Method and device for reporting channel state information in wireless communication system

Abstract

The present invention may comprise the steps of: allowing a terminal to be allocated an uplink resource for artificial intelligence (AI) channel state information (CSI) reporting in a wireless communication system; determining whether to adjust an AI CSI payload size on the basis of the uplink resource; and if the AI CSI payload size is adjusted, performing AI CSI reporting through the uplink resource on the basis of the adjusted AI CSI payload size.

Description

Method and device for reporting channel state information in a wireless communication system

[0001] The present disclosure relates to a method and device for reporting channel state information (CSI) in a wireless communication system. Specifically, the present disclosure relates to a method and device for reporting CSI for an artificial intelligence / machine learning (AI / ML)-based system in a wireless communication system.

[0002]

[0003] The International Telecommunication Union (ITU) is developing the International Mobile Telecommunication (IMT) framework and standards, and is currently discussing fifth-generation (5G) communications through a program called "IMT for 2020 and beyond."

[0004] To meet the requirements presented in "IMT for 2020 and beyond," the 3rd Generation Partnership Project (3GPP) NR (New Radio) system is being discussed to support various numerologies based on time-frequency resource units, taking into account various scenarios, service requirements, and potential system compatibility.

[0005] Additionally, 5G communications can support the transmission of physical signals or physical channels through multiple beams to overcome adverse channel conditions such as high path loss, phase noise, and frequency offset that occur at high carrier frequencies. This enables 5G communications to support applications such as enhanced Mobile Broadband (eMBB), massive Machine Type Communications (mMTC), and Ultra Reliable and Low Latency Communication (URLLC).

[0006]

[0007] The technical problem of the present disclosure is a method and device for performing channel state information (CSI) reporting corresponding to at least one of AI / ML model input and output in a wireless communication system.

[0008] The technical problem of the present disclosure is a method and device for performing AI (artificial intelligence) CSI reporting in a wireless communication system.

[0009] The technical problem of the present disclosure is a method and apparatus for adjusting the AI ​​CSI payload size by considering uplink resources scheduled for uplink control information (UCI) transmission in a wireless communication system.

[0010] The technical problem of the present disclosure is a method and device for allocating each subsequence corresponding to a plurality of AI encoders to uplink resources in a wireless communication system and omitting some subsequences.

[0011] The technical problems to be achieved in the present disclosure are not limited to the technical problems mentioned above, and other technical problems not mentioned can be clearly understood by a person having ordinary skill in the technical field to which the present disclosure belongs from the description below.

[0012]

[0013] According to one aspect of the present disclosure, the method may include a step of allocating an uplink resource for an artificial intelligence (AI) channel state information (CSI) report to a terminal, a step of determining whether to adjust an AI CSI payload size based on the uplink resource, and a step of performing AI CSI reporting through the uplink resource based on the adjusted AI CSI payload size when the AI ​​CSI payload size is adjusted.

[0014] Also, according to one aspect of the present disclosure, a wireless user device includes at least one processor and a memory storing instructions for causing the wireless user device to perform a specific operation by the at least one processor, wherein the specific operation includes: allocating uplink resources for artificial intelligence (AI) channel state information (CSI) reporting, determining whether to adjust an AI CSI payload size based on the uplink resources, and performing AI CSI reporting through the uplink resources based on the adjusted AI CSI payload size when the AI ​​CSI payload size is adjusted.

[0015] Additionally, the following may equally apply:

[0016] According to one aspect of the present disclosure, if uplink resources for AI CSI reporting are sufficient, the terminal performs AI CSI reporting through the uplink resources, and if uplink resources for AI CSI reporting are not sufficient, the AI ​​CSI payload size can be adjusted.

[0017] Additionally, according to one aspect of the present disclosure, the AI ​​CSI payload size can be adjusted by adjusting the AI ​​encoder output size in a quantization method determined based on at least one of AI CSI model performance, reliability, and channel environment.

[0018] In addition, according to one aspect of the present disclosure, the AI ​​encoder in the terminal-side model performs AI encoding and quantization procedures through a quantization method determined based on AI CSI inputs to generate an AI encoder output, wherein the AI ​​encoder output is a bit string derived through the AI ​​encoder output size and the number of quantization bits per output node, and the AI ​​CSI payload size can be adjusted by adjusting at least one of the AI ​​encoder output size, the number of quantization bits per output node, and the bit string.

[0019] Additionally, according to one aspect of the present disclosure, the AI ​​encoder output may report additional output values ​​together through differential reporting operations based on the highest or most reliable output value based on the quantization step and range.

[0020] Additionally, according to one aspect of the present disclosure, if the uplink resources are sufficient to perform AI CSI reporting based on the adjusted AI CSI payload size, the terminal may perform AI CSI reporting through the uplink resources, and if the uplink resources are not sufficient to perform AI CSI reporting based on the adjusted AI CSI payload size, the terminal may omit transmission of some subsequences among at least one subsequence of the AI ​​encoder output.

[0021] Additionally, according to one aspect of the present disclosure, the AI ​​encoder output may include at least one subsequence based on a subsequence corresponding to each of at least one AI encoder, wherein the subsequence corresponding to each of at least one AI encoder may be assigned to the AI ​​CSI payload in an order of subsequences having a higher priority based on AI encoder priorities.

[0022] Additionally, according to one aspect of the present disclosure, each AI encoder corresponds to at least one of a layer, a rank, a sub-band, and a port, and the AI ​​encoder priority may be determined based on at least one of a layer, a rank, a sub-band, and a port corresponding to each AI encoder.

[0023] Additionally, according to one aspect of the present disclosure, each AI encoder corresponds to AI CSI content, and AI encoder priorities may be determined based on AI CSI content corresponding to each AI encoder.

[0024]

[0025] According to the present disclosure, a method for performing CSI reporting in a wireless communication system can be provided.

[0026] According to the present disclosure, a method for performing AI CSI reporting in a wireless communication system can be provided.

[0027] According to the present disclosure, a method for performing CSI reporting in a wireless communication system can be provided.

[0028] According to the present disclosure, a method for performing AI CSI reporting in a wireless communication system can be provided.

[0029] According to the present disclosure, a method for adjusting an AI CSI payload size in consideration of uplink resources scheduled for UCI transmission in a wireless communication system can be provided.

[0030] According to the present disclosure, a method for allocating each subsequence corresponding to a plurality of AI encoders to uplink resources and omitting some subsequences in a wireless communication system can be provided.

[0031] 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 will be clearly understood by a person having ordinary skill in the art to which the present disclosure pertains from the description below.

[0032]

[0033] FIG. 1 is a drawing for explaining the frame structure of a wireless communication system to which the present disclosure can be applied.

[0034] FIG. 2 is a diagram showing the resource structure of a wireless communication system to which the present disclosure can be applied.

[0035] FIG. 3 is a diagram illustrating CSI-RS resources to which the present disclosure can be applied.

[0036] FIG. 4 is a diagram illustrating a type 1 CSI applicable to the present disclosure.

[0037] FIG. 5 is a diagram illustrating a method of using multiple SRSs that can be applied to the present disclosure.

[0038] FIG. 6 is a diagram illustrating a method for performing non-codebook based precoding that can be applied to the present disclosure.

[0039] FIG. 7 is a diagram illustrating CSI omission applicable to the present disclosure.

[0040] Figure 8 is a diagram illustrating AI / ML model life cycle management applicable to the present disclosure.

[0041] FIG. 9 is a diagram illustrating a method for performing AI / ML model training based on Type 1 applicable to the present disclosure.

[0042] FIG. 10 is a diagram illustrating a method for performing AI / ML model training based on Type 2 applicable to the present disclosure.

[0043] FIG. 11 is a diagram illustrating a method for performing AI / ML model training based on Type 3 applicable to the present disclosure.

[0044] FIG. 12 is a diagram illustrating an AI encoder structure applicable to the present disclosure and a method for deriving output from input based thereon.

[0045] FIG. 13 is a diagram illustrating how scalar quantization is performed in an AI encoder applicable to the present disclosure.

[0046] FIG. 14 is a diagram illustrating a method for adjusting the AI ​​CSI payload size according to an adaptive quantization method within an AI encoder applicable to the present disclosure.

[0047] FIG. 15 is a diagram illustrating a method for adjusting the AI ​​CSI payload size according to an adaptive quantization method within an AI encoder applicable to the present disclosure.

[0048] FIG. 16 is a diagram illustrating a method for adjusting the AI ​​CSI payload size according to an adaptive quantization method within an AI encoder applicable to the present disclosure.

[0049] FIG. 17 is a diagram illustrating a method for adjusting an AI CSI payload size according to an adaptive quantization method in an AI encoder applicable to the present disclosure.

[0050] FIG. 18 is a diagram illustrating a method for adjusting AI CSI payload size based on AI encoder switching applicable to the present disclosure.

[0051] FIG. 19 is a diagram illustrating a method for adjusting the AI ​​CSI payload size through one or more AI encoder combinations applicable to the present disclosure.

[0052] FIG. 20 is a diagram illustrating a method for adjusting the AI ​​CSI payload size through one or more AI encoder combinations applicable to the present disclosure.

[0053] FIG. 21 is a diagram illustrating a method for adjusting the AI ​​CSI payload size through one or more applicable AI encoder combinations applicable to the present disclosure.

[0054] FIG. 22 is a diagram illustrating a method for adjusting the AI ​​CSI payload size through one or more applicable AI encoder combinations applicable to the present disclosure.

[0055] Figure 23 is a flowchart showing terminal operations applicable to the present disclosure.

[0056] FIG. 24 is a diagram illustrating a method for multiplexing and transmitting AI CSI content applicable to the present disclosure and existing CSI content.

[0057] FIG. 25 is a diagram illustrating a method for performing AI CSI reporting applicable to the present disclosure.

[0058] Figure 26 is a drawing showing a base station device and a terminal device to which the present disclosure can be applied.

[0059]

[0060] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present disclosure. However, the present disclosure may be implemented in various different forms and is not limited to the embodiments described herein.

[0061] In describing embodiments of the present disclosure, detailed descriptions of known configurations or functions will be omitted if they are deemed to obscure the gist of the present disclosure. Furthermore, portions of the drawings that are irrelevant to the description of the present disclosure have been omitted, and similar portions are designated with similar reference numerals.

[0062] In the present disclosure, when a component is said to be "connected," "coupled," or "connected" to another component, this may include not only a direct connection, but also an indirect connection in which another component exists in between. Furthermore, when a component is said to "include" or "have" another component, unless otherwise specifically stated, this does not exclude the other component, but rather implies that the other component may be included.

[0063] In this disclosure, terms such as first, second, etc. are used solely to distinguish one component from another, and do not limit the order or importance of components unless specifically stated otherwise. Accordingly, within the scope of this disclosure, a first component in one embodiment may be referred to as a second component in another embodiment, and similarly, a second component in one embodiment may be referred to as a first component in another embodiment.

[0064] In this disclosure, distinct components are used to clearly illustrate their respective characteristics, and do not necessarily imply that the components are separated. That is, multiple components may be integrated into a single hardware or software unit, or a single component may be distributed into multiple hardware or software units. Therefore, even if not specifically mentioned, such integrated or distributed embodiments are also included within the scope of this disclosure.

[0065] In the present disclosure, the components described in various embodiments are not necessarily essential components, and some may be optional components. Therefore, embodiments comprising a subset of the components described in one embodiment are also within the scope of the present disclosure. Furthermore, embodiments including other components in addition to the components described in various embodiments are also within the scope of the present disclosure.

[0066] The present disclosure describes a wireless communication network, and operations performed in the wireless communication network may be performed in a process of controlling the network and transmitting or receiving a signal in a system (e.g., a base station) that manages the wireless communication network, or in a process of transmitting or receiving a signal in a terminal connected to the wireless network.

[0067] It is self-evident that various operations performed for communication with terminals in a network consisting of multiple network nodes including a base station can be performed by the base station or other network nodes other than the base station. 'Base station (BS)' can be replaced by terms such as fixed station, Node B, eNodeB (eNB), ng-eNB, gNodeB (gNB), and access point (AP). In addition, 'terminal' can be replaced by terms such as UE (User Equipment), MS (Mobile Station), MSS (Mobile Subscriber Station), SS (Subscriber Station), and non-AP station (non-AP STA).

[0068] In the present disclosure, transmitting or receiving a channel means transmitting or receiving information or a signal through the channel. For example, transmitting a control channel means transmitting control information or a signal through the control channel. Similarly, transmitting a data channel means transmitting data information or a signal through the data channel.

[0069] In the following description, the term NR (New Radio) system is used for the purpose of distinguishing the system to which various examples of the present disclosure are applied from existing systems; however, the scope of the present disclosure is not limited by this term.

[0070] NR systems support a variety of subcarrier spacings (SCS) to accommodate diverse scenarios, service requirements, and potential system compatibility. Furthermore, NR systems can support the transmission of physical signals / channels across multiple beams to overcome challenging channel conditions, such as high path loss, phase noise, and frequency offsets that occur at high carrier frequencies. This enables NR systems to support applications such as enhanced Mobile Broadband (eMBB), massive Machine Type Communications (mMTC) / ultra Machine Type Communications (uMTC), and Ultra Reliable and Low Latency Communications (URLLC).

[0071] Hereinafter, 5G mobile communication technology can be defined to include not only the NR system, but also the existing LTE-A (Long Term Evolution-Advanced) system and LTE (Long Term Evolution) system. 5G mobile communication may include technology that operates in consideration of backward compatibility with previous systems as well as the newly defined NR system. Therefore, the 5G mobile communication below may include technology that operates based on the NR system and technology that operates based on previous systems (e.g., LTE-A, LTE), and is not limited to a specific system.

[0072] First, we would like to briefly explain the physical resource structure of the NR system to which the present invention is applied.

[0073] FIG. 1 is a drawing for explaining an NR frame structure to which the present disclosure can be applied.

[0074] In NR, the basic unit of time domain is It can be, , and N can be 4096. Meanwhile, in LTE, the basic unit of the time domain is It can be, And, =2048. The constant for the multiplication relationship between the NR time base unit and the LTE time base unit is k= can be defined as

[0075] Referring to Figure 1, the time structure of a frame for downlink / uplink (DL / UL) transmission is can have. Here, one frame is It consists of 10 subframes corresponding to time. The number of consecutive OFDM symbols in each subframe is = It can be. In addition, each frame is divided into two half frames of the same size, half frame 1 can be composed of sub frames 0-4, and half frame 2 can be composed of sub frames 5-9.

[0076] represents the timing advance (TA) between the downlink (DL) and uplink (UL). Here, the transmission timing of the uplink transmission frame i is determined based on the downlink reception timing at the terminal, based on the following mathematical expression 1.

[0077] [Mathematical Formula 1]

[0078]

[0079]

[0080] Here, It may be a TA offset value that occurs due to differences in duplex mode, etc. In FDD (Frequency Division Duplex), has a value of 0, but in TDD (Time Division Duplex), it takes into account the margin for DL-UL switching time. It can be defined as a fixed value. For example, in TDD (Time Division Duplex) of FR1 (Frequency Range 1), which is a frequency below 6 GHz, is 39936 or 25600 It could be 39936 is 20.327μs, and 25600 is 13.030μs. Also, at FR2 (Frequency Range 2), which is a millimeter wave (mmWave) frequency, is 13792 It could be. At this time, 13792 is 7.020 μs.

