Beam measurement method, user equipment and storage medium

Abstract

Provided in an embodiment of the present invention are a beam measurement method, user equipment and a storage medium, the beam measurement method comprising: a UE performing beam scanning to generate one or more measurement matrixes; compressing the one or more measurement matrices by performing matrix training to obtain one or more perceptual matrices based on the one or more measurement matrices, one perceptual matrix being a network perceptual matrix or a UE perceptual matrix; and performing weighted beam measurement using the one or more sensing matrices to generate one or more weighted beam measurement matrices. According to the invention, beam measurement can be carried out better.

Description

Technical Field

[0001] The present invention relates to wireless communications, and more particularly to compressed sensing based on artificial intelligence (AI) for beam prediction. Background Art

[0002] In traditional 3GPP 5G new radio (NR) networks, sub-6GHz and millimeter wave (mmWave) frequency bands are widely used. The former is mainly used for basic network coverage, while the latter has been widely used in ultra-low latency and ultra-high bandwidth communication scenarios due to its rich spectrum resources and short wavelength. However, millimeter wave signals are more susceptible to attenuation and absorption by obstacles such as buildings, trees and even rain. They also have a shorter propagation range and cannot penetrate solid objects like traditional wireless signals. With the rapid development of mobile applications, network efficiency needs to be improved. In order to overcome the large path loss in millimeter wave communications, beamforming technology is introduced to enhance signal strength and propagation capabilities.

[0003] In beam management, the optimal beam is selected through beam sweeping and beam measurement, which can be collectively referred to as beam searching / training. Traditional beam search methods are exhaustive and introduce a large amount of beam training overhead. It is time-inefficient to measure the beam quality information of all candidate beams in the beamforming codebook one by one. Another hierarchical beam search method can greatly save training overhead by first searching in a wide beam and then searching for a narrow beam within the angular range of the selected wide beam. The beam is selected based on an indicator that partially reflects the beam quality. In addition, in order to further improve the system efficiency, accurate and efficient beam prediction is required.

[0004] Improvements are needed in compressed sensing measurements and predictions for wireless networks. Summary of the invention

[0005] An embodiment of the present invention provides a method for performing beam measurement, comprising: performing beam scanning to generate one or more measurement matrices; compressing the one or more measurement matrices by performing matrix training to obtain one or more perception matrices based on the one or more measurement matrices, wherein a perception matrix is ​​a network perception matrix or a UE perception matrix; and performing weighted beam measurement using the one or more perception matrices to generate one or more weighted beam measurement matrices.

[0006] An embodiment of the present invention provides a user equipment, comprising: a radio frequency transceiver for sending and receiving radio signals; a measurement module for performing beam scanning to generate one or more measurement matrices; a compression module for compressing the one or more measurement matrices by performing matrix training to obtain one or more perception matrices based on the one or more measurement matrices, wherein a perception matrix is ​​a network perception matrix or a user equipment perception matrix; and a weighted measurement module for performing weighted beam measurement using the one or more perception matrices to generate one or more weighted beam measurement matrices.

[0007] An embodiment of the present invention provides a storage medium storing a program, which, when executed, enables a device to perform the steps of the method for performing beam measurement in the present application.

[0008] By utilizing the present invention, better beam measurements can be performed. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The accompanying drawings illustrate embodiments of the present invention, wherein like numerals indicate like components.

[0010] Figure 1 It is a system diagram of a MIMO wireless network using AI-based compressed sensing for beam prediction according to an embodiment of the present invention.

[0011] Figure 2 is an exemplary schematic diagram of predicting an optimal beam through a sensing matrix according to an embodiment of the present invention.

[0012] Figure 3 is an exemplary overall flow chart of the training phase according to an embodiment of the present invention.

[0013] Figure 4 is an exemplary overall flow chart of the prediction stage according to an embodiment of the present invention.

[0014] Figure 5 is an exemplary overall flow chart of performing optimal beam prediction in the spatial domain by reconstructing a compressed RSRP image according to an embodiment of the present invention.

[0015] Figure 6 is an exemplary overall flow chart of performing optimal beam prediction in the spatial and temporal domains by predicting and reconstructing compressed RSRP images according to an embodiment of the present invention.

[0016] Figure 7 is an exemplary flow chart of performing beam measurement according to an embodiment of the present invention. DETAILED DESCRIPTION

[0017] Reference will now be made in detail to some embodiments of the present invention, examples of which are illustrated in the accompanying drawings.

