A fast and low-complexity beam scanning method

Abstract

The present invention discloses a fast and low-complexity beam scanning method, which relates to a hash beam generation method and a multi-arm beam generation and scanning method. Taking a RIS / IRS-assisted communication system as an example, first, a RIS / IRS-assisted communication scenario is modeled, then hash multi-arm beams and codebooks are generated. After pilot signal transmission, according to the signals received in each time slot, the access point separates the signals from different users based on existing modulation and demodulation technologies; demultiplexes the superimposed signals of the same user received by the access point, and makes a decision vote according to the demultiplexing results. The randomness and independence of the hash function itself in the present invention ensure the effective progress of demultiplexing and beam scanning, and the scanning overhead meets the complexity of the logarithmic level. Compared with the existing beam scanning technologies, the scanning method proposed by the present invention can quickly and accurately identify the best beam alignment direction, and shows excellent performance in terms of accuracy, speed and overhead.

Description

Technical Field

[0001] The present invention relates to the technical field of beam scanning, and particularly to a fast and low-complexity beam scanning method. Background Art

[0002] With the development of wireless communication, Reconfigurable Intelligent Surfaces (RIS) or Intelligent Reflecting Surfaces (IRS) have shown great potential in improving spectrum / energy efficiency, suppressing interference, and increasing the capacity of wireless communication systems. RIS / IRS consists of a large number of low-cost tunable elements that can be dynamically controlled to affect the radio propagation environment. In millimeter-wave communication at high frequency bands, the direct link between an access point (AP) and the users it serves is severely blocked by obstacles and suffers from path loss. RIS / IRS is usually used to establish a virtual AP-RIS / IRS-user line-of-sight (LoS) channel to improve communication quality. At the same time, the highly directional antennas in millimeter waves have high requirements for beam alignment between the transmitter and the receiver during communication. A large number of RIS / IRS reflecting elements generate pencil-shaped sharp beams, so a large number of beam directions are required to cover the space of interest, which makes the beam alignment of RIS / IRS a challenge.

[0003] In the existing technologies, the exhaustive search method selects the best beam by exhausting all possible beam directions, which requires scanning the entire beam space and will result in a long delay. The hierarchical search method performs spatial scanning according to the corresponding codebook at each stage. The scanning process is similar to the exhaustive search method, and then the best beam is found according to the received signal power, and a sub-codebook with higher resolution is subdivided in the next stage until the required spatial resolution is satisfied. The main disadvantage of hierarchical beam search is that using wide beams in the early stage will lead to a reduction in beamforming gain. Therefore, spatial scanning may not be able to identify the correct wide beam at low signal-to-noise ratios, and ultimately the best beam direction cannot be detected. Due to the severe path loss of the cascaded channel, this situation becomes worse in RIS / IRS-assisted millimeter-wave systems. In addition, since hierarchical search requires comparing and judging the scanning results of each stage to determine the scanning sub-codebook of the next stage, this will result in additional delay. The equal-spacing search method adopts a pre-determined multi-arm beam scanning order and performs a set operation after multiple rounds of scanning to obtain the best beam direction. The problem with this method is that it highly depends on the result of the first scan, which limits the accurate recognition rate of scanning to a certain extent.

[0004] In addition, all three of these technologies have a fatal flaw, that is, they are only effective in simple communication scenarios without RIS / IRS assistance or single RIS / IRS assistance. However, due to the cost-effectiveness of RIS / IRS and its important role in assisting communication, the scenario of multi-RIS / IRS-assisted communication is common and necessary. Different from the case of single RIS / IRS path reflection, the superposition of signals reflected by multiple RIS / IRS makes it very difficult to distinguish the optimal alignment directions of different RIS / IRS. None of the existing scanning technologies can perform simultaneous scanning of multiple RIS / IRS to determine the possible users in space and the alignment directions of each RIS / IRS with respect to the users from the received superimposed signals assisted by multiple RIS / IRS.