[0081] FIG. 2 is a diagram showing an NR resource structure to which the present disclosure can be applied.

[0082] Resource elements (REs) within a resource grid can be indexed according to each subcarrier spacing. Here, one resource grid can be created for each antenna port and each subcarrier spacing. Uplink and downlink transmission and reception can be performed based on the corresponding resource grid.

[0083] In the frequency domain, one Resource Block (RB) consists of 12 REs, and each of the 12 REs can be configured with an index (nPRB) for one RB. The index for an RB can be utilized within a specific frequency band or system bandwidth. The index for an RB can be defined as in the following mathematical expression 2. Here, represents the number of subcarriers per RB, and k represents the subcarrier index.

[0084] [Equation 2]

[0085]

[0086]

[0087] Different numerologies can be configured to meet the diverse services and requirements of NR systems. For example, while LTE / LTE-A systems can support a single subcarrier spacing (SCS), NR systems can support multiple SCSs.

[0088] A new numerology for NR systems supporting multiple SCSs can operate in frequency ranges or carriers such as below 3GHz, 3GHz-6GHz, 6GHZ-52.6GHz or above 52.6GHz to address the issue of not being able to use wide bandwidth in frequency ranges or carriers such as 700MHz or 2GHz.

[0089] Table 1 below shows examples of numerologies supported by the NR system.

[0090] [Table 1]

[0091]

[0092] Referring to Table 1 above, the numeral can be defined based on the subcarrier spacing (SCS), cyclic prefix (CP) length, and number of OFDM symbols per slot used in the Orthogonal Frequency Division Multiplexing (OFDM) system. The above values ​​can be provided to the terminal through the upper layer parameters DL-BWP-mu and DL-BWP-cp for the downlink, and through the upper layer parameters UL-BWP-mu and UL-BWP-cp for the uplink.

[0093] In Table 1 above, when the subcarrier spacing setting index (u) is 2, the subcarrier spacing (Δf) is 60 kHz, and normal CP and extended CP can be applied. For other numerology indices, only normal CP can be applied.

[0094] A normal slot can be defined as the basic time unit used to transmit a single piece of data and control information in an NR system. The length of a normal slot can be set to the number of OFDM symbols, which is 14 by default. Furthermore, unlike slots, a subframe has an absolute time length equivalent to 1 ms in an NR system and can be used as a reference time for the length of other time intervals. Here, for coexistence or backward compatibility between LTE and NR systems, a time interval similar to an LTE subframe may be required in the NR standard.

[0095] For example, in LTE, data can be transmitted based on a unit of time called a Transmission Time Interval (TTI), which can be set to one or more subframes. Here, one subframe can be set to 1ms and can contain 14 OFDM symbols (or 12 OFDM symbols).

[0096] In addition, a non-slot can be defined in NR. A non-slot can mean a slot having a number that is at least one symbol smaller than a normal slot. For example, in the case of providing low latency such as URLLC service, latency can be reduced through a non-slot having a number of symbols smaller than a normal slot. Here, the number of OFDM symbols included in a non-slot can be determined by considering the frequency range. For example, a non-slot with a length of 1 OFDM symbol can be considered in a frequency range of 6 GHz or higher. As an additional example, the number of OFDM symbols defining a non-slot can include at least 2 OFDM symbols. Here, the range of the number of OFDM symbols included in a non-slot can be set as the length of a mini-slot up to a predetermined length (e.g., the normal slot length - 1). However, as a specification of a non-slot, the number of OFDM symbols may be limited to 2, 4, or 7 symbols, but is not limited thereto.

[0097] Additionally, for example, in unlicensed bands below 6 GHz, subcarrier spacing where u equals 1 and 2 may be used, and in unlicensed bands above 6 GHz, subcarrier spacing where u equals 3 and 4 may be used. For example, when u equals 4, it may be used for SSB (Synchronization Signal Block).

[0098] [Table 2]

[0099]

[0100]

[0101] Table 2 shows the number of OFDM symbols per slot for normal CP, by subcarrier spacing setting (u). ), number of slots per frame ( ), number of slots per subframe ( ) is shown. Table 2 shows the above-described values ​​based on a normal slot having 14 OFDM symbols.

[0102] [Table 3]

[0103]

[0104]

[0105] Table 3 shows the number of slots per frame and the number of slots per subframe based on normal slots with 12 OFDM symbols per slot when extended CP is applied (i.e., when u is 2 and the subcarrier spacing is 60 kHz).

[0106] As mentioned above, one subframe may correspond to 1ms on the time axis. Furthermore, one slot may correspond to 14 symbols on the time axis. For example, one slot may correspond to 7 symbols on the time axis. Accordingly, the number of slots and symbols that can be considered within 10ms corresponding to one radio frame may be set differently. Table 4 may show the number of slots and symbols according to each SCS. In Table 4, the 480kHz SCS may not be considered, but the present invention is not limited to these examples.

[0107] [Table 4]

[0108]

[0109]

[0110] A terminal can calculate downlink CSI parameters (e.g., CQI, PMI, RI, L1-RSRP, etc.) through a DL (downlink) CSI (channel state information) reporting procedure and report the calculated CSI information to a base station. The base station can utilize the received CSI information for downlink data transmission. The CSI parameters are measurement values ​​associated with a channel state, and the base station can perform downlink data transmission based on the measurement values. The base station can instruct the terminal about data scheduling information based on the CSI parameters. The terminal generally reports the CSI calculation and derived CSI parameters to the base station as feedback information by receiving a CSI-RS transmitted by the base station. Alternatively, the terminal can measure the channel state through another downlink signal (e.g., SSB (synchronization signal block)) in addition to the CSI-RS and report the measured value to the base station.

[0111]

[0112] Select RI (rank indicator):

[0113] RI can be a CSI parameter that indicates the number of layers available for downlink transmission in a specific channel environment. RI can also indicate the maximum number of uncorrelated paths available for downlink transmission. Here, other CSI parameters (e.g., PMI, CQI) can be calculated based on the rank provided by RI.

[0114] Select precoding matrix index (PMI):

[0115] A PMI may include a set of indices corresponding to a precoding matrix. The base station may apply the PMI reported by the terminal to downlink transmission. However, the base station may have the freedom to apply a precoding matrix other than the PMI. For example, the standard may support multiple types of codebooks to report the precoding matrix for the PMI. Each codebook type is described below.

[0116] Codebook type

[0117] - Type 1 single-panel codebooks

[0118] - Type 1 multi-panel codebooks

[0119] - Type 2 codebooks

[0120] - Enhanced Type 2 codebooks

[0121]

[0122] For example, PMI selection may be performed based on the codebook type, the number of transmission layers represented by RI values, and other CSI reporting configuration parameters (e.g., antenna panel dimensions). Each codebook may include a set of precoding matrices. For example, in NR, a dual-stage precoding matrix codebook design may be applied, but is not limited thereto.

[0123] The dual-stage precoding matrix may be as shown in Equation 3 below. In Equation 3, W1 may be a matrix representing a DFT (discrete Fourier transform) beam or a beam group for both polarizations according to the codebook type, and W2 may be selected from one of the following according to the codebook type.

[0124] W2 codebook type

[0125] - Beam selection from W1

[0126] - Weighting coefficients for the beams in W1

[0127] - Cophasing values ​​between two polarizations

[0128]

[0129] [Equation 3]

[0130]

[0131]

[0132] The terminal calculates a PMI value that represents the highest SINR (signal interference noise ratio) value at the receiver using the codebook selected for "Type I single-panel codebook" and "Type I multi-panel codebook" as the codebook type and all possible precoding matrices in the given channel environment. For example, a Type 1 codebook may be considered for PMI selection in a single user multi-input multi-output (SU-MIMO) scenario, but is not limited thereto.

[0133] A Type II codebook can consider a set of orthogonal DFT beams. In Equation 3, the W1 matrix contains information associated with the DFT beams, and the PMI selection function calculates beam amplitude scaling and cophasing values ​​(W2) for all DFT beams within each orthogonal beam group for a given channel environment and number of DFT beams. The terminal can report the i1 and i2 values ​​to the base station based on the PMI selection function, and each index corresponds to a precoding matrix (W) that provides the maximum SINR value. A Type II codebook-based PMI can provide more accurate channel information than a Type I codebook-based PMI. A Type II codebook can provide more accurate channel information because multiple beams are applied to estimate the channel eigenvector. For example, a Type II codebook-based PMI can be used in a MU (multi)-MIMO scenario because it considers multiple beams, but is not limited thereto. For example, a Type 2 codebook can only support up to Rank 2, while an enhanced Type 2 codebook can support up to Rank 4 with overhead reduction methods, which will be described later.

[0134] Furthermore, based on the above, CSI reporting performed by a terminal may consider explicit CSI feedback and implicit CSI feedback. Explicit CSI feedback is a method in which the terminal directly feeds back channel information (e.g., channel eigenvector / eigenvalue, SVD) indicating the channel state measured by the terminal, and may not consider how the base station processes the reported CSI. On the other hand, implicit CSI feedback may be a method in which a selected precoder matrix (desired precoder matrix) W within a set of candidate precoder matrices from a possible precoder codebook is indicated.

[0135] For example, CSI parameters (or CSI contents) related to CSI feedback may be as shown in Table 5 below, but may not be limited thereto.

[0136] [Table 5]

[0137]

[0138]

[0139] The CSI content within a CSI report can be set by different combinations of the CSI parameters in Table 5. Here, the CSI content within a CSI report can be set by reportQuantity.

[0140] For example, when set to "none", the base station may trigger aperiodic CSI-RS transmissions. Aperiodic CSI-RS transmissions may be triggered by the terminal for channel measurement and receiver parameter adjustment, and related CSI reporting may not be performed by the terminal. Aperiodic CSI-RS transmissions may be, but are not limited to, aperiodic transmissions of CSI-RS or TRS (tracking reference signal) for tracking purposes for synchronization purposes, or spatial aperiodic receive beam sweeping for the terminal to adjust its receive filter. The following operations do not require the base station to report aperiodic CSI to the terminal, but may operate within the same framework as aperiodic CSI requests.

[0141] - Aperiodic transmission of the Tracking Reference Signal (TRS or CSI-RS for tracking)

[0142] - Aperiodic Rx beam sweeping (so called P3 beam sweep) - adjusts the Rx spatial received filter to the UE

[0143]

[0144] Additionally, CSI contents within a single CSI report may be set as shown in Table 6 below, but may not be limited thereto. For example, the combinations of CSI contents in Table 6 below are representative combinations of CSI contents that a base station will primarily need.

[0145] [Table 6]

[0146]

[0147]

[0148] For example, “cri-RI-i1” in Table 6 can operate based on “Hybrid beamformed / non-precoded CSI acquisition operation” as shown in FIG. 3.

[0149] The base station can mainly use UE-specific beamformed CSI-RS (320) with several ports per CSI-RS resource for the purpose of CSI acquisition. However, in order for the base station to recognize how to beamform the CSI-RS for each UE, non-precoded CSI-RS resources (310) can be transmitted in the middle together with multiple antenna ports (e.g., 32 ports). Here, the UE can report PMI corresponding to all antenna ports (e.g., 32 ports) based on the received CSI-RS, thereby indicating a preferred beam direction. For example, beamforming of the CSI-RS can be based on the wideband / long-term part of the reported precoder. That is, the reported W1 metrics can be used to perform UE-specific CSI-RS beamforming. Additionally, the short-term subband CS (short-term / subband CSI) (W2) can be determined from the beamforming CSI-RS.

[0150] For example, non-precoded CSI-RS (310) can be used to identify how CSI-RS beamforming is performed, and based on this, the terminal may only need to report CSI corresponding to W1 and not report for W2. This can reduce overhead.

[0151] Additionally, the terminal can select a PMI based on a specific number of transmit antennas. Specifically, the terminal can select a PMI based on a specific number of transmit antennas given by the number of antenna ports of a configured CSI-RS associated with the reporting configuration.

[0152] Additionally, in the case of MU-MIMO scheduling, the base station may select a PMI other than the PMI reported by the target terminal, taking into account interference or channel conditions from other terminals simultaneously scheduled. Considering the above, two types of CSI can be defined in a wireless communication system, as shown in Table 7 below.

[0153] [Table 7]

[0154]

[0155]

[0156] Specifically, type 1 CSI may have different sub-codebook formats based on different antenna configurations at the transmitter. As an example, FIG. 4 is a diagram illustrating type 1 CSI applicable to the present disclosure. Specifically, FIG. 4(a) may be a case where 16 antenna ports are configured with (N1, N2) = (4, 2). The precoding matrix W for type 1 single panel CSI may be expressed as the product of two matrices (W1, W2), each of which may be independently reported as a different part of the overall PMI. For example, W1 may reflect long-term frequency independent channel characteristics, and thus may report channel information corresponding to the entire bandwidth (wideband reporting). For example, W1 may be as shown in Equation 4 below. W1 may indicate a set of beams pointing in different directions, and matrix B may define one beam. In Equation 4, based on the 2 X 2 block structure, there can be two polarizations, and the selection of W1 can be a limited set selection with the same beam direction.

[0157] [Equation 4]

[0158]

[0159]

[0160] As a concrete example, for rank 1 or 2, a single beam or four neighboring beams can be indicated by the W1 matrix. The correct beam among the four neighboring beams can be selected by W2 reported for each subband, thereby selecting an optimized beam for each subband. Furthermore, W2 can provide cophasing between two polarizations. If W1 defines a single beam, then W2 in B, which is a single-column matrix, can only provide cophasing between two polarizations. On the other hand, if multiple neighboring beams are defined by W1, W2 can select one correct beam and provide cophasing between two polarizations within that beam.

[0161] On the other hand, if the rank is greater than 2, W1 can define N neighboring orthogonal beams, which can be as shown in Equation 5. In Equation 5, N beams used for transmission of R transmission layers can be indicated, and W2 can only provide cophasing between two polarizations. Here, transmission to the same terminal can be possible for up to 8 transmission layers.

[0162] [Equation 5]

[0163]

[0164]

[0165] On the other hand, W2 is short-term and can potentially indicate different channel characteristics depending on the frequency. W2 may not be reported in certain cases. Here, the base station can randomly select W2 for each physical resource block group (PRG), and the terminal can select the channel quality indicator (CQI) based on this assumption.

[0166] Also, Fig. 4(b) may be a Type 1 multi-panel CSI. Multi-panel CSI may be applied considering a case where multiple antenna panels are used at a base station. Here, coherence between different panels may be difficult to guarantee. For example, Fig. 4(b) may be a case where (N1, N2) = (4, 1) and there are 32 antenna ports by 4 panels, but the present invention is not limited thereto. Here, the difference between the single panel of Fig. 4(a) and the multi-panel of Fig. 4(b) is that coherence may be difficult to guarantee between the antenna ports of different antenna panels. In the multi-panel, W1 may be the same set of beams for different polarizations and panels, just like in the single panel. On the other hand, in the case of the multi-panel, W2 may provide cophasing not only between polarizations but also between multi-panels for each subband.