[0018] Several aspects of telecommunication systems will now be presented with reference to various devices and methods. These devices and methods will be described in the detailed description below and illustrated in the accompanying drawings by various blocks, components, circuits, processes, algorithms, etc. (collectively referred to as "elements"). These elements can be implemented using electronic hardware, computer software, or any combination thereof. Whether these elements are implemented as hardware or software depends on the specific application and the design constraints imposed on the entire system. The present invention provides methods, devices, processing systems, and computer-readable media for NR (or 5G technology), 6G, or other radio access technologies. NR can support various wireless communication services. These services may have different quality of service (QoS) requirements, such as delay and reliability requirements.

[0019] Figure 1 1 is a system diagram of a multiple input multiple output (MIMO) wireless network using AI-based compressed sensing for beam prediction according to an embodiment of the present invention. The wireless communication network 100 includes one or more fixed infrastructure units to form a network distributed over a geographical area. These basic units may also be referred to as access points, access terminals, base stations, node Bs, evolved node Bs (eNBs), next generation node Bs (gNBs), or other terms used in the art. For example, a base station serves multiple mobile stations within a service area (e.g., a cell or a cell sector). In some systems, one or more base stations are coupled to a controller to form an access network, wherein the access network is coupled to one or more core networks. gNB 106, gNB 107, and gNB 108 are base stations in a wireless network, and their service areas may or may not overlap. For example, user equipment (UE) 101 or mobile station 101 is located in a service area covered by gNB 106 and gNB 107; UE 102 or mobile station 102 is only in the service area of ​​gNB 107 and is connected to gNB 107. gNB 106 is connected to gNB 107 via Xn interface 121. gNB 106 is connected to gNB 108 via Xn interface 122. 5G network entity 109 is connected to gNB 106, 107 and 108 via NG connections 131, 132 and 133, respectively. In one embodiment, UE 101 is configured to be able to transmit data in an INACTIVE state without transitioning to a CONNECTED state.

[0020] In an example 110, artificial intelligence-based compressed sensing is performed on a MIMO wireless network 100 for beam prediction. In step 111, the UE performs beam scanning and measurement. The beam scanning generates one or more (complete) measurement matrices 116. The (complete) measurement matrix 116 is a measurement matrix of beam quality. Signal-to-noise ratio (SNR), reference signal received power (RSRP) and received signal strength (RSS) are typically used to indicate beam quality. In one embodiment, the complete measurement matrix is ​​an RSRP image. In step 112, the UE performs a training phase process, i.e., performs training / compression to generate one or more sensing matrices 117. The sensing matrix is ​​a network sensing matrix Φ1 or a UE sensing matrix Φ2. In step 113, the UE performs weighted beam measurement and generates one or more compressed measurement matrices 118. In one embodiment, the compressed measurement matrix is ​​a compressed RSRP image. In step 114, the UE performs matrix prediction to generate one or more prediction matrices and / or measurement matrix reconstruction. In one embodiment, the UE predicts a compressed measurement matrix (e.g., based on historical weighted beam measurement results). In one embodiment, the UE predicts a complete measurement matrix (e.g., based on historical weighted beam measurement results). In one embodiment, the UE predicts the optimal beam based on the predicted matrix. In one embodiment, the predicted compressed measurement matrix is ​​a compressed RSRP image. In another embodiment, the UE reconstructs the predicted compressed measurement matrix (e.g., a compressed RSRP image) into a complete measurement matrix, such as a complete RSRP image. In another embodiment, the UE reconstructs one or more compressed measurement matrices (e.g., a compressed RSRP image) into corresponding one or more complete measurement matrices, such as a complete RSRP image. The UE predicts the optimal beam based on one or more reconstructed measurement matrices (e.g., a complete RSRP image).

[0021] Figure 1 Further illustrated is a simplified block diagram of a base station and a mobile device / UE supporting AI-based compressed sensing for beam prediction. Figure 1 The simplified block diagram of a UE (e.g., UE 101) is included. The UE has an antenna 165 for sending and receiving radio signals. An RF transceiver circuit 163 coupled to the antenna receives an RF signal from the antenna 165, converts the RF signal to a baseband signal, and sends the baseband signal to a processor 162. The RF transceiver 163 also converts the baseband signal received from the processor 162 to an RF signal and sends it to the antenna 165. The processor 162 processes the received baseband signal and calls different functional modules to perform functional features in the UE 101. The memory 161 stores program instructions and data 164 to control the operation of the UE 101.