[0005] Based on the above problems, in order to improve the accuracy and efficiency of beam scanning technology and at the same time extend the applicable scenarios to a more common and practical multi-user communication model with multi-RIS / IRS assistance, a fast and low-complexity beam scanning method for RIS / IRS-assisted wireless communication systems is proposed. Summary of the Invention

[0006] The object of the present invention is to address the problems of single usage scenario, large beam recognition error, and high time delay of existing beam scanning technologies. Through a multi-arm beam composition method based on hash functions and a new decision voting mechanism according to received signal power, a fast beam scanning technology applicable to single RIS / IRS multi-user, multi-RIS / IRS single-user, and multi-RIS / IRS multi-user communication scenarios is proposed. The optimal beam direction obtained by this method has the highest accuracy except for exhaustive search and achieves a low scanning overhead of logarithmic level, showing excellent performance and generality.

[0007] The specific technical solution adopted by the present invention to solve its technical problems is: a fast and low-complexity beam scanning method, which can be applied to different communication scenarios. The method includes the following steps:

[0008] Step 1, perform RIS / IRS-assisted communication scenario modeling;

[0009] Step 2, generate hash multi-arm beams and codebooks: Discretize the beam space into N directions. For each RIS / IRS, L mutually independent randomly generated hash functions are used for hashing, where each hash function evenly hashes these N directions into B multi-arm beams, and construct the codebook corresponding to each RIS / IRS;

[0010] The codebook corresponding to each RIS / IRS contains BL rows, and each row consists of R segments of equal length, representing a multi-arm beam containing R = N / B sub-beams. Each segment corresponds to a sub-beam direction;

[0011] Step 3, Pilot signal transmission: In the uplink, each RIS / IRS uses a multi-arm beam to receive the pilot signals sent by each user and reflect them to the access point, and finally they are superimposed at the access point;

[0012] Step 4, The access point uses modulation and demodulation techniques to separate the signals from different users received in each time slot, and then demultiplexes the superimposed signals of the same user received; The demultiplexing includes:

[0013] In the first round of loop, for the L time slots corresponding to the L values with the largest soft decision signal-to-noise ratio, the time slots contain the signals reflected by the user through the strongest RIS / IRS reflection path; In the second round of loop, for the L time slots corresponding to the L values with the second largest soft decision signal-to-noise ratio, the time slots contain the signals reflected by the user through the second strongest RIS / IRS reflection path; And so on until the loop end condition is reached;

[0014] The loop condition is as follows: If the number of RIS / IRS is known, the number of loop rounds is equal to the number of RIS / IRS; If the number of RIS / IRS is unknown, the loop stops until the remaining time slots only contain noise and interference signals;

[0015] Step 5, Conduct decision voting. The access point votes for each sub-beam direction in the corresponding multi-arm beam of each RIS / IRS according to the time slots containing the user signals obtained by demultiplexing in Step 4, and selects the direction of the RIS / IRS with the highest number of votes as the best alignment direction for the user.

[0016] Further, the RIS / IRS-assisted communication scenario is that the RIS / IRS assists the user and the access point in communication. The RIS is composed of tunable reflection elements, and the access point communicates with the RIS / IRS through a control channel to adjust the phase shift of the reflection elements.

[0017] Further, the number L of the hash functions can be proven and verified by simulation, and the formula is as follows: L = O(logN).

[0018] Further, in Step 2, the codebook is a two-dimensional hash multi-arm beam codebook, including:

[0019] In the horizontal / vertical direction, the azimuth / pitch angle of the beam space is discretized into N directions, and the direction set is represented as [1, 2,..., N]; Use the hash function to calculate the hash values of the keywords 1, 2,..., N, and the storage location of the keywords, that is, the bucket, can be found through the hash values as indexes; Select L mutually independent hash functions from the hash function family, and use each hash function to evenly hash the keywords 1, 2,..., N into B buckets, and a total of BL bucket groups are obtained, and each bucket represents a multi-arm beam.