[0167] Type 2 CSI can provide channel information with higher spatial granularity than Type 1 CSI and can be primarily applied to MU-MIMO scenarios. For example, Type 2 CSI may be limited to a maximum of two ranks, while Enhanced Type 2 CSI may be extended to four ranks, but may not be limited to a specific form. For convenience of explanation, the terms Type 2 CSI and Enhanced Type 2 CSI are referred to below, but may not be limited thereto.

[0168] Here, in the above-described mathematical expressions 3 and 4, W1 may be a wideband report. Type 2 CSI may report up to four beams, corresponding to four columns within B. W2 may provide amplitude values ​​(partially wideband and partially subband reporting) and phase values ​​(subband reporting) for each of the four reported beams and two polarizations within B.

[0169] Type 2 CSI can provide more detailed channel model information than Type 1, including main ray and amplitude and phase information. Based on the CSI reported from multiple terminals, the base station can determine which combinations of terminals are appropriate for simultaneous transmission on the same time / frequency resource.

[0170] Additionally, enhanced Type 2 CSI (e.g., Rel-16 enhanced Type 2 CSI) may be considered. Enhanced Type 2 CSI can report a set of beams on the wideband, similar to Type 2 CSI, along with a set of combining coefficients on an additional narrowband. Here, the reported beams can be linearly combined using the combining coefficients to provide a set of precoder vectors for each transmission layer. For example, in the existing Type 2 CSI, the combining coefficient value can be reported independently for each subband. Therefore, even considering the fact that the channels between adjacent subbands are highly correlated, the existing Type 2 CSI must report a relatively large combining coefficient value on a per-subband basis, which can increase the overhead. Enhanced Type 2 CSI can utilize frequency correlation to reduce this overhead. Furthermore, enhanced Type 2 CSI can improve the frequency-domain granularity of the PMI report. For example, compression operations can be applied to subbands or half-subbands based on frequency domain units. While conventional Type 2 CSI uses one precoder per subband, enhanced Type 2 CSI can use a recommended precoder for each frequency domain unit. However, actual reporting can be performed together for all frequency units. Specifically, given layer k, the reported precoders for all frequency domain units can be expressed as in Equation 6 below, where N can be the number of frequency domain units.

[0171] [Equation 6]

[0172]

[0173]

[0174] Here, Equation 6 may not be a mapping of layers and antennas, but may mean a set of precoder vectors for the entire set of frequency domain units for layer k. For the entire k layers, there may be a set of k precoder vectors, each set may be composed of N precoder vectors, and the layer and antenna port mapping for a specific frequency domain unit n may be as shown in Equation 6 below.

[0175] Additionally, in the expression for vectors in Equation 7, W1 can be the same as Equation 4 described above, which is the same as the existing Type 2 CSI.

[0176] [Equation 7]

[0177]

[0178]

[0179] Here, the B columns can correspond to the L selected beams, and W1 can be the same for all frequency domain units. (wideband reporting) Also, W1 can be the same for all layers.

[0180] Additionally, as an example, the enhanced Type 2 CSI has a compression matrix (size M x N) can be considered. can be constructed as a set of row vectors from the DFT basis, and can provide a transformation from N dimensions in the frequency domain (N frequency domain units) to M dimensions in the smaller delay domain. can be frequency independent, but can be reported independently for each layer. That is, can be a single common metric for all frequency domain units, but can be reported independently for each layer. The number of rows can be equal to M in Equation 8. In Equation 8, R can be the number of frequency domain units per subframe (R=1 or 2), and p can be a configurable parameter that controls the amount of compression.

[0181] [Equation 8]

[0182]

[0183]

[0184] Finally can perform mapping from the delay domain to the beam domain. For example, W2 of the existing Type 2 CSI can generate a similar matrix for all subbands, and the matrix can be mapped from the frequency (subband) domain to the beam domain. On the other hand, can perform allocation from a smaller delay domain to a beam domain. can be adjusted by the reported rank and compression factor parameters p, which allows reporting based on fewer parameters. In addition, The amount of CSI to be reported can be adjusted through the β value of 2LM, which allows for a smaller Fewer parameters can be reported based on size.

[0185] Additionally, as an example, the enhanced Type 2 CSI may be expanded to a maximum of 4 reported ranks and a maximum number of selectable beams to 6, with reduced overhead and improved frequency domain granularity. Table 8 below may illustrate possible configurations of the enhanced Type 2 CSI, but is not limited thereto.

[0186] [Table 8]

[0187]

[0188]

[0189] Additionally, multi-antenna precoding can be considered in uplink transmission. PUSCH (physical uplink shared channel) MIMO precoding can support up to four layers. For uplink transmission, DFT-based precoding can only support single-user (SU) MIMO. For example, a terminal can perform codebook-based transmission and non-codebook-based transmission with two different PUSCH precoding modes. In case of codebook-based precoding, the uplink grant transmitted by the base station can include precoder-related information (e.g., PMI). It is assumed that the terminal uses the precoder provided by the base station in the uplink differently from the downlink transmission. Furthermore, coherence can be assumed between the terminal antennas in uplink MIMO transmission, and the phase of the signals transmitted on the two antennas can be adjusted. Here, if there is no coherence between the antenna ports, each antenna port may result in a random relative phase, which may mean that the use of antenna port specific weight factors is limited.

[0190] For example, the coherence between antenna ports can correspond to any one of full coherence, partial coherence, and no coherence, and the corresponding information can be provided as terminal capability information. In addition, the use of multiple sounding reference signals (SRS) for codebook-based PUSCH can be considered. FIG. 5 is a diagram illustrating a method of using multiple SRSs applicable to the present disclosure. Referring to FIG. 5, a terminal can transmit relatively large beams on a multi-port SRS. For example, the beams can correspond to different terminal antenna panels (in different directions), and each panel can include a set of antenna elements. That is, each panel can correspond to antenna ports of a multi-port SRS. The base station can transmit an SRI (SRS Resource Indicator) to the terminal to determine whether to perform transmission through which beam. The terminal can determine which transmission to perform within the selected beam based on the number of layers and precoder using SRI and precoder information.

[0191] FIG. 6 is a diagram illustrating a method for performing non-codebook-based precoding applicable to the present disclosure. Referring to FIG. 6, when performing non-codebook-based precoding, a terminal can perform transmission based on a precoder indication and channel measurement received from a base station. In codebook-based precoding, the terminal performs transmission based on an uplink precoder selected based on channel measurement by the base station, but in non-codebook, uplink transmission can be performed through a precoder selected based on measurement based on channel reciprocity. Here, each column of the precoder matrix W can be a digital beam for a corresponding layer. When the terminal selects a precoder for N layers, it can be regarded as selecting N different beam directions, and each beam can correspond to one possible layer.

[0192] For example, referring to FIG. 6, a terminal may be capable of PUSCH transmission based on the selected precoding. However, the precoder selected by the terminal based on downlink channel measurements may not be the optimal precoder from the base station's perspective. Considering the above, the base station may modify the precoder selected by the terminal. For example, the base station may remove some beams or some columns selected by the terminal from the precoder and transmit information about this to the terminal. Specifically, the base station may indicate the result of beam selection based on measured SRS information back to the terminal through SRI. Based on the indicated information, the terminal may perform only some beam transmissions, which may implicitly indicate reduced layer transmission.

[0193] A terminal can transmit uplink control information (UCI) to a base station. For example, the terminal can simultaneously transmit UCI on a physical uplink control channel (PUCCH) and a PUSCH. Specifically, the terminal can be instructed with a "simultaneous PUCCH and PUSCH transmission" parameter through upper layer signaling or dynamic signaling, and can configure simultaneous PUCCH and PUSCH transmission for UCI. That is, the terminal can be configured to perform simultaneous transmission for PUCCH and PUSCH. Here, if PUCCH and PUSCH are transmitted on the same carrier, "non-single carrier waveform transmission" may occur. Accordingly, PUCCH and PUSCH can be transmitted based on a large transmission backoff (e.g., 10 dB) based on frequency location resource allocation size, transmission power, and other factors. On the other hand, when PUCCH and PUSCH are transmitted on different carriers, a large power backoff may or may not be required based on a single power amplitude (PA) or multiple PAs.

[0194] As another example, if simultaneous transmission on PUCCH and PUSCH is required, the UE may drop the PUCCH and transmit UCI on the PUSCH. For example, the UCI type piggybacked on the PUSCH may be at least one of HARQ (hybrid automatic repeat and request), CSI type 1, and CSI type 2. On the other hand, the SR (scheduling request) may not be piggybacked on the PUSCH and may be replaced with a BSR (buffer state report), but is not limited thereto.

[0195] Upper layer parameter beta The value is used to determine the amount of radio resources transmitted on the PUSCH, and this value can be indicated to the UE from the base station. The beta value can have a wide range based on the amount of radio resources. When the UE transmits UCI on the PUSCH, the UE can determine the amount of radio resources required for UCI transmission by considering the UCI payload size on the PUSCH and the PUSCH spectral efficiency along with the above-mentioned beta value. This can prevent unnecessary resource usage for UCI, and the corresponding operation can be controlled by upper-layer parameters.

[0196] Additionally, as an example, the size of the UCI information bits may vary depending on the UCI type, and thus different UCI handling methods may be considered. Consider the cases in Table 9 below.

[0197] [Table 9]

[0198]

[0199]

[0200] [Table 10]

[0201]

[0202]

[0203] For example, in case of PUCCH-based subband CSI reporting and PUSCH-based reports with Type 1 CSI feedback, CSI Part 1 may include RI (if reported) and CSI-RS Resource Indicator (CRI) (if reported). In addition, CSI Part 1 may include CQI for the first codeword (first CW). CSI Part 2 may include CQI and PMI for the second CW when RI is greater than 4.

[0204] CSI Part 1 may additionally include an indication of the number of "non-zero wideband amplitude coefficients per layer" for Type 2 CSI feedback over PUSCH. Here, the "wideband amplitude coefficient" is part of Type 2 CB, and the PMI payload size may vary depending on whether the coefficient value is zero or not.

[0205] Additionally, CSI Part 1 and CSI Part 2 can each be set based on Table 11 and Table 12.

[0206] [Table 11]

[0207]

[0208]

[0209] [Table 12]

[0210]

[0211]

[0212] Additionally, diverse uplink data scheduling may be required. The beta value can be dynamically indicated to the UE, taking into account the potentially different QoS requirements for each UCI and UCI of the PUSCH. For example, the base station can indicate one of four different beta values ​​pre-configured through higher layers via downlink control information (DCI).

[0213] For example, when allocating UCI on PUSCH, HARQ may have the highest priority, followed by CSI Type 1 and CSI Type 2. The priority may be determined based on whether the UCI is mapped closely to the DMRS on the PUSCH on which it is transmitted.

[0214] Here, in case of UCI 1 or 2-bit HARQ, some REs can be reserved for puncturing by pretending to be 2-bit HARQ bits, and the corresponding information can be transmitted through puncturing. In addition, in case of UCI greater than 2-bit HARQ and CSI Part 1 and CSI Part 2, rate matching can be performed starting from the available REs from the right non-DMRS symbol after the first DMRS symbol, and HARQ can be performed first, followed by CSI Part 1 and CSI Part 2.

[0215] Additionally, for example, CSI Part 1 may not be allocated to REs reserved for HARQ to avoid being punctured by HARQ. On the other hand, CSI Part 2 and uplink data may be allocated to REs reserved for HARQ (i.e., potentially entailing puncturing).

[0216] UCI on PUSCH can be allocated in frequency order and then in time order. That is, UCI on PUSCH can be allocated starting from the lowest available RE of the lowest available symbol. As described above, the base station can decode UCI quickly. In order to maximize the frequency diversity gain for UCI on PUSCH within one uplink symbol, REs for UCI can be allocated in a 'frequency-distributed' manner, but is not limited thereto. Specifically, the distance (d) between two adjacent REs within one orthogonal frequency division multiplexing (OFDM) symbol to which UCI is allocated can be set to 1 (d=1) when the number of modulation symbols to be allocated for UCI is greater than or equal to the number of REs available within the uplink symbol.

[0217] Otherwise, d can be determined by floor(number of available REs in an uplink symbol / number of modulation symbols to be allocated for UCI). Here, d>1 for a UL symbol, and the allocation of UCI modulation symbols in that symbol can be frequency distributed. In addition, all UCI types can be allocated to all layers of PUSCH transmission and use the same PUSCH modulation. For example, CSI can be allocated only to the layer with the highest modulation coding scheme (MCS), but is not limited thereto.

[0218] For example, in UL MIMO transmission, 2 TBs are used in LTE, but in 5G NR, 4-layer UL MIMO transmissions are used for one TB, so CSI can be allocated only to the layer with the highest MCS.

[0219] Additionally, when frequency hopping is enabled on the PUSCH, UCI modulation symbols can be divided into two parts of approximately equal size. These two parts can be assigned to each of the two hops, and can be divided into approximately equal parts. Accordingly, the negative impact on uplink data due to UCI piggybacking on the PUSCH can be equally distributed across the two hops.

[0220] Additionally, partial CSI omission can be considered in PUSCH-based CSI reporting. In PUSCH-based CSI reporting and Type 2 CSI reporting, the CSI payload size can dynamically change depending on the rank indicator (RI) selection. For example, when Type 2 CSI reporting is performed, the PMI payload when RI is 2 can be approximately twice as large as the PMI payload when RI is 1. Since the base station cannot recognize the RI selection in advance, it can perform scheduling by predicting the RI value that the UE is likely to select when performing PUSCH scheduling. Therefore, a misalignment due to RI selection may occur between the base station and the UE. If the RI selected by the UE is larger than the RI value expected by the base station, the CSI payload size may increase, which may prevent it from being included on the PUSCH. In other words, the code rate may be too high or uncoded systematic bits may not be included. In the above-described case, instead of dropping the CSI report, the terminal can report some high-priority CSI information to the base station via PUSCH through partial CSI omission.

[0221] As an example, FIG. 7 is a diagram illustrating CSI omission applicable to the present disclosure. Referring to FIG. 7, some (710) of the CSI information may be included in the report, and the remaining part (720) may be omitted from the report. As an example, the terminal may perform ordering of CSI contents within CSI Part 2. When multiple CSI reports are transmitted on the PUSCH, wideband CSI components (i.e., wideband PMI and CQI) in all CSI reports are allocated to the most significant bit (MSB) of the UCI, and subband CSI (each CSI report) may be allocated in the manner of "even SB CSI - odd SB CSI - even SB CSI", which may be as shown in FIG. 7.

[0222] For example, if the UCI code rate exceeds a threshold, some of the UCI least significant bit (LSB) bits may be omitted. Here, the omission (720) may be performed until the code rate falls below the threshold, which may be as shown in FIG. 7. As an example, in FIG. 7, "SB CSI for odd numbered" may be omitted first, but the present invention may not be limited thereto.

[0223] For example, a base station can interpolate PMI / CQI by estimating missing PMI / CQI values ​​in the frequency domain. That is, omission can be performed based on the above-described order so that the base station can estimate the maximum interpolation for some of the omitted SB CSIs. This can maintain higher performance than omitting all consecutive SB CSIs.

[0224] For example, AI / ML (artificial intelligence / machine learning)-based operations may be performed below. The terms in Table 13 below may be used within the AI / ML-based framework, but may not be limited thereto.