[0022] The UE also includes a set of control modules for performing functional tasks. These control modules can be implemented by circuits, software, firmware or a combination thereof. The state control module 191 controls the UE radio resource control (RRC) state according to the network's commands and UE conditions. RRC supports the following states: RRC idle (RRC_IDLE), RRC connected and RRC inactive. The DRB controller 192 establishes / adds, reconfigures / modifies and releases / removes DRBs based on different condition sets for DRB establishment, reconfiguration and release. The protocol stack controller 193 adds, modifies or removes the protocol stack of DRB, and the protocol stack includes SDAP, PDCP, RLC, MAC and PHY layers. In one embodiment, the SDAP layer supports data transmission, mapping between QoS flows and DRBs, marking QoS flow IDs, reflecting QoS flows to DRB mapping for UL SDAP data PDUs and other functions. The PDCP layer supports data transmission, maintenance of PDCP SN, header compression and decompression using the ROHC protocol, encryption and decryption, integrity protection and integrity verification, timer-based SDU discard, split bearer routing, duplication, reordering and in-order delivery, out-of-order delivery and duplicate discard. The RLC layer supports error correction, segmentation and reassembly, resegmentation, duplicate detection, reconstruction and other functions through ARQ. In one embodiment, a new process of RLC reconfiguration is performed, and the RLC entity can be reconfigured to be associated with one or two logical channels. The MAC layer supports the following functions: mapping between logical channels and transport channels, multiplexing / demultiplexing, HARQ, radio resource selection, etc.

[0023] The control module can perform tasks to implement beam prediction based on AI-based compressed sensing. The measurement module 181 performs beam scanning to generate one or more measurement matrices. The compression module 182 compresses one or more measurement matrices by performing matrix training to obtain one or more perception matrices based on one or more measurement matrices, wherein a perception matrix is ​​a network perception matrix or a user equipment perception matrix. The weighted measurement module 183 uses one or more perception matrices to perform weighted beam measurement to generate one or more weighted beam measurement matrices. In one embodiment, the control module also includes a prediction module 184, which performs prediction based on one or more weighted beam measurement matrices to generate one or more prediction matrices. In one embodiment, the control module also includes a reconstruction module 185, which performs measurement matrix reconstruction based on one or more weighted beam measurement matrices and one or more perception matrices.

[0024] Figure 1Also included is a simplified block diagram of a gNB (e.g., gNB 106). The gNB 106 has an antenna 156 that sends and receives radio signals. An RF transceiver circuit 153 coupled to the antenna receives an RF signal from the antenna 156, converts the RF signal to a baseband signal, and sends the baseband signal to the processor 152. The RF transceiver 153 also converts the baseband signal received from the processor 152 to an RF signal and sends it to the antenna 156. The processor 152 processes the received baseband signal and calls different functional modules to perform functional features in the gNB 106. The memory 151 stores program instructions and data 154 to control the operation of the gNB 106. The gNB 106 also includes a set of control modules 155 that perform functional tasks to communicate with mobile stations. These control modules may be implemented by circuits, software, firmware, or a combination of the above.

[0025] Figure 2 2 is an exemplary schematic diagram of predicting an optimal beam using a sensing matrix according to an embodiment of the present invention. UE 201 is connected to gNB 202. In step 211, UE 201 performs beam scanning. In step 212, UE 201 generates a beam having C T ×C R The complete measurement matrix 210 of dimension C T is the number of transmitters in the transmission beam, C R is the number of receivers receiving the beam. In one embodiment, the measurement matrix is ​​an RSRP measurement matrix. The RSRP measurements of all beam pairs in the codebook (i.e., the RSRP matrix) can be displayed as an image. The RSRP matrix X 210 is also called the RSRP image X 210, which is obtained by UE measurement and includes C T ×C R Each element of the matrix represents the RSRP measured for a beam pair between the network and the UE.