[0020] Further, in step 2, the codebook is a three-dimensional hash multi-arm beam codebook, including:

[0021] The azimuth angle and elevation angle in the beam space are discretized into N1 and N2 directions respectively, and there are N1N2 directions [φ, θ] composed of azimuth angles and elevation angles; L mutually independent hash functions are selected from the hash function family, and each hash function evenly hashes the keywords 1, 2,..., N1N2 into B buckets, resulting in a total of BL bucket groups, and each bucket represents a multi-arm beam.

[0022] Further, the demultiplexing and voting methods for each user are the same and are carried out simultaneously.

[0023] Further, in the single RIS / IRS multi-user communication scenario, since there is only a single RIS / IRS assisting communication, for each user, the signal received by the access point is only reflected by a single RIS / IRS; according to the power of the received signal, it is determined whether the received signal in each time slot contains useful signals from the user by means of hard decision or soft decision.

[0024] Further, in the single RIS / IRS multi-user communication scenario, the hard decision is: for the received signal in each time slot, if the signal power value is greater than the preset threshold, it is determined that the user signal is included, otherwise it is determined that only noise and interference signals are included.

[0025] Further, in the single RIS / IRS multi-user communication scenario, the soft decision is: among the received signals in a total of BL time slots, the L time slots corresponding to the L largest signal power values contain user signals, that is, the multi-arm beam of the RIS / IRS can scan to the user, and the remaining time slots only contain noise or interference signals.

[0026] Further, in the multi-RIS / IRS scenario, the voting process in step 5 is specifically as follows: according to the signal-to-noise ratio of the received signal, first find the largest L values from it, and the corresponding time slot serial numbers are defined as The same decision and voting operations are performed for each RIS / IRS, and the sub-beams in the multi-arm beam of the time slot are voted on, and the direction of the RIS / IRS with the highest number of votes is the best alignment direction of the user; then the above voting method is repeated for the time slots with the second largest L signal-to-noise ratios until the alignment directions of all RIS / IRSs are calculated.

[0027] The beneficial effects of the present invention are as follows:

[0028] The randomness and independence of the hash function ensure a satisfactory successful recognition rate. The scanning method of the multi-arm beam and the simultaneous scanning of multiple users and multiple RIS / IRSs significantly reduce the scanning complexity. The scanning complexity is independent of the number of users and the number of RIS / IRSs, and the decision voting process is carried out at the AP. Therefore, the scanning method of the present invention has low complexity, low power consumption, and high accuracy.

[0029] (1) The random performance of the hash function brings higher accuracy. Compared with fixed scanning beam methods such as equidistant scanning and hierarchical search scanning, the randomness of the hash function reduces the leakage interference of multi-round averaging, including the interference between multiple sub-beams within a single multi-arm beam and the interference between different multi-arm beams. The hash multi-arm beam scanning method can achieve the highest accuracy except for the exhaustive search scanning, and is significantly better than the equidistant multi-arm beam scanning and hierarchical search scanning.

[0030] (2) In the multi-RIS / IRS assisted scenario, since the signals of users are superimposed after being reflected by multiple RIS / IRSs, only the hash multi-arm beam scanning method can realize the simultaneous scanning of multiple RIS / IRSs, making the scanning complexity independent of the increase in the number of RIS / IRSs or the number of users. This is because the randomness and independence of the hash function itself guarantee the premise for the effective operation of the proposed demultiplexing algorithm. The signals after demultiplexing processing bring more known information to recover the directions of each RIS / IRS for the user in the subsequent voting. While equidistant scanning, hierarchical search scanning, and exhaustive search scanning can only obtain the same high accuracy by scanning each RIS / IRS separately, so the complexity is proportional to the number of RIS / IRSs and the training cost is very high.