[0225] [Table 13]

[0226]

[0227]

[0228]

[0229] For example, deep learning, a subcategory of machine learning (ML), can be a method that focuses on parameterizing multi-layer neural networks. Deep learning can be a method for learning representativeness of data and has demonstrated high performance in image classification, speech recognition, and natural language processing. However, deep learning can require a large amount of data during the training process, and the computational complexity required to train deep learning-based models on large data sets can be high, necessitating high hardware performance.

[0230] Regarding AI / ML, AI / ML-based mobile communications can be implemented. For example, AI has been able to reduce CAPEX and OPEX costs by managing data from devices operating on existing wireless communication systems. Furthermore, the introduction of next-generation networks may require low operating costs and high return-to-cost compensation, and AI can help address the complexity and optimization challenges of network operations. Furthermore, network intelligence and automation may be critical issues in next-generation mobile communication systems, and these challenges can be addressed using AI / ML. For example, budget management, network management, equipment lifecycle management, service level agreement (SLA) management, network performance management, and network planning can be implemented using AI / ML. AI / ML can enhance the Quality of Experience (QoE) of network operation and usage, but is not limited to this. Furthermore, AI / ML can improve network quality and enhance the provision of more personalized services, but is not limited to a specific form.

[0231] For example, AI / ML-based operations may consider two types of operations: model generation and inference operations. Model generation may be performed based on model training, model validation, model testing, and application stages. For example, model training may be performed based on at least one of input / output, pre-processing / post-processing, and online / offline application, but may not be limited thereto.

[0232] In addition, inference operations can be performed based on input / output, pre / post-processing, and application stages. A general AI / ML framework and an AI / ML model life cycle management (LCM) procedure can be performed based on the following steps. As an example, FIG. 8 is a diagram illustrating an AI / ML model life cycle management applicable to the present disclosure. Referring to FIG. 8, data collection for an AI / ML model can be performed. Here, the collected data can include data for model training, monitoring data for model management and performance monitoring, and inference data for model inference. Model training can perform model learning based on training data, and the learned and updated model can be stored. Here, the model can be stored based on a function or model identifier or an identifier related to the model and additional information, but is not limited to a specific form. In addition, model training can be performed based on model training control information based on model management and performance monitoring.

[0233] Model management and performance monitoring can control model inference based on monitoring data. Model management and performance monitoring can control the activation / deactivation / selection / switching / fallback of model inference, as well as other operations. Model management and performance monitoring can obtain information about the results of model inference. Model inference can obtain inference data, perform inference operations, and obtain result information. Model inference can receive models stored in storage and perform inference operations based on them.

[0234] In relation to the LCM (life cycle management) process, one AI / ML model may have one model ID and related information and / or model functions for AI / ML operations, as described above.

[0235]

[0236] Actions within the LCM procedure

[0237] - Data Collection

[0238] - Model Training

[0239] - Functionality / model identification

[0240] - Model transfer

[0241] - Model Inference

[0242] - Functionality / model selection, activation, deactivation, switching and fallback operation

[0243] - Functionality / model monitoring

[0244] - Model update

[0245] - Model Storage

[0246] - UE capability

[0247]

[0248] For example, AI / ML-based operations in wireless communication systems can consider the following cases. However, this is only one example, and AIML-based operations can also be applied to other cases.

[0249]

[0250] AI / ML-based operation of wireless communication systems

[0251] CSI (Channel State Information): compression (**), temporal prediction (*)

[0252] BM(Beam Management): spatial prediction (*), temporal prediction (*)

[0253] Positioning: direct (*), assisted (*)

[0254] (**) is a two-side model (*) is a single-side model

[0255]

[0256] As a concrete example, we can consider AI / ML model training collaborations for CSI compression using a two-sided model.

[0257] For example, with respect to CSI compression, AI / ML model inference can be performed on both the terminal and the base station. That is, AI / ML model inference can be performed on the two-side. To this end, a CSI generation part can be configured on the terminal, and a CSI reconstruction part corresponding to the CSI generation part can be configured on the base station. The CSI generation part is a part for CSI compression, and the CSI reconstruction part can be used to obtain (recover) more accurate CSI for better MU (multi-user) scheduling operation in a massive MIMO scenario, but is not limited thereto, and cooperation on the two-side AI / ML model can be performed in various forms.

[0258] As another example, considering the limitations of the terminal's capability or computing power, it may be considered that a third entity (e.g., a server for the terminal) performs AI / ML model training or inference operations. That is, the terminal-side operations (e.g., model training / inference / monitoring / switching / activation / deactivation / validation / testing) in the above-described AI / ML operations may include the operations of other entities and are not limited to a specific form.

[0259] For example, training collaboration can consider Types 1 through 3 below. In the following, all AI / ML model-related operations performed at the base station or network are assumed to be performed at the network level. However, this may not be limited. Since the "network" is an inclusive term that includes the base station, the two terms may be used interchangeably. What is performed at the network can encompass both the base station side and the broader network side. For convenience of explanation, the following description will be described as being performed at the network.

[0260] For example, joint training can be performed based on Type 1. Type 1 can be a method in which training for an AI / ML model is performed on one side / entity. Other sides / entities can then obtain specific model parts through downloading / uploading, and CSI feedback operations can be performed based on these.

[0261] As a specific example, FIG. 9 is a diagram illustrating a method for performing AI / ML model training based on Type 1. Referring to FIG. 9, model training for the CSI generation part of the terminal and the CSI reconstruction part of the base station can be performed on the network (920) side in a joint training manner. Thereafter, the network side can provide the trained model information to the terminal (910) through downloading and can also apply it to the network (920). As another example, model training for the CSI generation part of the terminal and the CSI reconstruction part of the base station can be performed on the terminal (910) side in a joint training manner. Thereafter, the terminal side can apply the trained model information to the terminal (910) and provide it to the network (920) through an uplink. In other words, an AI / ML model can be trained on one side / entity, and the trained AI / ML model can be transferred to another side / entity.

[0262] As a specific example, network-side model 1 training can train the CSI generation part using a data set related to CSI generation for training purposes in the network, and train the CSI reconstruction part by transferring the result value into the training loop of the CSI reconstruction part. Here, the CSI reconstruction part located in the base station can be utilized for subsequent AI / ML model-based CSI feedback. On the other hand, the CSI generation part (i.e., CSI encoder) that performed model training by the base station can transfer the trained model information to the terminal via the wireless interface (UE download AI / ML model). The terminal can perform the AI / ML model-based CSI feedback operation using the downloaded CSI generation part model. That is, CSI compression can be performed through the CSI generation part to perform CSI feedback.

[0263] On the other hand, when the AI / ML model is trained on the terminal side, terminal-side Type 1 training can train the CSI generation part using a data set of CSI generation for training purposes at the terminal. Then, the output value can be provided to the training loop of the CSI reconstruction part, and the CSI reconstruction part can be trained based on the output value. For example, the CSI generation part located at the terminal can be used to generate CSI feedback information based on the AI / ML model. The CSI reconstruction part (i.e., the CSI decoder) that performed the model training by the base station can transmit the trained model information to the base station via the wireless interface (UE upload AI / ML model). Thereafter, the base station can perform the AI / ML model-based CSI feedback operation using the CSI reconstruction part model received through the terminal upload. That is, CSI feedback can be performed by recovering CSI feedback information through the CSI reconstruction part.

[0264] Additionally, AI / ML model joint training type 1 may require AI / ML model exchange. For example, AI / ML model inference may require a common reference for model inference and a common AI / ML model inference algorithm, which may require AL / ML model exchange. For example, terminal-specific joint training type 1 may perform AI / ML model transfer based on the open format of the AI / ML model structure recognized by the terminal. On the other hand, non-terminal-specific joint training type 1 may mean model transfer within an open format without the terminal being aware of the AI / ML model structure. For example, when training a terminal-side model on the network side, the terminal-specific model may be trained on the network side based on the terminal capabilities and the terminal model structure, but may not be limited to a specific form.

[0265] As another example, FIG. 10 is a diagram illustrating a method for performing AI / ML model training based on Type 2 applicable to the present disclosure. Referring to FIG. 10 , Type 2 model training may be identical to Type 1 in that both the CSI generation part and the CSI reconstruction part are performed within a single, identical loop. However, Type 2 model training differs from Type 1 in that training is not performed on a specific side / entity, but on both sides / entities. That is, both the network and the terminal may be involved in the training.

[0266] For example, referring to FIG. 10, the terminal side can perform a forward propagation (FP) operation based on a data sample. That is, the terminal (1010) can generate a CSI feedback result and transmit the FP result as information thereon to the network side. Thereafter, the network (1020) can perform a CSI reconstruction operation based on the FP result, thereby training the CSI reconstruction part. The network side can generate BP (backward propagation) information (e.g., gradients) based on the above-described method and transmit the generated BP information back to the terminal side. The terminal side can perform training for the CSI generation part based on the BP information received from the network side. Here, the data set can be aligned on both the terminal side and the network side, and training can be performed through the above-described method.

[0267] Also, as an example, FIG. 11 is a diagram illustrating a method for performing AI / ML model training based on Type 3 applicable to the present disclosure. Referring to FIG. 11, Type 3 model training is a sequential training method, and unlike the joint training of Types 1 and 2, model training can be performed sequentially / independently for each node. That is, Type 3 model training can be performed based on the exchange of aligned data set information between a branch station and a terminal within a separate loop, without training the CSI generation part and the CSI reconstruction part within the same loop. As an example, Type 3 model training can be distinguished into network-first training or terminal-first training, similar to Type 1 model training.

[0268] Referring to Figure 11, in the case of network-first training, the network may use the CSI generation part for training the CSI reconstruction model set on the network side. However, the CSI generation part is for training purposes and may not be used in subsequent model inference.

[0269] The network (1120) side inputs the CSI generation part's input value ( ) based on the corresponding CSI generation part and its output value ( ) and the final output value ( through model training of the CSI reconstruction part based on this ) can be generated. The CSI reconstruction part for which training is completed can be used for the AI / ML-based CSI feedback operation of the network. Thereafter, the network (1120) side can share the data set information (e.g., dataset including labels and intermediate results) utilized in the model training of the network side with the terminal after the model training is completed. The terminal (1110) side can apply the data set information to the CSI generation part training. Thereafter, a joint inference operation can be performed through the CSI generation part performed on the terminal (1110) side and the CSI reconstruction part performed on the network (1120). In addition, information on the CSI generation part trained on the terminal (1110) side (e.g., model / functionality identification process - information on available CSI generation parts) can be reported or transmitted to the network (1120). The network (1120) can select or optimize a CSI reconstruction part corresponding to the CSI generation part model of the terminal based on information about the CSI generation part obtained from the terminal (1110).

[0270] As an example, options for training a terminal-side CSI generation model for network-first type 3 joint training can be considered as shown in Table 14 below.

[0271] [Table 14]

[0272]

[0273]

[0274] The terminal-side CSI output format or the base station-side CSI input format may be as shown in Table 15 below. That is, the CSI generated by the terminal by considering AI / ML and transmitted to the base station may be in the form of Table 15 below. Option 1 in Table 15 may be a case where the terminal-side CSI output format or the base station-side CSI input format is in the form of a precoding matrix similar to the existing CSI feedback. The terminal may report CSI feedback to the base station in the form of a precoding matrix together with the confirmed layer number information. For example, the terminal may report CSI feedback in the angular-delay domain as the domain of option 1-b in Table 15, but may not be limited to the embodiment. (e.g., via spatial-frequency DFT domain transformation to angular / delay domain)

[0275] On the other hand, option 2 in Table 15 may be a case where the terminal-side CSI output format or the base station-side CSI input format is a channel matrix with explicit CSI feedback. That is, the explicit CSI feedback of option 2 may not include precoding vectors assigned to each layer as a channel matrix. The explicit channel matrix may be based on channels in the space-frequency domain (option 2-a) or based on channels in the angle-delay domain (option 2-b), but is not limited to a specific form. For example, in the angle-delay domain of option 2-b, loss may be caused by sub-selection of angle-delay domain-based indices, but is not limited thereto.

[0276] [Table 15]

[0277]

[0278]

[0279] Hereinafter, a method for performing AI / ML-based CSI reporting by performing AI / ML-based CSI compression in a wireless communication system is described. Specifically, when applying AI / ML-based CSI compression in a wireless communication system, it may be a method for performing UCI (uplink control information) reporting while considering the CSI payload size. As an example, in the following, for convenience of explanation, the CSI generated based on AI / ML-based CSI compression in a wireless communication system is referred to as AI CSI, but is not limited to that name, and the same can be applied to CSI derived based on an operation based on AI / ML-based CSI compression. Here, the AI ​​CSI payload can be adjusted in consideration of AI / ML-based CSI reporting, and this is described below.

[0280] In addition, the AI-CSI payload adjustment method proposed below can also be applied as a payload adjustment method for specific information in other use cases based on AI / ML, when the terminal must transmit specific information to the base station for AI / ML model operation.

[0281] Therefore, the methods proposed below are not limited to the use case for AI-CSI reporting, and can be applied to uplink reporting in other use cases.

[0282] FIG. 12 is a diagram illustrating an AI encoder structure applicable to the present disclosure and a method for deriving output from input based thereon. Referring to FIG. 12, an AI encoder (1210) may obtain AI CSI inputs (AI CSI inputs) 1221, perform AI / ML-based CSI compression, and generate an AI encoder output (AI encoder output) 1222. The AI ​​CSI inputs (1221) may include various information related to AI / ML-based operations and may not be limited to a specific form. The AI ​​encoder (1210) may perform AI encoding for AI / ML-based CSI compression using AI / ML model-related AI CSI inputs (1221), and then perform a quantization procedure. The AI ​​encoder (1210) may derive an AI encoder output (1222) after the AI ​​encoding and quantization procedures. The AI ​​encoder output may be, but is not limited to, information in the form of a bit string or a bit sequence.

[0283] The terminal can transmit AI encoder output (1222) information for AI / ML-based CSI compression to the base station. The base station can obtain AI encoder output (1222) information from the terminal, and perform a decoding procedure on the AI ​​encoder output information through an AI decoder corresponding to a CSI reconstruction part located in the base station, thereby reconstructing information on AI / ML-based CSI inputs. For example, the above-described AI CSI information transmission procedure based on AI CSI compression can be performed in at least one of the training / inference / monitoring and other procedures within the LCM (Life Cycle Management) procedure. In addition, the above-described procedure for transmitting AI CSI information based on AI CSI compression can also be utilized for other AI model monitoring or model switching, and may not be limited to a specific form.

[0284] Here, the AI ​​encoder output (1222) may be the aforementioned AI CSI, and the terminal may report the AI ​​CSI to the base station to perform the aforementioned operation. Below, specific operations for performing AI CSI reporting are described. For example, applicable new content and transmission methods between AI CSI reports through AI / ML-based CSI compression may be as shown in Tables 16 and 17 below.

[0285] [Table 16]

[0286]

[0287]

[0288] [Table 17]

[0289]

[0290]

[0291]

[0292] For example, the information in Tables 16 and 17 may be information about combinations of various forms of AI CSI reports at each stage within the LCM procedure based on at least one of base station settings, channel conditions, terminal capabilities, and other information. AI CSI reports may be configured and set based on Tables 16 and 17. Furthermore, AI CSI reports may be configured and set considering other contents or stages other than those in Tables 16 and 17, and are not limited to a specific form.

