / or flowcharts of operation disclosed in this document. One or more processors (102, 202) may generate messages, control information, data, or information according to the descriptions, functions, procedures, proposals, methods, and / or flowcharts of operation disclosed in this document. One or more processors (102, 202) may generate a signal (e.g., baseband signal) containing a PDU, SDU, message, control information, data, or information according to the functions, procedures, proposals, and / or methods disclosed in this document and provide it to one or more transceivers (106, 206). One or more processors (102, 202) may receive a signal (e.g., baseband signal) from one or more transceivers (106, 206) and may obtain a PDU, SDU, message, control information, data, or information according to the descriptions, functions, procedures, proposals, methods, and / or flowcharts disclosed in this document.

[0275] One or more processors (102, 202) may be referred to as a controller, microcontroller, microprocessor, or microcomputer. One or more processors (102, 202) may be implemented by hardware, firmware, software, or a combination thereof. For example, one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), one or more Digital Signal Processing Devices (DSPDs), one or more Programmable Logic Devices (PLDs), or one or more Field Programmable Gate Arrays (FPGAs) may be included in one or more processors (102, 202). The descriptions, functions, procedures, proposals, methods, and / or flowcharts disclosed in this document may be implemented using firmware or software, and the firmware or software may be implemented to include modules, procedures, functions, etc. Firmware or software configured to perform the descriptions, functions, procedures, proposals, methods, and / or operation sequences disclosed in this document may be contained in one or more processors (102, 202) or stored in one or more memories (104, 204) and driven by one or more processors (102, 202). The descriptions, functions, procedures, proposals, methods, and / or operation sequences disclosed in this document may be implemented using firmware or software in the form of code, instructions, and / or sets of instructions.

[0276] One or more memories (104, 204) may be connected to one or more processors (102, 202) and may store various forms of data, signals, messages, information, programs, codes, instructions, and / or commands. One or more memories (104, 204) may be composed of ROM, RAM, EPROM, flash memory, hard drive, registers, cache memory, computer read storage media, and / or combinations thereof. One or more memories (104, 204) may be located inside and / or outside of one or more processors (102, 202). Additionally, one or more memories (104, 204) may be connected to one or more processors (102, 202) through various technologies such as wired or wireless connections.

[0277] One or more transceivers (106, 206) may transmit user data, control information, wireless signals / channels, etc., as mentioned in the methods and / or operation flowcharts, etc., of this document to one or more other devices. One or more transceivers (106, 206) may receive user data, control information, wireless signals / channels, etc., as mentioned in the descriptions, functions, procedures, proposals, methods and / or operation flowcharts, etc., disclosed in this document from one or more other devices. For example, one or more transceivers (106, 206) may be connected to one or more processors (102, 202) and may transmit and receive wireless signals. For example, one or more processors (102, 202) may control one or more transceivers (106, 206) to transmit user data, control information, or wireless signals to one or more other devices. Additionally, one or more processors (102, 202) may control one or more transceivers (106, 206) to receive user data, control information, or wireless signals from one or more other devices. Additionally, one or more transceivers (106, 206) may be connected to one or more antennas (108, 208), and one or more transceivers (106, 206) may be configured to transmit and receive user data, control information, wireless signals / channels, etc., as described in the descriptions, functions, procedures, proposals, methods, and / or flowcharts of operation disclosed in this document through one or more antennas (108, 208). In this document, one or more antennas may be multiple physical antennas or multiple logical antennas (e.g., antenna ports). One or more transceivers (106, 206) can convert the received wireless signal / channel, etc. from an RF band signal to a baseband signal in order to process the received user data, control information, wireless signal / channel, etc. using one or more processors (102, 202).One or more transceivers (106, 206) can convert user data, control information, wireless signals / channels, etc. processed using one or more processors (102, 202) from baseband signals to RF band signals. To this end, one or more transceivers (106, 206) may include (analog) oscillators and / or filters.

[0278]

[0279] The detailed description of the preferred embodiments of the present invention disclosed above is provided to enable those skilled in the art to implement and practice the present invention. Although the present invention has been described with reference to preferred embodiments, those skilled in the art will understand that various modifications and changes can be made to the present invention without departing from the scope of the invention. For example, those skilled in the art may utilize each configuration described in the above embodiments in a manner that combines with one another.

[0280] Accordingly, the present invention is not intended to be limited to the embodiments shown herein, but to be given the broadest scope consistent with the principles and novel features disclosed herein.

[0281] The codebook for AI / ML-based CSI feedback according to the embodiments of the present invention as described above is suitable for use in a communication environment utilizing AI / ML of 5G systems in 3GPP, but can also be used in the same way in 3GPP's 6G and subsequent next-generation mobile communication systems.