[0031] (3) A decision-making method using soft decision is proposed, which can achieve higher accuracy than hard decision in the absence of the direct path. Because soft decision is based on the relative value of power / SNR rather than the absolute value, it shows better robustness in the case of a large number of sub-beams, increased background noise, or interference from other RIS / IRSs. Description of the Drawings

[0032] Figure 1 is the communication scenario diagram of the research problem of the present invention;

[0033] Figure 2 is the example diagram of uniform hashing of the present invention;

[0034] Figure 3 is the schematic diagram of the two-dimensional multi-arm beam generated by the present invention;

[0035] Figure 4 is the schematic diagram of the three-dimensional multi-arm beam generated by the present invention;

[0036] Figure 5 It is a schematic diagram of the voting mechanism in the single RIS / IRS multi-user scenario of the present invention;

[0037] Figure 6 It is a schematic diagram of the voting mechanism in the multi-RIS / IRS single-user scenario of the present invention;

[0038] Figure 7 It is a schematic diagram of the voting mechanism for three-dimensional beam scanning in the multi-RIS / IRS scenario of the present invention;

[0039] Figure 8 It is a simulation diagram of the accurate recognition rate in the single RIS / IRS multi-user scenario of the present invention;

[0040] Figure 9 It is a simulation diagram of the scanning complexity in the single RIS / IRS multi-user scenario of the present invention;

[0041] Figure 10 It is a simulation diagram of the accurate recognition rate in the multi-RIS / IRS scenario of the present invention;

[0042] Figure 11 It is a simulation diagram of the scanning complexity in the multi-RIS / IRS scenario of the present invention. Detailed implementation manners

[0043] The following further elaborates on the detailed implementation manners of the present invention in conjunction with the accompanying drawings.

[0044] Many specific details are set forth in the following description to facilitate a thorough understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0045] The present invention provides a fast and low-complexity beam scanning method;

[0046] 1. The specific implementation steps of the fast and low-complexity beam scanning method of the present invention in the single RIS / IRS multi-user scenario are as follows:

[0047] (1) Modeling of the single RIS / IRS multi-user communication scenario: As Figure 1 shown, the present invention studies the uplink of a multi-user communication system assisted by a single RIS / IRS, where I = 1 distributed RIS / IRS is deployed to assist from K single-antenna user groups to a single one with N AThe transmission of the access point (AP) of the antenna. The RIS / IRS consists of N0 reflection elements. To minimize the path loss between the RIS / IRS and the served users, in this embodiment, the RIS / IRS is deployed near the user cluster in the actual scenario. The direct link between the AP and the users is blocked by obstacles. The AP communicates with the RIS / IRS through the control channel to adjust the phase shift of the reflection elements. Since the positions of the access point and the RIS / IRS are fixed, in this embodiment, the beamforming of the transmitter and receiver of the access point-RIS / IRS channel is fixed to focus on studying the beam scanning problem of the RIS / IRS-users.

[0048] In the millimeter-wave band, the wireless channel follows the geometric channel model. The steering vector can be expressed as

[0049]

[0050] where for N0 = N h ×N v antenna uniform planar array (UPA), N h is the number of horizontal array elements, N v is the number of vertical array elements; the horizontal array element spacing is d h , the vertical array element spacing is d v , assuming that the azimuth angle of the incident wave from user k arriving at the RIS / IRS is φ k , the elevation angle is θ k , λ represents the wavelength of the incident wave, and Since the positions of the AP and the RIS / IRS are fixed, while the positions of the users are mobile, therefore, relative to the RIS / IRS-users channel, the AP-RIS / IRS channel can be regarded as a quasi-static channel. The beam alignment between the AP and the RIS / IRS can be determined according to the previous channel state information. Therefore, the uniform linear array of the AP can be equivalent to a single antenna. The transmission signal of user k arriving at the AP through the reflection channel of the RIS / IRS can be written as

[0051]

[0052] where, is the signal sent by user k with transmission power P, is the additive white Gaussian noise (AWGN), h k represents the complex gain of the user k-RIS / IRS LoS channel, g represents the complex gain of the effective channel from the RIS / IRS to the AP, represents the reflection diagonal matrix of the RIS / IRS, where ψ n is the phase shift of the nth reflection unit of the RIS / IRS. The superscript H is the conjugate transpose operation, the superscript r represents reception, and the superscript t represents transmission.