[0293]

[0294] For example, the eigenvectors of the raw channel matrix (H) or the ray channel matrix (V) may be provided as input to the AI ​​encoder. If the raw channel matrix is ​​provided as input to the AI ​​encoder, the eigenvector calculation may be performed at the base station, which may increase the complexity of the base station. In addition, the raw channel matrix may include more unnecessary information related to precoder selection, which may increase the overhead. Considering the above-described characteristics, the AI ​​decoder performance may vary based on the AI ​​CSI content information provided as the input value of the AI ​​encoder, and considering the above-described points, there is a need to perform the AI ​​CSI reporting procedure based on a combination of appropriate input values ​​among the input values ​​in Tables 16 and 17.

[0295] For example, when an AI / ML-based CSI feedback procedure is performed, AI / ML-based quantized CSI compression information and existing CSI report / contents information may be transmitted through the same air interface. For example, the AI / ML-based quantized CSI compression information may be AI CSI information derived based on an AI encoder, and the existing CSI report / contents may be CSI information reported considering channel conditions. For example, the existing CSI report / contents may include various forms of wireless channel-related information that are not based on an AI / ML model, and may not be limited to a specific form. However, for convenience of explanation, it is referred to as existing CSI information below.

[0296] For example, AI CSI reporting fields for AI / ML-based quantized CSI compression information may differ from existing CSI reporting fields. Specifically, AI CSI information with feedback information corresponding to AI / ML-based quantized CSI compression information may be classified based on existing CSI reporting, but may be determined to have a different form of CSI content within the CSI report.

[0297] As another example, AI CSI information can be determined as a new CSI report type with feedback information corresponding to AI / ML-based quantized CSI compression information. In other words, the AI-based CSI report may be a new type of report different from the existing CSI reporting method, but may not be limited to the present embodiment.

[0298] Additionally, a new codebook type corresponding to AI-based CSI feedback reporting that includes PMI information (e.g., Type-3 codebook) may be determined to support autoencoders in a two-sided model. The new codebook type may be determined based on, but not limited to, different forms of CSI content and quantization methods newly defined to efficiently support "AI-based CSI compressed feedback."

[0299] As another example, regarding the AI ​​CSI reporting field, the base station can be configured to feed back actual training data for an AI / ML model to the terminal. If the terminal feeds back actual training data for the AI / ML model, the AI / ML CSI report transmitted by the terminal to the base station may include CSI parameters corresponding to both existing PMI information and the actual training data.

[0300] As another example, it may be possible for the AI ​​CSI reporting field to apply new CSI fields (e.g., AI-based auto-encoder / NN parameters).

[0301] As another example, the terminal may transmit information on the computational complexity to be performed on the terminal side to the base station in the form of an AI CSI report, taking into account signaling for the computational complexity value. For example, the signaling for the computational complexity value may be configured differently based on the number of CPUs, the number of AI-based CSI reports, and other information. As a specific example, the number of AI-based CSI reports may be determined differently in relation to the computational measurement corresponding to the AI-based CSI report, and thus the computational complexity value may be different. As another example, the number of AI-based CSI reports may be determined differently based on whether the terminal performs on multiple component carriers and other information, and thus the computational complexity value may be different, but may not be limited to a specific form.

[0302]

[0303] For example, in existing wireless communication systems, sufficient resources may not be provided to a terminal to transmit all CSI contents that must be reported within the time required for a single CSI report. The terminal may be provided with uplink resources based on uplink scheduling for CSI reporting, and may transmit a CSI report containing CSI contents to the base station based on these resources. Here, if the terminal is not allocated sufficient uplink resources to report all CSI contents based on the uplink scheduling provided by the base station, the terminal may omit some of the CSI contents that must be reported and report only the remaining CSI contents. In other words, the terminal may perform CSI omission. As a specific example, consider a case where the terminal performs Type 2 CSI reporting as described in Table 7, where the RI is 2. The terminal may perform CSI reporting based on a limited number of layers considering support for MU-MIMO scheduling. Here, when the RI is 2, the PMI payload may be approximately twice as large as when the RI is 1. In addition, since the base station cannot recognize the RI selected in advance by the terminal, the base station can perform scheduling based on the RI value expected to be selected by the terminal when performing PUSCH scheduling. Considering the above, misalignment in RI selection between the base station and the terminal may occur, and in the above case, the terminal cannot transmit all of the CSI payload on the PUSCH. For example, information with a high code rate or 'uncoded systematic bits' may not be transmitted on the PUSCH. Therefore, some CSI contents in the CSI report may be dropped or omitted. However, some important CSI parts (i.e.If the contents) are dropped, performance degradation may occur throughout the entire system, and considering the above, the CSI part 2 for the "eType II codebook (CB)" can be configured to be divided into three groups as described above. Here, based on the CSI group, information corresponding to important CSI parts among the CSI parts can be given a higher priority and transmitted on the PUSCH.

[0304]

[0305] For example, AI / ML-based CSI compression feedback may not include 'non-zero coefficients', 'SD basis', 'FD basis', 'indication of non-zero coefficients', etc., unlike the above-described "eType II Code Book (CB)". That is, with AI / ML-based CSI compression feedback, AI CSI information may mean compressed quantization information rather than CSI-related information itself. Specifically, in the CSI generation part of the AI / ML model for CSI compression feedback, AI encoding and quantization procedures may be performed based on AI CSI inputs, and thus an AI encoder output may be derived. When sufficient uplink resources are not allocated to the terminal for AI / ML CSI reporting, the terminal may selectively transmit some bit information to the base station based on the importance of the quantized bit information as the AI ​​encoder output. For example, if AI CSI reporting is to be performed in at least one of the model training and model inference procedures, the terminal may selectively transmit some quantized bit information to the base station based on the importance of the quantized bit information.

[0306] The following describes a case where, when the uplink resources allocated to a terminal by a base station are not sufficient to transmit all AI CSI information, a UCI (or CSI) payload that at least includes the corresponding AI CSI information is adjusted for AI CSI information transmission. As another example, the following describes AI CSI omission, in which some of the determined quantized CSI information bits are omitted based on the quantized CSI bit subsequence, priority, and other information, which will be described later.

[0307] For example, operations based on options 1 to 4 of Table 18 may be performed, but may not be limited thereto. Specifically, the terminal may adjust the payload size of the AI ​​encoder output by variably setting the quantization resolution in relation to the quantization procedure within the AI ​​encoder. (Option 1) That is, the terminal may control at least one of the quantization type, level, and pattern in consideration of the uplink resource size within the AI ​​encoder. In addition, quantization alignment for the base station and the terminal may be performed based on the above. As a specific example, in option 1 of Table 18, quantization may be a method of adjusting the quantization level according to AI CSI content. At least one of the quantization method and the bit string for quantization may be applied differently according to the priority among AI CSI contents. In addition, a different CSI payload size may be determined for each CSI generation part, but may not be limited thereto.

[0308] As another example, the output size of the terminal-side AI encoder may be fixed in consideration of other CSI parameters. (Option 2) The AI / ML-based CSI feedback compression level may be adjusted within the AI ​​encoder so that the AI ​​encoder output size does not change depending on channel characteristics or RI values ​​selected by the terminal, but is not limited to this embodiment.

[0309] As another example, AI / ML-based CSI compression may be performed by dividing the AI ​​CSIs into multiple groups with different priorities based on at least one of layers, ranks, sub-bands, and ports. (Option 3) Here, among the AI ​​CSIs derived as AI encoder outputs based on AI / ML-based CSI compression, transmission of AI CSIs corresponding to groups with lower priorities may be omitted.

[0310] As another example, a single AI CSI report for AI CSI information may be performed by being separated into multiple sub-reports (Option 4). Here, each sub-report may be reported at a different transmission time due to limited uplink resources, and it is necessary to determine the relationship between previously transmitted sub-reports and subsequently transmitted sub-reports by taking into account the above-described point. As an example, multiple sub-reports related to a single AI CSI report may be reported at different times by considering uplink transmission resources, and through the above-described point, AI / ML-based CSI compression information may not be omitted.

[0311] [Table 18]

[0312]

[0313]

[0314] The options proposed in Table 18 can be basically assumed to be applied to AI / ML-based CSI compression procedures based on a combination of at least one option on the AI / ML encoder side on the terminal side. However, they can also be applied to other use cases based on a "two-sided AI / ML model" between the base station and the terminal (a model in which AI / ML models are located on the base station and the terminal side, and operate based on mutual compatibility). In addition, for example, at least one of the encoding and quantization procedures can be performed within the encoder, but depending on the implementation, the quantization procedure may be performed independently of the encoding procedure.

[0315] Quantization alignment can be performed between the base station and the terminal. Here, quantization alignment can be performed for adaptive CSI payload size adjustment, thereby enabling efficient AI CSI reporting under limited uplink resources.

[0316] Specifically, the base station needs to know in advance the size of the quantized AI CSI information reported by the terminal for decoding the AI ​​CSI information using the CSI generation part output value transmitted from the terminal. If the quantization method performed by the terminal is not aligned with that of the base station, the base station cannot successfully decode the terminal-side CSI generation part output value. Alternatively, if the base station does not know in advance the quantization method of the terminal and the parameter setting information therefor, the base station cannot properly perform quantization alignment with the terminal and cannot successfully decode the terminal-side CSI generation part output value. That is, quantization alignment needs to be performed in a situation where the encoder output of the terminal-side CSI generation part is transmitted as the input of the base station-side CSI reconstruction part.

[0317] FIG. 13 is a diagram illustrating a method for performing scalar quantization in an AI encoder applicable to the present disclosure. Referring to FIG. 13, for quantization alignment, a base station-side decoder may receive a bit string or bit sequence derived by Equation 9 below from a terminal, taking into account the quantization procedure. In Equation 9, S may be a bit string or a bit sequence, and may be an output value of a terminal-side AI encoder. Here, K may be the number of nodes in the encoder output layer, which may be the encoder output size. In addition, Q may be the number of quantization bits per node. That is, the size of the AI ​​encoder output value may be determined by the product of the sizes of the output nodes and the number of quantization bits per node.

[0318] [Equation 9]

[0319]

[0320]

[0321] For example, when an inference procedure is performed based on AI / ML, the number of bits used in a CSI report can be minimized through quantization, thereby reducing uplink resource consumption. Referring to FIG. 13, a terminal-side encoder and a base station-side decoder can recognize in advance a bit string (or bit sequence) S (1330) exchanged between the terminal and the base station. For example, the bit string S (1330) can be fixed to a value set in advance. As another example, the bit string S (1330) can be a value signaled to the terminal by the base station or a value determined based on the signaling, but is not limited to a specific form. The bit string S (1330) is determined by the output size (or size, K, 1310) of the encoder and the number of quantization bits per node (Q, 1320), and thus may be determined differently based on one or more combinations of the output size K (1310) of the encoder and the number of quantization bits per node Q (1320). For example, the above-described quantization nodes may correspond to each or some combinations of different AI-CSI contents.

[0322] As a specific example, if one output node #0 corresponds to the L1-RSRP value, the Q value (e.g., 7 bits) and the range of the Q value (e.g., [-140, 44] dBm with 1 dB step) corresponding to the node can be determined. Here, if another output node, Output node #1, corresponds to the RI value, the Q value and the range of the Q value corresponding to the node can be determined independently. Accordingly, the entire bit string S value can be determined according to the reported combination and quantization method considering the reported AI CSI contents. However, the above-described case is only one example and may not be limited thereto. The quantization method applied according to all possible AI CSI and parameters / contents for CSI can be applied according to the proposed bit string determination method.

[0323] For example, even if the bit string S (1330) is recognized in advance, it may be difficult for the base station and the terminal to accurately derive the encoder output size K (1310) and the number of quantization bits per node Q (1320). As a specific example, in the case where the encoder part and the decoder part have different structures in the CSI compression cooperative type 2 and type 3 training, the base station and the terminal may not recognize the encoder output size K (1310) and the number of quantization bits per node Q (1320) even if they recognize the bit string S (1330) through quantization alignment. As a more specific example, since there may be multiple combinations of "S = K1xQ1 = K2xQ2" for a fixed bit string S (1330), the base station and the terminal need to recognize not only the bit string S (1330) but also the encoder output size K (1310) and the number of quantization bits per node Q (1320), and quantization alignment can be performed based on this.

[0324] However, in cases where the quantization procedure does not operate in the above-described manner, only the "S" value may be indicated by at least one of a pre-set value between the base station and the terminal, an upper layer setting, and physical layer signaling (L1 signaling), and is not limited to a specific form.

[0325] For example, for quantization alignment between a base station and a terminal, the base station may, taking into account capability information of the terminal, instruct the terminal through upper layer signaling a setting (or a quantization level related parameter setting) related to at least one of the encoder output size K and the number of quantization bits per node Q for AI encoder quantization operation. The terminal may operate according to the instructed setting information based on the base station signaling, or may apply a quantization level to the AI ​​encoder based on selection of a preferred value in the terminal AI / ML model within a limited range of values ​​based on the instructed information. Thereafter, the terminal may additionally transmit the selected quantization level related information (or parameter) to the base station together with the AI ​​CSI report in the form of side information.

[0326] As another example, it may be possible for a setting (or a quantization level related parameter setting) related to at least one of the encoder output size K and the quantization bits per node Q for the AI ​​encoder quantization operation to be preset between the base station and the terminal. The terminal may operate according to the preset information or apply a preferred quantization level to the AI ​​encoder based on terminal selection within a limited range based on the preset information. Thereafter, the terminal may transmit the quantization level related information (or parameter) to the base station together with the AI ​​CSI report in the form of side information. As another example, it may be possible for the S value, rather than the K and Q values ​​as described above, to be preset between the base station and the terminal.

[0327] As another example, a quantization method may be shared between an AI encoder and an AI decoder. In the case of the collaboration type 1 described above, information about the quantization method may be shared during the model transfer process. For example, collaboration type 1 may be a type in which an AI / ML model is determined by one entity and the model is transferred to another entity, and information about the quantization method may be transferred together with the AI / ML model transfer. As a specific example, when a base station performs modeling, the base station may transfer information about the quantization method together during the model transfer process to the terminal. Alternatively, when a terminal performs model training based on the collaboration type model 1, the terminal may transfer information about the quantization method together during the model transfer process to the base station, but the present invention is not limited to this embodiment. As another example, in the case of collaboration type 2 or collaboration type 3, information about the quantization method may be shared during the initial information exchange between the terminal and the network, but the present invention is not limited to this embodiment.

[0328] Furthermore, if each terminal vendor uses its own unique quantization method for quantization alignment, the complexity of processing CSI reports on the network side can be significant. Furthermore, if each terminal vendor uses its own unique quantization method, storage processing and switching delays can occur. Considering the above, the base station can determine the quantization method and perform AI model training on the network side.

[0329]

[0330] As described above, when the quantization method is dynamically determined, the CSI payload size can be adjusted as an output of the AI / ML model-based CSI generation part on the terminal side. For example, at least one of the encoder output size K and the number of quantization bits per node Q of the CSI generation part of the AI / ML model can be changed through quantization adjustment, and accordingly, the size of the AI ​​CSI bit string S can be adjusted differently. That is, based on Equation 9, S can be adjusted based on at least one of the K and Q described above, or directly without using the K and Q values.