Claims

1. In a method for a user device (UE) to communicate with a network in a mobile communication system, Receive AI (Artificial Intelligence) / ML (Machine Learning) related configuration information from the above network; and Based on the above configuration information, Channel Status Information (CSI) feedback is transmitted to the above network, The above CSI feedback is, A first mode that displays channel information in the form of a bitmap, A second mode that represents the above channel information based on beam spacing and start position, or A third mode representing the above channel information based on beam spacing, start position, and number of beams Represented in any one of the ways, Communication method.

2. In Paragraph 1, The first mode corresponds to an irregular codebook-based operation mode, the second mode corresponds to a uniform codebook-based operation mode, and the third mode corresponds to a localized codebook-based operation mode, Communication method.

3. In Paragraph 1, The above channel information is based on either a first type that transmits the CSI feedback information based on a codebook or a second type that transmits the CSI feedback information without being based on a codebook. Communication method.

4. In Paragraph 1, The above AI / ML-related configuration information is received via a Radio Resource Control (RRC) message representing one or more of the above first mode, the above second mode, or the above third mode, Communication method.

5. In Paragraph 4, The above CSI feedback information is, A first type method that determines any one of the first mode, the second mode, or the third mode based on a determination of the channel environment in the UE based on the information of the above RRC message, or A second type method that specifies any one of the first mode, the second mode, or the third mode in the information of the above RRC message. Based on one or more of the following, Communication method.

6. In Paragraph 4, The above CSI feedback is information for beam management of the above network, and The above RRC message includes information representing one or more of the above first mode, the above second mode, or the above third mode as information regarding Set B representing a measurement target for input or training of an AI / ML model. Communication method.

7. In Paragraph 1, The above AI / ML-related configuration information includes assignment information for different set B to each of a plurality of UEs for MU-MIMO (Multi-User Multiple Input Multiple Output) communication, and The above CSI feedback includes information about Set B assigned to the UE, Communication method.

8. In Paragraph 7, The above AI / ML-related configuration information additionally includes information on Set A representing the AI / ML-based estimation result beam combination, and Based on the inclusion relationship between the above set A and the above set B, the omission of information regarding either the above set A or the above set B is determined. Communication method.

9. In Paragraph 1, The above CSI feedback is information for CSI compression, and The above AI / ML-related configuration information includes a Pairing ID that distinguishes the pairing of the UE-side model and the network-side model, Communication method.

10. In Paragraph 9, It additionally includes transmitting a message to the above network that transmits UE Capability Info, which includes information about one or more UE-side models that the UE can support, and The AI / ML-related configuration information including the pairing ID is received from the network via an RRC reconfiguration message based on the UE performance information, Communication method.

11. In Paragraph 9, The above AI / ML-related configuration information additionally includes parameter combination information that defines the output message of the UE-side model corresponding to the above pairing ID, and The above parameter combination information includes at least two parameter combination information among (i) the number of real number items (d) of the output message, (ii) the length of the segment (L), and (iii) a quantization coefficient (Q) corresponding to the number of bits for quantization of the segment, Communication method.

12. In Paragraph 11, A method of representing the CSI feedback information in any one of the first to third modes through the output message defined according to the above parameter combination information, Communication method.

13. A method in which a network in a mobile communication system communicates with a user device (UE), Transmit AI (Artificial Intelligence) / ML (Machine Learning) related configuration information to the above UE; and Based on the above configuration information, receive CSI (Channel Status Information) feedback from the UE, The above CSI feedback is, A first mode that displays channel information in the form of a bitmap, A second mode that represents the above channel information based on beam spacing and start position, or A third mode representing the above channel information based on beam spacing, start position, and number of beams Appearing in any one of the following ways Communication method.

14. In a user device (UE) that communicates with a network in a mobile communication system, At least one processor; and It includes at least one computer memory that can be operably connected to the at least one processor and stores instructions that cause the at least one processor to perform operations when executed, and The above operations are, Receive AI (Artificial Intelligence) / ML (Machine Learning) related configuration information from the above network; and Based on the above configuration information, Channel Status Information (CSI) feedback is transmitted to the above network, The above CSI feedback is, A first mode that displays channel information in the form of a bitmap, A second mode that represents the above channel information based on beam spacing and start position, or A third mode representing the above channel information based on beam spacing, start position, and number of beams Represented in any one of the ways, User device.

15. In a network that communicates with multiple user devices (UEs) in a mobile communication system, At least one processor; and It includes at least one computer memory that can be operably connected to the at least one processor and stores instructions that cause the at least one processor to perform operations when executed, and The above operations are, Transmit AI (Artificial Intelligence) / ML (Machine Learning) related configuration information to the above UE; and Based on the above configuration information, receive CSI (Channel Status Information) feedback from the UE, The above CSI feedback is, A first mode that displays channel information in the form of a bitmap, A second mode that represents the above channel information based on beam spacing and start position, or A third mode representing the above channel information based on beam spacing, start position, and number of beams Appearing in any one of the following ways network.

Images