[0026] The sparsity of the complete measurement matrix (RSRP image X) 210, reflected in the shadows of the small grid, can be well exploited. Representing the complete RSRP image X 210, in step 221, the UE obtains one or more sensing matrices to find a suitable user equipment sensing matrix Φ2, and optionally, a network sensing matrix Φ1. Similarly, the network obtains one or more sensing matrices to find a suitable network sensing matrix Φ1, and optionally, a UE sensing matrix Φ2. According to compressed sensing, F = Φ1XΦ2. Where represents a compressed measurement matrix 220 or a compressed RSRP image 220, as well as are two measurement / sensing matrices. The above expression can be converted into a vector form f = Φx, where f = vec(F), x = vec(X), It can be called an equivalent sensing matrix. The multiplication result of the sensing matrices Φ1 and Φ2 on both sides of the RSRP matrix is ​​a compressed RSRP image 220 (i.e., a compressed RSRP matrix) F (F = Φ1XΦ2), which contains M×N elements. In step 222, the UE performs weighted / partial measurement. In one embodiment, in step 223, the UE predicts a future measurement matrix. In one embodiment, the predicted matrix is ​​a compressed measurement matrix.

[0027] Due to the sparsity of the original RSRP image, the compressed RSRP image can contain the maximum information from the original matrix under the appropriate perception matrix configuration. In one embodiment 231, the compressed measurement matrix / compressed RSRP image 220 is reconstructed into the original matrix 230 by certain methods. In one embodiment 235, the reconstruction algorithm uses a traditional greedy algorithm, such as orthogonal matching pursuit (OMP) or subspace pursuit (SP). In one embodiment 236, the reconstruction process is implemented by an AI model. In one embodiment, multiple long short-term memory (LSTM) layers are used for reconstruction of the AI ​​model, including refresh 261, LSTM1 262, LSTM2 263 and dense 264.

[0028] In one embodiment, in step 223, one or more compressed measurement matrices / compressed RSRP images are used to predict future matrices. Given the observed signal / compressed RSRP images and the perception matrix, a compressed reconstruction algorithm may be applied to solve the following optimization problem:

[0029] min‖x‖1

[0030] Make

[0031] The hyper-parameter ε is determined by the error of the reconstruction algorithm. In step 250, the UE predicts the optimal beam based on the predicted compressed measurement matrix / compressed RSRP image or the reconstructed complete measurement matrix / reconstructed complete RSRP image.

[0032] Figure 3is an exemplary overall flow chart of the training phase according to an embodiment of the present invention. A user equipment 301 is connected to a gNB 302 in a wireless network. In order to obtain detailed channel characteristics between the user equipment and the gNB, the user equipment performs a complete beam scan to generate an RSRP matrix X containing all Tx / Rx beam pairs and their corresponding RSRPs. In an embodiment 310, the user equipment performs multiple complete beam scans at different times to collect enough samples for model training and perception matrix parameter adjustment. In step 311, an exhaustive / complete beam scan is performed. In step 315, repeated exhaustive beam scans are performed. In step 320, the user equipment obtains / generates one or more (complete) measurement matrices 321. In an embodiment, the one or more (complete) measurement matrices are RSRP images. In an embodiment, in step 330, the user equipment generates and optimizes Φ1 for the network and Φ2 for the UE, and the compressed RSRP matrix (RSRP image) 331 can be derived as F=Φ1XΦ2, where Φ1 is the network perception matrix and Φ2 is the user equipment perception matrix. In step 340, the user equipment uses the trained AI model to reconstruct one or more compressed measurement matrices / compressed RSRP images into one or more reconstructed matrices / reconstructed RSRP images 341. In one embodiment, the principle of finding a suitable perception matrix and AI model is to reconstruct the matrix The error between (predicted from the compressed RSRP matrix) and the original RSRP matrix X is minimized. In one embodiment, the user equipment trains an AI model to reconstruct the compressed RSRP image F into the RSRP image X. In one embodiment, after training, the user equipment indicates the network-aware matrix Φ1 to the network for deployment.

[0033] Figure 4 is an exemplary overall flow chart of the prediction phase according to an embodiment of the present invention. User equipment 401 is connected to gNB 402. In an embodiment 410, the user equipment is trained in the training phase. In step 411, the user equipment 401 indicates the network perception matrix Φ1 to the gNB for deployment. In an embodiment 420, the user equipment 401 reports the measured RSRP matrix to the network, and the network performs training to optimize the perception matrices Φ1 and Φ2 for the network and UE 401. In an embodiment, one or more perception matrices are generated, such as the network perception matrix Φ1 and the user equipment perception matrix Φ2. The network perception matrix Φ1 can be generated by the user equipment or the wireless network. The user equipment perception matrix Φ2 can be generated by the user equipment or the wireless network. The user equipment 401 reports multiple historically measured RSRP matrices to the network to provide sufficient samples for model training and perception matrix parameter adjustment. The network trains an AI model for reconstructing the compressed RSRP image F into the RSRP image X. In an embodiment, in step 421, the network indicates the user equipment perception matrix Φ2 to the user equipment 401 for deployment after training.