[0331] However, even when performing the signaling as described above for quantization alignment between the base station and the terminal, it is necessary to be able to perform different quantization procedures depending on the degree of freedom in the terminal implementation. Therefore, rather than sharing a specific quantization model (e.g., scalar quantization or vector quantization) and related parameters (e.g., Q, K, S, etc. above), a method may be needed to define a quantization level and set a specific quantization model and related parameters associated with the quantization level. Table 19 below divides the quantization levels into 'High', 'Medium', and 'Low', and the base station can provide signaling according to Table 19 to the terminal. The terminal can perform the quantization procedure based on the value signaled from the base station. However, Table 19 below is just one example for the convenience of explanation and may not be limited thereto.

[0332] [Table 19]

[0333]

[0334]

[0335] As a specific example, the terminal may set or receive in advance from the base station the quantization models and related parameter values ​​associated with each of the scalar quantization levels ('High', 'Medium', and 'Low') in Table 19. The terminal may adjust the CSI payload size by performing a quantization procedure based on the provided values. Alternatively, the terminal may adjust the CSI payload size by performing a quantization procedure determined by the terminal's selection based on the provided values. Table 19 may be an example of setting and applying a quantization method based on a scalar quantization level, and may be set based on a different quantization model. In addition, it may be possible for parameter values ​​for different set quantization models to be mapped to the same quantization level as in the above table and set and indicated, and is not limited to a specific form.

[0336] Based on the above, even if multiple quantization models exist, the combination of relevant parameters can be restricted, simplifying quantization implementation between the base station and the terminal. Furthermore, for example, the use of AI models limited to specific quantization models and parameters can increase implementation complexity due to the need to store multiple AI models between the base station and the terminal, and can also lead to increased costs. The above-described approach allows for the adjustment of the CSI payload size while reducing the implementation burden and increased costs.

[0337] As another example, rather than defining the quantization level as in Table 19, the quantization model and related parameter values ​​may be associated and matched based on at least one of the AI ​​Model / Feature ID / Associated ID used to distinguish the AI ​​model, feature, and network condition within the LCM procedure and the AI ​​encoder index used within the corresponding model, as in Table 20. The terminal may obtain information regarding Table 20 from the base station through upper layer signaling. In addition, the information regarding Table 20 may be preset information, but is not limited thereto. The terminal may adjust the CSI payload size by performing the quantization procedure based on the provided value. Alternatively, the terminal may adjust the CSI payload size by performing the quantization procedure determined by the terminal's selection based on the provided value. That is, the CSI payload size associated with the AI ​​CSI report transmitted through the uplink resource may be adjusted by adaptively utilizing at least one of the AI ​​and quantization model / parameter related information, the CSI content, and the associated priority information.

[0338] [Table 20]

[0339]

[0340]

[0341] That is, the terminal may adjust the CSI payload size by performing a quantization procedure corresponding to quantization-related information or quantization-related parameters signaled or preset from the base station, or by performing a quantization procedure determined by the terminal's selection based on the preset quantization-related information or quantization-related parameters. Here, the above-described Tables 19 and 20 may be only examples, and the terminal may perform a quantization procedure based on different information (or parameters).

[0342]

[0343] AI / ML-based CSI content can be configured or configured with one or more CSI parts. These configurations or settings can be shared between the base station and the UE through base station configuration or UE capability reporting, or determined as preset values. While the AI ​​CSI report below is described based on the use of one or two AI-based CSI parts, this is for convenience of explanation and may not be limited to this. In other words, it may be possible for an AI CSI report to use two or more CSI parts.

[0344] In addition, for the convenience of explanation, the following describes the related contents based on AI CSI reports consisting of one CSI part and AI CSI reports consisting of at least two CSI parts. However, this is a configuration for the convenience of explanation and may not be limited thereto. For example, depending on which model type (e.g., layer-common model, layer-specific model, rank-common model, or rank-specific model) is applied for CSI compression based on a two-sided AI / ML model, either AI CSI reports consisting of one CSI part or AI CSI reports consisting of at least two CSI parts may be used, but are not limited thereto. For example, when only one CSI part is used, the base station and the terminal may determine the overall output CSI payload size of the AI / ML model to be fixed or preset depending on the model through quantization alignment. That is, the size of the AI ​​encoder output sequence may not be changed by the channel characteristics or the 'RI' value, 'number of non-zero coefficient values', 'number of SD' and 'number of FD basis' selected by the terminal.

[0345]

[0346] Hereinafter, a method is described to adjust the UCI (CSI) payload size that can be transmitted to a terminal through uplink resources allocated by a base station by applying different quantization methods based on priorities. As another example, the quantization method may be applied according to a method implemented by the terminal without considering the priorities, but is not limited thereto. Unlike the existing CSI reporting method, the AI-based compressed CSI (i.e. UCI) reporting procedure can report the AI ​​encoder output derived through the AI ​​encoding and quantization procedures within the AI ​​encoder as AI CSI. As an example, the quantization procedure can be applied by the terminal implementation through various implementation methods. Quantization alignment can be performed in advance by the base station and the terminal so that the receiving side (i.e. the base station decoder) can accurately decode the corresponding AI-based compressed CSI information.

[0347] Hereinafter, a method of controlling the AI-based UCI (CSI) payload size is described, in which an independent number of quantization bits Q is used for each encoder output node, node group, or sequence group among multiple Qs (quantization bits per node) that are preset. As described above, the overall AI encoder bit string size can be controlled by priority or preset rules. The terminal can adaptively determine the UCI payload size by considering the amount of uplink resources allocated by the base station to perform UCI reporting through the uplink resources. Here, the information determined (or selected) by the terminal can be transmitted to the base station as side information together with the AI ​​compressed CSI report, and the base station-side AI decoder can perform decoding using the same.

[0348] The methods described below can operate based on the signaling and procedures proposed for the quantization alignment method described above. As a specific example, the priority-related information shown in Table 17 and the parameter settings for quantization alignment in Tables 18 and 19 can be applied to the method described below, but are not limited thereto.

[0349]

[0350] FIG. 14 is a diagram illustrating a method for adjusting the AI ​​CSI payload size according to an adaptive quantization method within an AI encoder applicable to the present disclosure.

[0351] For example, a case may be considered where the output size K of the AI ​​encoder is determined in advance, and the number of quantization bits per node Q is adaptively determined. Referring to FIG. 14, the output size (K, 1410) of the AI ​​encoder may be determined as 4, and the number of quantization bits per node (Q, 1420) may be determined as Q1 or Q2. However, FIG. 14 is merely an example for convenience of explanation and is not limited thereto. Here, among the output nodes of the AI ​​encoder, an output node having a priority of high CSI contents may be determined to have a larger quantization bit (Q1) value, and an output node having a lower priority may be determined to have a smaller quantization bit (Q2) value than Q1. That is, the quantization bit value may be determined differently based on the priority of the AI ​​encoder output node, and the CSI payload size may be adjusted as described above.

[0352] As a specific example, in FIG. 14, the AI ​​encoder output size (K, 1410) may be determined based on at least one of the number of ranks, layers, or other CSI contents. The Q value may be individually applied to each AI encoder output node. As another example, after the Q value is individually applied to each AI encoder output node, the K value may be adaptively readjusted for the purpose of UCI size adjustment, but the present invention is not limited thereto.

[0353] Multiple Q values ​​with quantization bits per node can be indicated by base station signaling or can be preset values. The terminal can determine each output node where Q1 and Q2 are used among the multiple Q values. For example, the terminal can determine the Q value of the output node based on at least one of the priority information (e.g., layer information, sub-band information, antenna port information, number of ranks, AI compressed CSI content, quantization granularity, latent space size, and combinations thereof) of Table 17 described above.

[0354] After that, output nodes corresponding to each sequence group can be determined according to the quantization operation settings and decisions that are each mapped to the determined priorities, and a quantized bit sequence can be generated from each output node. The quantized bit sequence generated through each output node can pass through a multiplexer / controller to generate a sequence bit (S, 1430) as the output of the final AI encoder.

[0355] Also, as an example, the terminal may determine the Q value of the output node based on the terminal implementation information and / or the configuration / pre-configuration from the base station without considering the priority. In other words, which CSI content-related information corresponds to the Q value of each output node may be left to the terminal implementation. As an example, in FIG. 14, output node #0 and output node #1 may be included in sequence group 1 corresponding to Q1, and the remaining output nodes #2 and #3 may be included in sequence group 2 corresponding to Q2. Here, different quantization may be performed for each sequence group, and based on this, the final output sequence bits (S, 1430) may be determined, and the CSI payload size may be adjusted.

[0356] For example, a bit string corresponding to an output node may be associated with a single transmission layer. As another example, a bit string corresponding to an output node may correspond to all layers, not just the transmission layer, but may not be limited thereto.

[0357] Table 21 below shows priority-based quantization bit adjustment when performing AI-based CSI payload size adjustment. As described above, a different number of quantization bits may be determined for each output node of the AI ​​encoder, and Table 21 shows that CSI payload adjustment may be possible on a subsequence group basis. Specifically, the terminal may allocate the number of quantization bits Q1 to an output node included in sequence group 1 as a subgroup of AI encoder outputs corresponding to a high priority, considering the uplink resources allocated from the base station, and may allocate the number of quantization bits Q2 to an output node included in sequence group 2 as a subgroup of AI encoder outputs corresponding to a low priority. As described above, a bit string (S) may be adjusted as an AI encoder output, and the terminal may perform AI CSI reporting through uplink resources based on the adjusted bit string (S).

[0358] [Table 21]

[0359]

[0360]

[0361] FIG. 15 is a diagram illustrating a method for adjusting an AI CSI payload size according to an adaptive quantization method in an AI encoder applicable to the present disclosure. As an example, a case may be considered where the output size K of the AI ​​encoder is determined in advance and the number of quantization bits per node Q is adaptively determined. Referring to FIG. 15, the output size (K, 1510) of the AI ​​encoder is 4, and the number of quantization bits (Q, 1520) may be determined as Q1, Q2, or Q3, respectively. That is, in FIG. 15, the number of quantization bits per node and the number of sequence groups corresponding thereto may be distinguished in more detail than in FIG. 14. As an example, among the output nodes of the AI ​​encoder, an output node with a high priority may have a value of the largest quantization bit (Q1), an output node with the next priority may have a value of a quantization bit (Q2) smaller than Q1, and an output node with the last priority may have a value of a quantization bit (Q3) smaller than Q2. That is, the quantization bit value can be adjusted based on the priority of the AI ​​encoder output node, thereby reducing the CSI payload size.

[0362] For example, in FIG. 15, the AI ​​encoder output size (K, 1510) may be determined based on at least one of the rank number, layer, or other CSI content-related information. Here, the Q value may be individually applied to each AI encoder output node. As another example, after the Q value is individually applied to each AI encoder output node, the K value may be adaptively readjusted for the purpose of UCI size adjustment, but may not be limited thereto.

[0363] Among the multiple Q values, Q1, Q2, and Q3 can be determined by base station signaling or can be pre-determined. The terminal can determine the respective output nodes where Q1, Q2, and Q3 are used among the multiple Q values. For example, the terminal can determine the Q value of the output node based on at least one of the priority information (e.g., layer information, sub-band information, antenna port information, number of ranks, AI compressed CSI content, quantization granularity, latent space size, and combinations thereof) of Table 17 described above.

[0364] Output nodes corresponding to each sequence group can be determined based on the quantization operation settings and decisions mapped to each determined priority. Thereafter, the quantized bit sequence generated through each output node can be generated as the output sequence bits (S, 1530) of the final AI encoder through a multiplexer / controller.

[0365] Also, as an example, similar to the above, the terminal can determine the Q value of the output node based on the terminal implementation information without considering the priority. In FIG. 15, output node #0 and output node #1 may be included in sequence group 1 corresponding to Q1, output node #2 may be included in sequence group 2 corresponding to Q2, and output node #3 may be included in sequence group 3 corresponding to Q3. Here, a different quantization procedure may be performed for each sequence group, and the final output sequence bits (S, 1530) may be determined based on this, so that the CSI payload size may be adjusted. As an example, a bit string corresponding to an output node may be associated with one transport layer. As another example, a bit string corresponding to an output node may correspond to all layers in common, not just to a transport layer, but may not be limited thereto.

[0366] Table 22 below may represent priority-based quantization bit adjustment when performing AI-based CSI payload size adjustment. As described above, a different number of quantization bits may be determined for each output node of the AI ​​encoder, and Table 22 may allow CSI payload adjustment on a subsequence group basis. Specifically, the terminal may allocate the number of quantization bits Q1 to an output node included in sequence group 1 as a subgroup of AI encoder outputs corresponding to a high priority, considering the uplink resources allocated from the base station, the number of quantization bits Q2 to an output node included in sequence group 2 as a subgroup of AI encoder outputs corresponding to the next priority, and the number of quantization bits Q3 to an output node included in sequence group 3 as a subgroup of AI encoder outputs corresponding to the last priority. As described above, a bit string (S) may be adjusted as an AI encoder output, and the terminal may perform AI CSI reporting through uplink resources based on the adjusted bit string (S).

[0367] [Table 22]

[0368]

[0369]

[0370]

[0371] FIG. 16 is a diagram illustrating a method for adjusting an AI CSI payload size according to an adaptive quantization method within an AI encoder applicable to the present disclosure. Referring to FIG. 16, unlike FIGS. 14 and 15, the AI ​​encoder output size K (1610) can be adjusted, and accordingly, the CSI payload size can also be adjusted. For example, a terminal can adjust the output node (or layer) size by considering the uplink resources allocated by the uplink grant that schedules UCI (CSI) transmission, thereby adjusting the CSI payload size to suit the uplink resources.

[0372] FIG. 17 is a diagram illustrating a method for adjusting an AI CSI payload size according to an adaptive quantization method in an AI encoder applicable to the present disclosure. Referring to FIG. 17, the AI ​​encoder output size (K, 1710) and the number of quantization bits (Q, 1720) are not adjusted as parameters considered for the quantization procedure, and the CSI payload size can be determined by predetermined K (1710) and Q (1720). Here, AI UCI puncturing can be applied to the AI ​​encoder output node (layer) to adjust the CSI payload size based on the predetermined K (1710) and Q (1720). For example, the terminal can perform AI UCI puncturing by considering uplink resources allocated by an uplink grant that schedules UCI (CSI) transmission. Through this, the CSI payload size can be adjusted so that the bit string S can be transmitted through the uplink resource as the AI ​​encoder output. In Fig. 17, the terminal can puncture the portions for output node #2 and output node #3 among output node #0 to output node #3 considering the allocated uplink resource, thereby limiting the UCI payload size within the uplink resource. For example, which AI UCI information corresponding to which output node is to be punctured can be determined based on the priority of the corresponding UCI (CSI) content described above.

[0373]

[0374] As another example, an AI encoder may perform a quantization process for generating a CSI payload based on specified K and Q values. Here, the terminal may use multiple AI encoders, and may select and use at least one AI encoder among the multiple AI encoders.