[0053] (2) Hash beam and codebook generation: The single-beam search mode of the prior art scans the entire space, which will cause a large delay. The present invention adopts a multi-arm beam scanning method, which can greatly reduce the number of scans.

[0054] (2.1) Two-dimensional hash beam codebook: In the horizontal / vertical direction, the azimuth / pitch angle of the beam space is discretized into N directions, and the set is represented as [1, 2,..., N]. The hash values of the keywords 1, 2,..., N are calculated by using a hash function, and the storage location of the keyword, that is, the bucket, can be found through the hash value as an index. For example Figure 2 , it is shown that 3 hash functions evenly hash the directions 1 to 16 in the space into 4 buckets. Therefore, the hashing result of the first hash function is that the first multi-arm beam consists of directions 1, 4, 11, 15, the second multi-arm beam consists of directions 2, 6, 12, 14, the third multi-arm beam consists of directions 3, 8, 10, 16, and the fourth multi-arm beam consists of directions 5, 7, 9, 13. The present invention selects L mutually independent hash functions from the hash function family, and each hash function evenly hashes the keywords 1, 2,..., N into B buckets, and a total of BL groups of buckets can be obtained. Each bucket represents the sub-beam directions constituting a multi-arm beam.

[0055] Figure 3 As shown Figure 2 is the multi-arm beam image generated by the first-round hashing in, where the same line type represents a multi-arm beam. The corresponding multi-arm beam codebook C can be constructed from the DFT codebook: Each row of the multi-arm beam codebook C represents a multi-arm beam, denoted as C = [c1,..., c B T . Each multi-arm beam consists of R = N / B sub-beams, corresponding to the directions represented by the R keywords included in the bucket. The codebook of each sub-beam is intercepted from M adjacent elements in the row corresponding to the direction of the DFT codebook and can be expressed as where b(r) ∈ [1, 2,..., N] represents the keyword corresponding to the r-th sub-beam included in the b-th multi-arm beam, MN represents the number of antenna elements, represents the b(r)-th row of the DFT codebook .

[0056] ​(2.2) 3D Hash Beam Codebook: In practical situations, scanning needs to be performed in three-dimensional space directions. First, the azimuth and elevation angles in the beam space are discretized into N1 and N2 directions respectively. Therefore, there are N1·N2 directions [φ, θ] composed of azimuth and elevation angles. Then, L mutually independent hash functions are selected from the hash function family. Each hash function evenly distributes the keywords 1, 2,..., N1·N2 into B buckets. A total of BL bucket groups can be obtained, and each bucket represents the sub-beam directions of a multi-arm beam. The corresponding multi-arm beam codebook C is still constructed by intercepting from the DFT codebook. As Figure 4 shown, it shows a schematic diagram of a 3D hash beam in the 3D space with N1×N2 = 16×16 directions (formed by one of the hash functions), evenly distributed into B = 32 buckets, that is, each multi-arm beam contains 8 sub-beams. In the figure, the ranges of the φ-axis and θ-axis are from 0 to 180 degrees, representing the azimuth and elevation angles of the scanning beam in 3D space.

[0057] The advantage of this joint hash scanning scheme is that, compared with scanning the horizontal azimuth and vertical directions separately, it will not cause the superposition of scanning errors introduced in two dimensions, and can ensure higher accuracy under the condition of the same number of scanning times.

[0058] (3) Pilot Signal Transmission: In the uplink, each user continuously sends pilot signals to the AP during the beam scanning phase. During each training symbol duration, the RIS / IRS sequentially selects a row from its codebook, that is, uses a multi-arm beam to receive the pilot signals sent by each user and reflects them to the AP. We selected L = O(logN) mutually independent hash functions and performed L rounds of hash operations. Therefore, the total scanning time is Q = BL time slots.

[0059] (4) Decision and Voting Mechanism: According to the signals received in each time slot, the AP can separate the signals from different users based on existing modulation and demodulation technologies. The phase information of the received signals is not used here, and decisions and votes are made only based on the amplitude values / power values of the received signals.