[0034] In step 431, gNB 402 deploys a network sensing matrix Φ1. In step 432, user equipment 401 deploys a user equipment sensing matrix Φ2. In one embodiment, only the network sensing matrix Φ1 is deployed. In another embodiment, only the user equipment sensing matrix Φ2 is deployed. In another embodiment, as shown in steps 431 and 432, both the network sensing matrix Φ1 and the user equipment sensing matrix Φ2 are deployed. In one embodiment, the user equipment also deploys a well-trained artificial intelligence model. In the prediction stage, the user equipment 401 performs weighted beam measurement and obtains a compressed RSRP image F. In one embodiment, the weighted beam measurement is performed as a digital precoding for beamforming by deploying a sensing matrix. In one embodiment, in step 440, the network transmits a beamforming signal by controlling the power weights of different beams based on the sensing matrix Φ1. In one embodiment, the user equipment 401 receives multiple beams by beamforming through digital precoding to perform weighted beam measurement, and the digital precoding for beamforming is controlled by the user equipment sensing matrix Φ2. In one embodiment, at step 450, the user equipment 401 performs weighted beam measurement multiple times to collect enough samples to predict future compressed RSRP images.

[0035] In one embodiment 460, the network or user equipment generates and optimizes a network-aware matrix Φ1 or a user equipment-aware matrix Φ2. In one scenario, in step 461, a network-aware matrix Φ1 is generated. In step 462, a compressed RSRP matrix (RSRP image) can be derived as F′=Φ1X. In another scenario, in step 466, a user equipment-aware matrix Φ2 is generated. In step 467, a compressed matrix (RSRP image) can be derived as F′=XΦ2. In one embodiment, the compressed RSRP image can be unilaterally reconstructed into an RSRP image by the network or user equipment without signaling interaction between the user equipment and the network. In one embodiment, the network or user equipment trains an artificial intelligence model for reconstructing a compressed RSRP image F' into an RSRP image X. For example, in step 463, the network reconstructs F' into X based on the network-aware matrix Φ1. In another scenario, the user equipment 401 reconstructs F' into X based on the user equipment-aware matrix Φ2.

[0036] Figure 5 is an exemplary overall flow chart of performing optimal beam prediction in the spatial domain by reconstructing a compressed RSRP image according to an embodiment of the present invention. In an embodiment 520, the user equipment reconstructs the compressed RSRP image F 511 into the original RSRP image based on the perception matrix 512 configured on the user equipment and the network. 530. In one embodiment 521, the reconstruction algorithm is a traditional greedy algorithm, such as OMP, SP, etc. In one embodiment 522, the reconstruction process is implemented by an artificial intelligence model trained in the training phase. The compressed RSRP image F is reconstructed into the RSRP image The user equipment further performs optimal beam prediction based on the reconstructed RSRP image.

[0037] Figure 6 FIG. 6 is an exemplary overall flow chart of performing optimal beam prediction in the spatial and temporal domains by predicting and reconstructing a compressed RSRP image according to an embodiment of the present invention. In an embodiment 620, the user equipment performs optimal beam prediction in the spatial and temporal domains based on the compressed RSRP image 610 (e.g., F1, F2, ... F k ) Predict the future compressed RSRP image F N 621. In one embodiment 611, the prediction is implemented by an artificial intelligence model. In one embodiment 630, the UE reconstructs the predicted compressed RSRP image F based on the configured perception matrix 622 N The predicted RSRP image 641. In one embodiment, the UE directly predicts and reconstructs the future RSRP images based on the compressed RSRP images (e.g., F1, F2, F3, ...) measured in the past. In one embodiment, prediction and reconstruction are implemented by an AI model. In one embodiment 631, the reconstruction algorithm uses a traditional greedy algorithm, such as OMP, SP, etc. In one embodiment 632, the reconstruction process is implemented by an AI model trained in the training phase. The UE further performs future optimal beam prediction based on the reconstructed predicted RSRP image.