[0375] FIG. 18 is a diagram illustrating a method for adjusting an AI CSI payload size based on AI encoder switching applicable to the present disclosure. Referring to FIG. 18, a terminal may use two AI encoders and may select and use at least one of the two AI encoders. However, this is merely for convenience of explanation and may not be limited thereto. In FIG. 18, a first AI encoder (1810) may perform quantization based on an S1 value determined according to specified Q1 and K1 values, and a second AI encoder (1820) may perform quantization based on an S2 value determined according to specified Q2 and K2 values. Here, the terminal may determine AI CSI reporting by considering the size of uplink resources allocated according to an uplink grant that schedules UCI and the CSI payload to be reported. Specifically, the terminal may select at least one AI encoder among a plurality of AI encoders by considering the CSI payload size. For example, the terminal may perform selection for a specific encoder based on the number of AI encoder models and the terminal capability, taking into account the AI ​​encoder model. Referring to FIG. 18, the terminal may select one of two AI encoders (1810, 1820) considering the uplink resource, and transmit the CSI payload determined by the bit string S for the selected AI encoder to the base station through the uplink resource. For example, the priorities may be different for each of the multiple AI encoders configured in the terminal, and the terminal may first transmit the bit string for the AI ​​encoder with the highest priority, but is not limited thereto. The priorities are based on information related to the AI ​​CSI content described above, and the AI ​​encoder that processes the information is related to the corresponding AI CSI content, and if there are multiple AI CSI contents, the priority for the AI ​​encoder may be determined based on the highest or lowest priority among them.

[0376] The terminal can transmit the bit string S for the selected AI encoder to the base station via uplink resources, then switch the AI ​​encoder, and transmit the bit string of another AI encoder to the base station via uplink resources allocated to the terminal at a later time. That is, the bit string for each AI encoder can be transmitted in a TDM (time division multiplexing) manner at different times, but is not limited thereto.

[0377] As another example, if the allocated uplink resources are sufficient, the terminal may configure the bit strings of all of the multiple AI encoders as AI encoder outputs and transmit them together, and may not be limited to a specific form. In addition, as an example, the terminal may limit the number of AI encoder models in consideration of the implementation complexity of the terminal. Alternatively, the base station may set the number of AI encoder models for the terminal based on terminal capability information, and is not limited to a specific embodiment. The sequence allocation order between such multiple AI encoders may be determined by a predetermined method or from base station settings. For example, the sequence allocation order between the AI ​​encoders may be determined based on at least one of the encoder ID value order, the encoder priority order, the encoder configuration index order, the associated AI-CSI content, or the AI-CSI priority order.

[0378] In addition, although the description is based on the case where the quantization-related parameters (Q, K, S) of the AI ​​encoder in FIG. 18 are preset, the present invention is not limited thereto, and it may be possible for the quantization-related parameters (Q, K, S) to be adjusted for each AI encoder. That is, the methods of FIGS. 13 to 17 for adjusting the AI ​​CSI payload size by adjusting the quantization-related parameters (Q, K, S) within the quantization procedure can be combined and used with the methods of FIG. 18 and the following. For example, a method applied based on one AI encoding can be applied to each AI encoding in FIG. 18 and utilized in combination. That is, the methods based on each drawing of the method below can be combined and used, and are not limited to a specific form.

[0379] FIG. 19 is a diagram illustrating a method for adjusting an AI CSI payload size through one or more AI encoder combinations applicable to the present disclosure. Referring to FIG. 19, an AI encoder can be independently configured based on AI CSI content associated with at least one of a layer, a sub-band, and a port. To this end, the terminal and the base station need to be aware of configuration information associated with the AI ​​encoder model structure in advance, and the base station can instruct the terminal through upper layer signaling or can be preset. In addition, as an example, the base station and the terminal can select a preferred configuration within the information instructed based on upper layer signaling, but is not limited thereto.

[0380] The terminal can transmit the above-described configuration information to the base station in the form of side information along with the AI-CSI report, and the base station can perform AI CSI decoding using the configuration information.

[0381] For example, when uplink resources are insufficient, AI CSI information encoded through each independent AI encoder may be omitted based on the subsequence of the AI ​​encoder with the lowest priority. As another example, when uplink resources are insufficient, AI CSI information encoded through each independent AI encoder may be dropped based on the subsequence from the AI ​​encoder with the lowest priority. As another example, when uplink resources are insufficient, AI CSI information encoded through each independent AI encoder may be punctured based on the subsequence from the AI ​​encoder with the lowest priority. That is, the terminal may consider the allocated uplink resources to preferentially allocate and transmit the subsequence of the AI ​​encoder with a higher priority, and may not transmit the subsequence of the AI ​​encoder with a lower priority.

[0382] Referring to FIG. 19, each AI encoder (1910, 1920, 1930) can generate its own subsequence. For example, each AI encoder (1910, 1920, 1930) can be configured to correspond to at least one of a layer, a subband, and a port, and thus can generate its own subsequence based on different CSI content information. In FIG. 19, AI encoder #0 (1910) can generate a subsequence with the highest priority, and AI encoder #N-1 (1930) can generate a subsequence with the lowest priority. However, this is for convenience of explanation and is not limited thereto. Here, the subsequence with the lowest priority can be subject to at least one of omission, drop, and puncturing to adjust the CSI payload size according to the uplink resources allocated to the terminal.

[0383] As another example, there may be multiple AI encoders associated with at least one of a layer, a sub-band, and a port, and the AI ​​CSI quantized bit sequences associated with each of the multiple AI encoders may have the same priority, but may not be limited to a specific form.

[0384]

[0385] As another example, FIG. 20 is a diagram illustrating a method for adjusting the AI ​​CSI payload size through one or more AI encoder combinations applicable to the present disclosure. FIG. 20 illustrates that, unlike FIG. 19, each AI encoder (2010, 2020, 2030) may be mapped based on AI CSI content. In FIG. 19, an AI encoder corresponding to at least one of a layer, a sub-band, and a port performs AI encoding, but FIG. 20 may be a method for performing different AI encodings based on AI CSI content generated according to at least one of a common layer, a sub-band, and a port.

[0386] As a specific example, ground-truth CSI information, channel eigenvector / eigenvalue, and target CSI information may be AI contents with the highest priority, and the corresponding AI encoder may have the highest priority. The subsequence of the AI ​​encoder corresponding to the AI ​​CSI contents with the highest priority may be mapped to the most significant bit (MSB) in the AI ​​encoder output sequence, and AI CSI contents with lower priorities may be mapped toward the least significant bit (LSB). In other words, the subsequence corresponding to the AI ​​CSI contents with the highest priority may be mapped with priority.

[0387] For example, priorities among AI CSI contents may be determined, and different AI encodings may be performed based on the determined priorities. Here, the quantization methods may be the same or independent of each other, but are not limited to a specific form. When the same quantization method is applied, the terminal may drop subsequence(s) with lower priorities to adjust the CSI payload size. On the other hand, when independent quantization methods are applied to each, the terminal may control the proposed quantization degree to be adjusted according to the AI ​​CSI contents and included in the CSI payload, but is not limited to the embodiment.

[0388]

[0389] FIG. 21 is a diagram illustrating a method for adjusting an AI CSI payload size through one or more applicable AI encoder combinations applicable to the present disclosure. AI CSI content may include content corresponding to CSI Part 1 and CSI Part 2. Here, CSI Part 1 and CSI Part 2 are for convenience of explanation and may not be limited to the corresponding names. AI CSI content information corresponding to CSI Part 1 and AI CSI content information corresponding to CSI Part 2 may each independently perform AI CSI encoding. For example, CSI parts may be divided into CSI Part 1 having a fixed size and CSI Part 2 having a variable size. Here, the overall size of CSI Part 2 may be indicated through a specific part (or field) within CSI Part 1 having a fixed size. For example, parameters indicating the size of CSI Part 2 according to quantization characteristics may be provided within CSI Part 1. As another example, the CSI part 2 size information may be determined based on the AI / ML model ID information, feature information, and other information paired between the terminal and the base station, and K, Q, and S may be determined as described above. For example, referring to FIG. 21, the subsequences of the AI ​​encoders (2111, 2112, 2113) corresponding to the AI ​​CSI content corresponding to CSI part 1 may be assigned to the AI ​​encoder output with priority over the subsequences of the AI ​​encoders (2121, 2122, 2123) corresponding to the AI ​​CSI content corresponding to CSI part 2. In addition, within CSI part 1, each AI encoder may be configured according to the AI ​​CSI content, and thus N subsequences may be separated. Here, the N subsequences may be given respective priorities according to the priorities of the AI ​​CSI content.In addition, within CSI Part 2, each AI encoder may be configured according to at least one of a layer, a rank sub-band, and a port, and thus N sub-sequences may be separated. Here, priorities may also be assigned to each of the N sub-sequences, as described above. Sub-sequences for CSI Part 1 may be allocated first in the AI ​​encoder output based on the priorities, and sub-sequences for CSI Part 2 may be allocated thereafter based on the priorities. If the allocated uplink resources are insufficient, the terminal may omit and not transmit some of the sub-sequences with lower priorities. In FIG. 21, transmission of some sub-sequences within CSI Part 2 may be omitted in consideration of the uplink resources allocated to the terminal, but this is only one example and is not limited thereto.

[0390] FIG. 22 is a diagram illustrating a method for adjusting the AI ​​CSI payload size through one or more applicable AI encoder combinations applicable to the present disclosure. Referring to FIG. 22, CSI Part 1 may be generated as a single subsequence by outputs generated by AI encoders (2211, 2212, 2213) corresponding to each AI CSI content. On the other hand, CSI Part 2 may be generated as subsequences by AI encoders (2221, 2222, 2223) corresponding to each AI CSI content, and each subsequence may be allocated after CSI Part 1 according to priority. If the terminal does not have sufficient uplink resources to allocate both CSI Part 1 and CSI Part 2, the terminal may omit transmission of some subsequences of CSI Part 2 having a lower priority.

[0391]

[0392] In addition, the AI ​​encoder output in the terminal as an AI / ML model generation part can perform at least one of CSI omission, drop, and puncturing based on a subsequence (or a subsequence group), and partial decoding can be possible based on this. For example, a bit string corresponding to the AI ​​encoder output can be divided into N subsequences (N groups). The subsequences can include only information corresponding to the first (RI / 2) layers. As another example, the first subsequence can include only the MSB of the quantized output sequence based on scalar quantization, and the second subsequence can include the remaining LSBs, and the allocation order of the CSI fields can be as shown in Table 23 below.

[0393] [Table 23]

[0394]

[0395]

[0396]

[0397] In Table 23, the subsequence with the highest priority is determined as the first group, and the Nth subsequence group with a lower priority can be formed in the same order as Table 23. If the scheduling resources allocated based on the subsequences in Part 2 are insufficient, the terminal may not transmit the subsequence with a lower priority. That is, the terminal may perform at least one of CSI omission, dropping, and puncturing for the subsequence with a lower priority.

[0398] Here, the priority can be configured to correspond to each layer value. In addition, as an inter-layer CSI report prioritization method, the entire AI-CSI bit string (or bit sequence) can be configured based on subsequences corresponding to each layer, and the priority can be determined in order from the layer with the lowest index to the layers with the highest index. Alternatively, the reverse mapping may also be possible, and is not limited to a specific form.

[0399] Here, transmission of some sequences in CSI Part 2 may be omitted (or dropped, punctured) based on the above-described priorities. CSI Part 1 and CSI Part 2 may each independently perform AI encoding as in FIGS. 21 and 22 described above. Here, CSI Part 1 (or Part 1) may provide layer common information and size information (side information) for CSI Part 2 as in Table 23 and Table 24 below. For example, CSI information in CSI Part 1 may be AI CSI information having a higher priority than CSI information in CSI Part 2. Accordingly, AI CSI information in CSI Part 1 may be subsequences for which omission (or dropping, puncturing) is not performed, but may not be limited thereto. On the other hand, information corresponding to CSI Part 2 may not transmit subsequences corresponding to lower priorities, similar to the above-described methods. If the uplink resources allocated to the terminal for AI CSI reporting are insufficient, the terminal may drop and not transmit some of the subsequences with lower priorities among the subsequences associated with the CSI information corresponding to CSI Part 2. That is, AI encoding for CSI Part 1 and CSI Part 2 may be performed independently, and each AI encoding may be performed within each part. For example, it may be possible to apply a common AI encoding within each part, and it is not limited to a specific form.

[0400] [Table 24]

[0401]

[0402]

[0403] As another example, referring to Table 24 described above, the priority between subsequences can be determined by grouping the subsequence corresponding to the bit string with the largest Q, K, or S value as the subsequence with the highest priority (e.g., Group 1) based on the number of quantization bits per node (Q) of the CSI encoder output. In the same manner, grouping of subsequences based on Q, K, or S values ​​can be performed, as shown in Table 24.

[0404]

[0405] As described above, the payload structure of UCI including CSI may be applied layer-by-layer or may be a structure commonly applied to all layers. If the uplink resources (PUSCH) allocated to transmit UCI to the terminal are insufficient, the terminal may omit (or drop, puncture, rate match) transmission of some CSI or subsequences (or subsequence groups). For example, the priority for omitting transmission of some CSI or subsequences (or subsequence groups) may be determined by considering the information in Table 25 below, but may not be limited thereto.

[0406] [Table 25]

[0407]

[0408]

[0409] Figure 23 is a flowchart illustrating terminal operations applicable to the present disclosure. Referring to Figure 23, the terminal may be allocated uplink resources (UL grants) for AI CSI reporting from a base station. The terminal may compare the amount (or size) of the uplink resources with the AI ​​payload size of the AI ​​CSI content (S2310). For example, the AI ​​CSI content may further include AI-related CSI content and existing CSI content, but the present disclosure may not be limited to this embodiment.

[0410] Referring to FIG. 23, the terminal can determine whether sufficient uplink resources are allocated for AI CSI reporting (S2320). That is, the terminal can compare the amount of uplink resources with the AI ​​CSI payload size. Here, if the terminal is allocated sufficient uplink resources, the terminal can perform AI CSI reporting by including a bit string for the AI ​​encoder output in the AI ​​CSI payload (S2330).

[0411] On the other hand, if the terminal is not allocated sufficient uplink resources, the terminal can adjust the AI ​​CSI payload size and the AI ​​encoder output size. (S2340) Here, the AI ​​encoder output size can be adjusted through the quantization method applied based on the AI ​​CSI model performance, reliability, channel environment, use case, and scenario, and can be as described above. For example, the terminal can adaptively determine whether the quantization-related method and parameters can be utilized in relation to the AI ​​encoder. If not, the terminal can adjust the quantization method and related parameters in the AI ​​encoder according to the proposed method to adjust the AI ​​encoder size. Here, even if the AI ​​encoder size described above is adjusted, the AI ​​encoder size can be adjusted again within a limited range for the sake of the complexity of the implementation between the terminal and the base station and for an efficient quantization alignment procedure. That is, the terminal determines whether sufficient uplink resources (or threshold values) are allocated for transmitting the adjusted AI encoder output (S2350), and if sufficient uplink resources exist, the terminal can perform AI CSI reporting including the AI ​​encoder output (S2360). On the other hand, if sufficient uplink resources do not exist, the terminal can omit (or drop, puncture) some subsequences for the AI ​​encoder output (S2370).

[0412]

[0413] As another example, a terminal can simultaneously transmit both AI / ML-based CSI content (AI CSI content) and legacy CSI content to a base station. That is, the terminal can transmit both AI CSI content and legacy CSI content within the uplink resources allocated for UCI transmission. Here, a single CSI report can be either a legacy CSI report or an AI CSI report. For example, legacy CSI content can be transmitted and received for the purpose of AI CSI performance monitoring. In other words, it can serve as a comparative indicator of how reliably AI-CSI reports are reported and used. In the aforementioned case, AI CSI content and legacy CSI content can be reported from the terminal to the base station based on a single CSI report or multiple CSI reports at a specific time. For example, AI CSI content and legacy CSI content can be multiplexed within a single CSI report. Here, a decision may need to be made regarding how to handle legacy CSI content in relation to the CSI payload adjustment method described above.

[0414] As an example, FIG. 24 is a diagram illustrating a method for multiplexing and transmitting AI CSI content applicable to the present disclosure and existing CSI content.

[0415] Referring to Figure 24(a), existing CSI content can always take precedence over AI CSI content. That is, the terminal can prioritize existing CSI content to the MSB on the allocated uplink resources, and then allocate AI CSI to the remaining uplink resources. Here, some AI CSI content may be omitted depending on the amount (or size) of uplink resources.

[0416] As another example, referring to FIG. 24(b), the terminal may preferentially allocate CSI Part 1 to the MSB on the allocated uplink resources, and then allocate CSI Part 2 to the remaining uplink resources. Here, existing CSI content may be allocated with priority over AI CSI content in each of CSI Part 1 and CSI Part 2. For example, some of the CSI content allocated to the LSB side may be omitted depending on the amount (or size) of the uplink resources.

[0417] As another example, referring to FIG. 24(c), the terminal may first allocate CSI Part 1 to the MSB on the allocated uplink resources, and then allocate CSI Part 2 to the remaining uplink resources. Here, within CSI Part 1, existing CSI content may be allocated first, and AI CSI content may be allocated later. On the other hand, within CSI Part 2, existing CSI content and AI CSI content may be allocated alternately. For example, depending on the amount (or size) of uplink resources, some of the CSI content allocated to the LSB side may be omitted.

[0418] As another example, referring to FIG. 24(d), the terminal may first allocate CSI Part 1 to the MSB on the allocated uplink resources, and then allocate CSI Part 2 to the remaining uplink resources. Here, existing CSI content and AI CSI content may be allocated alternately within CSI Part 1 and CSI Part 2. For example, depending on the amount (or size) of uplink resources, some of the CSI content allocated to the LSB side may be omitted.

[0419] FIG. 25 is a diagram illustrating a method for performing AI CSI reporting applicable to the present disclosure. Referring to FIG. 25, a terminal may be allocated uplink resources for AI CSI reporting (S2510). That is, the terminal may obtain an uplink grant from a base station for UCI transmission and perform AI CSI reporting through the obtained uplink resources. Here, the terminal may determine whether to adjust the AI ​​CSI payload size based on the uplink resources (S2520). For example, if the uplink resources for AI CSI reporting are sufficient, the terminal may perform AI CSI reporting through the uplink resources. On the other hand, if the uplink resources for AI CSI reporting are insufficient, the terminal may adjust the AI ​​CSI payload size. That is, if the amount of uplink resources is less than the amount of AI CSI payload, the terminal may adjust the AI ​​CSI payload size. When the terminal adjusts the AI ​​CSI payload size, the terminal can perform AI CSI reporting through uplink resources based on the adjusted AI CSI payload size. (S2530) Here, the AI ​​CSI payload size can be adjusted by adjusting the AI ​​encoder output size in a quantization method determined based on at least one of AI CSI model performance, reliability, and channel environment. For example, the AI ​​encoder can generate an AI encoder output by performing an AI encoding and quantization procedure using a quantization method determined based on AI CSI inputs. The AI ​​encoder output is a bit string derived through the AI ​​encoder output size and the number of quantization bits per output node, and the AI ​​CSI payload size can be adjusted by adjusting at least one of the AI ​​encoder output size, the number of quantization bits per output node, and the bit string, and can be as shown in FIGS. 13 to 17 described above.

[0420] Additionally, as an example, the AI ​​encoder output described above can report additional output values ​​together through differential reporting operations based on the highest or most reliable output value based on the quantization step and range.

[0421] In addition, if the uplink resources are sufficient to perform AI CSI reporting based on the adjusted AI CSI payload size, the terminal may perform AI CSI reporting through the uplink resources. On the other hand, if the uplink resources are not sufficient to perform AI CSI reporting based on the adjusted AI CSI payload size, the terminal may omit (or drop, puncture) transmission of some subsequences among at least one subsequence of the AI ​​encoder output. Here, the AI ​​encoder output may include at least one subsequence based on the subsequence corresponding to each of the at least one AI encoder. The subsequences corresponding to each of the at least one AI encoders may be allocated to the AI ​​CSI payload in an order of subsequences having a higher priority based on AI encoder priorities. For example, each of the AI ​​encoders may correspond to at least one of a layer, a rank, a subband, and a port, and the AI ​​encoder priorities may be determined based on at least one of the layer, the rank, the subband, and the port corresponding to each of the AI ​​encoders. As another example, each AI encoder corresponds to AI CSI content, and the AI ​​encoder priority can be determined based on the AI ​​CSI content corresponding to each AI encoder, as shown in FIGS. 18 to 22.

[0422]

[0423] Figure 26 is a drawing showing a base station device and a terminal device to which the present disclosure can be applied.

[0424] The base station device (2600) may include a processor (2620), an antenna unit (2612), a transceiver (2626), and a memory (2616).

[0425] The processor (2620) performs baseband-related signal processing and may include a higher layer processing unit (2630) and a physical layer processing unit (2640). The higher layer processing unit (2630) may process operations of a MAC (Medium Access Control) layer, an RRC (Radio Resource Control) layer, or higher layers. The physical layer processing unit (2640) may process operations of a physical (PHY) layer (e.g., uplink reception signal processing, downlink transmission signal processing). In addition to performing baseband-related signal processing, the processor (2620) may also control the overall operation of the base station device (2600).

[0426] The antenna unit (2612) may include one or more physical antennas, and when it includes multiple antennas, it may support MIMO (Multiple Input Multiple Output) transmission and reception. In addition, it may support beamforming.

[0427] The memory (2616) can store information processed by the processor (2620), software related to the operation of the base station device (2600), an operating system, applications, etc., and may also include components such as a buffer.

[0428] The processor (2620) of the base station (2600) may be configured to implement the operations of the base station in the embodiments described in the present invention.

[0429] The terminal device (2650) may include a processor (2670), an antenna unit (2662), a transceiver (2664), and a memory (2666). For example, the terminal device (2650) in the present invention may communicate with a base station device (2600). As another example, the terminal device (2650) in the present invention may perform sidelink communication with another terminal device. That is, the terminal device (2650) in the present invention refers to a device that can communicate with at least one of the base station device (2600) and another terminal device, and is not limited to communication with a specific device.

[0430] The processor (2670) performs baseband-related signal processing and may include a higher layer processing unit (2680) and a physical layer processing unit (2690). The higher layer processing unit (2680) may process operations of the MAC layer, the RRC layer, or higher layers. The physical layer processing unit (2690) may process operations of the PHY layer (e.g., downlink reception signal processing, uplink transmission signal processing). In addition to performing baseband-related signal processing, the processor (2670) may also control the overall operation of the terminal device (2650).

[0431] The antenna unit (2662) may include one or more physical antennas, and when it includes multiple antennas, it may support MIMO transmission and reception. In addition, it may support beamforming.

[0432] The memory (2666) can store information processed by the processor (2670), software related to the operation of the terminal device (2650), an operating system, applications, etc., and may also include components such as a buffer.

[0433] A terminal device (2650) according to an example of the present invention may be associated with a vehicle. For example, the terminal device (2650) may be integrated into, positioned in, or positioned on the vehicle. Furthermore, the terminal device (2650) according to the present invention may be the vehicle itself. Furthermore, the terminal device (2650) according to the present invention may be at least one of a wearable terminal, an AV / VR, an IoT terminal, a robot terminal, and a public safety terminal. The terminal device (2650) to which the present invention is applicable may include any type of communication device that supports interactive services utilizing sidelink for services such as Internet access, service execution, navigation, real-time information, autonomous driving, and safety and risk diagnosis. Furthermore, any type of communication device capable of sidelink operation or an AR / VR device or a sensor that performs relay operation may be included.

[0434] Here, the vehicles to which the present invention is applied may include autonomous vehicles, semi-autonomous vehicles, non-autonomous vehicles, etc. Meanwhile, while the terminal device (2650) according to one example of the present invention is described as being associated with a vehicle, one or more of the UEs may not be associated with a vehicle. This is merely an example, and should not be construed as limiting the application of the present invention to the described example.

[0435] In addition, the terminal device (2650) according to an example of the present invention can be allocated uplink resources for AI CSI reporting from the base station device (2600). That is, the terminal device (2650) can obtain an uplink grant from the base station device (2600) for UCI transmission and perform AI CSI reporting through the obtained uplink resources. Here, the terminal device (2640) can determine whether to adjust the AI ​​CSI payload size based on the uplink resources. For example, if the uplink resources for AI CSI reporting are sufficient, the terminal device (2650) can perform AI CSI reporting through the uplink resources. On the other hand, if the uplink resources for AI CSI reporting are not sufficient, the terminal device (2650) can adjust the AI ​​CSI payload size. That is, if the amount of uplink resources is less than the amount of AI CSI payload, the terminal device (2650) can adjust the AI ​​CSI payload size. When the terminal device (2650) adjusts the AI ​​CSI payload size, the terminal device (2650) can perform AI CSI reporting through uplink resources based on the adjusted AI CSI payload size. Here, the AI ​​CSI payload size can be adjusted by adjusting the AI ​​encoder output size in a quantization method determined based on at least one of AI CSI model performance, reliability, and channel environment. For example, the AI ​​encoder can generate an AI encoder output by performing an AI encoding and quantization procedure using a quantization method determined based on AI CSI inputs. The AI ​​encoder output is a bit string derived through the AI ​​encoder output size and the number of quantization bits per output node, and the AI ​​CSI payload size can be adjusted by adjusting at least one of the AI ​​encoder output size, the number of quantization bits per output node, and the bit string, and can be as shown in FIGS. 13 to 17 described above.

[0436] Additionally, as an example, the AI ​​encoder output described above can report additional output values ​​together through differential reporting operations based on the highest or most reliable output value based on the quantization step and range.

[0437] In addition, if the uplink resources are sufficient to perform AI CSI reporting based on the adjusted AI CSI payload size, the terminal device (2650) may perform AI CSI reporting through the uplink resources. On the other hand, if the uplink resources are not sufficient to perform AI CSI reporting based on the adjusted AI CSI payload size, the terminal device (2650) may omit (or drop, puncture) transmission of some subsequences among at least one subsequence of the AI ​​encoder output. Here, the AI ​​encoder output may include at least one subsequence based on the subsequence corresponding to each of the at least one AI encoder. The subsequences corresponding to each of the at least one AI encoder may be allocated to the AI ​​CSI payload in an order of subsequences having a higher priority based on AI encoder priorities. For example, each AI encoder may correspond to at least one of a layer, a rank, a sub-band, and a port, and the AI ​​encoder priority may be determined based on at least one of the layer, the rank, the sub-band, and the port corresponding to each AI encoder. As another example, each AI encoder may correspond to AI CSI content, and the AI ​​encoder priority may be determined based on the AI ​​CSI content corresponding to each AI encoder, which may be as illustrated in FIGS. 18 to 22.

[0438] Additionally, various embodiments of the present disclosure may be implemented by hardware, firmware, software, or a combination thereof. In the case of hardware implementation, the embodiments may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), general processors, controllers, microcontrollers, microprocessors, etc.

[0439] The scope of the present disclosure includes software or machine-executable instructions (e.g., operating systems, applications, firmware, programs, etc.) that cause operations according to the methods of various embodiments to be executed on a device or a computer, and a non-transitory computer-readable medium having such software or instructions stored thereon and executable on the device or computer.

[0440] The various embodiments of the present disclosure are not intended to list all possible combinations but rather to illustrate representative aspects of the present disclosure, and the matters described in the various embodiments may be applied independently or in combination of two or more.

[0441]

[0442] The above may also apply to other systems.

Claims

1. In the method, A step in which a terminal is allocated uplink resources for reporting artificial intelligence (AI) channel state information (CSI); A step of determining whether to adjust the AI ​​CSI payload size based on the above uplink resources; and A method comprising, when the AI ​​CSI payload size is adjusted, performing the AI ​​CSI reporting through the uplink resource based on the adjusted AI CSI payload size.

2. In paragraph 1, If the uplink resources for the AI ​​CSI report are sufficient, the terminal performs the AI ​​CSI report through the uplink resources, A method for adjusting the AI ​​CSI payload size when the uplink resources for the AI ​​CSI report are insufficient.

3. In paragraph 2, A method in which the AI ​​CSI payload size is adjusted by adjusting the AI ​​encoder output size in a quantization method determined based on at least one of AI CSI model performance, reliability, and channel environment.

4. In paragraph 3, The AI ​​encoder generates an AI encoder output by performing an AI encoding and quantization procedure using the quantization method determined based on the AI ​​CSI inputs. The above AI encoder output is a bit string derived from the AI ​​encoder output size and the number of quantization bits per output node, A method for adjusting the AI ​​CSI payload size by adjusting at least one of the AI ​​encoder output size, the number of quantization bits per output node, and the bit string.

5. In paragraph 1, If the uplink resource is sufficient to perform the AI ​​CSI report based on the adjusted AI CSI payload size, the terminal performs the AI ​​CSI report through the uplink resource, A method in which, if the uplink resources are not sufficient to perform the AI ​​CSI reporting based on the adjusted AI CSI payload size, the terminal skips transmission of some subsequences among at least one subsequence of the AI ​​encoder output.

6. In paragraph 5, The above AI encoder output comprises at least one subsequence based on a subsequence corresponding to each of at least one AI encoder, A method wherein the subsequences corresponding to each of the at least one AI encoder are allocated to the AI ​​CSI payload in a subsequence order with a higher priority based on the AI ​​encoder priority.

7. In paragraph 6, Each AI encoder corresponds to at least one of a layer, a rank, a sub-band, and a port, A method wherein the AI ​​encoder priority is determined based on at least one of the layer, rank, sub-band, and port corresponding to each of the AI ​​encoders.

8. In paragraph 6, Each AI encoder corresponds to AI CSI content, A method wherein the AI ​​encoder priority is determined based on the AI ​​CSI content corresponding to each AI encoder.

9. For wireless devices, at least one processor; and A memory storing instructions that cause the wireless user device to perform a specific operation by the at least one processor, The above specific actions are: Uplink resources are allocated for AI (artificial intelligence) channel state information (CSI) reporting, Decide whether to adjust the AI ​​CSI payload size based on the above uplink resources, and A device that performs AI CSI reporting through the uplink resource based on the adjusted AI CSI payload size when the AI ​​CSI payload size is adjusted.

Images