[0038] Figure 7 is an exemplary flow chart of performing beam measurement according to an embodiment of the present invention. In step 701, the UE performs beam scanning to generate one or more measurement matrices. In step 702, the UE compresses the one or more measurement matrices by performing matrix training to obtain one or more sensing matrices based on the one or more measurement matrices, wherein a sensing matrix is ​​a network sensing matrix or a UE sensing matrix. In step 703, the UE performs weighted beam measurement using the one or more sensing matrices to generate one or more weighted beam measurement matrices.

[0039] Although the present invention has been disclosed above with respect to the preferred embodiments, it is not intended to limit the present invention. Various changes, modifications and combinations may be made to the features in the various embodiments without departing from the protection scope of the present invention defined by the claims.

Claims

1. A method for performing beam measurement, comprising: The user equipment UE performs beam scanning to generate one or more measurement matrices; Compressing the one or more measurement matrices by performing matrix training to obtain one or more sensing matrices based on the one or more measurement matrices, wherein a sensing matrix is ​​a network sensing matrix or a UE sensing matrix; as well as Weighted beam measurements are performed using the one or more sensing matrices to generate one or more weighted beam measurement matrices.

2. The method for performing beam measurement as described in claim 1, wherein the one or more measurement matrices are reference signal received power (RSRP) matrices, and wherein the UE scans and measures all beams to generate the RSRP matrix.

3. The method for performing beam measurement as described in claim 1, wherein the matrix training is performed by the UE using an artificial intelligence (AI) model.

4. The method for performing beam measurement as claimed in claim 3, further comprising: The network-aware matrix is ​​indicated to a wireless network.

5. The method for performing beam measurement as claimed in claim 1, wherein the matrix training is performed by the wireless network using an AI model.

6. The method for performing beam measurement as claimed in claim 5, further comprising: The user equipment perception matrix is ​​received from the wireless network.

7. The method for performing beam measurement as claimed in claim 1, further comprising: A measurement matrix reconstruction is performed based on the one or more weighted beam measurement matrices and the one or more sensing matrices.

8. The method for performing beam measurement as claimed in claim 7, wherein the reconstruction is performed using a greedy algorithm or using a trained AI model.

9. The method for performing beam measurement as claimed in claim 1, further comprising: Predictions are performed based on the one or more weighted beam measurement matrices to generate one or more prediction matrices.

10. The method for performing beam measurement as claimed in claim 9, further comprising: An optimal beam is predicted based on the one or more prediction matrices.

11. The method for performing beam measurement of claim 10, wherein the one or more prediction matrices are reconstructed to predict the optimal beam.

12. A user equipment, comprising: Radio frequency transceivers, used to send and receive radio signals; A measurement module, configured to perform beam scanning to generate one or more measurement matrices; A compression module, configured to compress the one or more measurement matrices by performing matrix training to obtain one or more perception matrices based on the one or more measurement matrices, wherein a perception matrix is ​​a network perception matrix or a user equipment perception matrix; as well as A weighted measurement module is used to perform weighted beam measurement using the one or more sensing matrices to generate one or more weighted beam measurement matrices.

13. The user equipment of claim 12, wherein the one or more measurement matrices are reference signal received power (RSRP) matrices, and wherein the user equipment scans and measures all beams to generate the RSRP matrix.

14. The user device of claim 12, wherein matrix training is performed by the user device using an AI model.

15. The user equipment of claim 12, wherein the matrix training is performed by the wireless network using an AI model.

16. The user equipment of claim 15, further comprising a reconstruction module for performing measurement matrix reconstruction based on the one or more weighted beam measurement matrices and the one or more sensing matrices.

17. The user equipment of claim 16, wherein the reconstruction is performed using a greedy algorithm or using a trained AI model.

18. The user equipment according to claim 12, further comprising a prediction module, configured to perform prediction based on the one or more weighted beam measurement matrices to generate one or more prediction matrices.

19. The user equipment of claim 18, wherein the prediction module further predicts an optimal beam based on the one or more prediction matrices.

20. The user equipment of claim 19, wherein the one or more prediction matrices are reconstructed to predict the optimal beam.

21. A storage medium storing a program, wherein when the program is executed, the user equipment executes the steps of the method for performing beam measurement according to any one of claims 1 to 11.

Images