[0060] Given the signal power values from each user and the multi-arm beamforming pattern of the RIS / IRS in each time slot, determine the time slots containing the user signals using hard decision or soft decision methods. For hard decision, a threshold T needs to be set in advance according to the current channel environment. If the power value of the received signal is greater than the threshold T, it is considered that the RIS / IRS can scan user k in the corresponding time slot, that is, the pilot signal of user k can be reflected to the AP in this time slot. Then, vote for each sub-beam direction contained in the bucket of this time slot. Otherwise, it is determined that the received power in this time slot is only noise and interference signals and no vote is given. For soft decision, since there is exactly one bucket among the B buckets in each round of hashing that contains the alignment direction of the RIS / IRS for user k, among the BL time slots, there are exactly L time slots that contain useful information from user k. Compare the time slots corresponding to the L largest received signal power values and vote for each sub-beam direction of the L buckets corresponding to these L time slots.

[0061] The same decision and voting operations are performed for each user. After the signal decision and voting for all time slots are completed, the direction with the highest number of votes is selected as the best alignment direction of the RIS / IRS relative to user k.

[0062] As Figure 5 shown in the figure, the two-round hashing multi-arm beams of the RIS / IRS are shown in the case of two users. In each round, different and independent hashing functions are used to uniformly hash the 16 directions in the example, and four multi-arm beams are generated in each round. In each training time slot, the RIS / IRS sequentially uses one of the multi-arm beams to receive and reflect the pilot signals from user 1 and user 2. The AP separates and judges the received signals among users and obtains that the multi-arm beam marked in light gray contains useful signals from user 1, and the multi-arm beam marked in dark gray contains useful signals from user 2. For user 1, vote for each sub-beam contained in the first bucket of the first round and the first bucket of the second round. Thus, direction 6 gets two votes, directions 1, 2, 8, 11, 12, 14 each get one vote, and the remaining directions get no votes. Therefore, direction 6 is the best alignment direction of the RIS / IRS for user 1; similarly, for user 2, vote for each sub-beam contained in the third bucket of the first round and the fourth bucket of the second round. Thus, direction 13 gets two votes, directions 2, 3, 5, 7, 8, 11, 16 each get one vote, and the remaining directions get no votes. Therefore, direction 13 is the best alignment direction of the RIS / IRS for user 2.

[0063] 2. The specific implementation steps of the fast low-complexity beam scanning method of the present invention in the multi-RIS / IRS single-user scenario are as follows:

[0064] (1) Modeling of the multi-RIS / IRS single-user communication scenario: As Figure 1As shown, the present invention studies the uplink of a multi-RIS / IRS assisted single-user communication system, where I distributed RISs / IRSs are deployed to assist the transmission from K = 1 single-antenna user to an access point (AP) with N A antennas. Each RIS / IRS (denoted by ) consists of N i reflecting elements. To minimize the path loss between the RIS / IRS and the user, we assume that the RIS / IRS is deployed near the user in an actual scenario. The direct link between the AP and the user is blocked by obstacles. The AP communicates with each RIS / IRS through a control channel to adjust the phase shift of the reflecting elements. Since the positions of the access point and the RIS / IRS are fixed, for simplicity, we assume that the beamforming of the transmitter and receiver of the access point-RIS / IRS channel is fixed to focus on studying the beam scanning problem of the RIS / IRS-user.

[0065] In the millimeter-wave band, the wireless channel follows a geometric channel model. For a uniform planar array (UPA) with N i = N h ×N v antennas, N h is the number of horizontal array elements, and N v is the number of vertical array elements; the horizontal array element spacing is d h , and the vertical array element spacing is d v . Assuming that the azimuth angle of the incident wave from the user to the i-th RIS / IRS is φ i , and the elevation angle is θ i , the steering vector can be expressed as

[0066]

[0067] where λ represents the wavelength of the incident wave, and Since the positions of the AP and the RIS / IRS are fixed, while the position of the user is mobile, the AP-RIS / IRS channel can be regarded as a quasi-static channel with respect to the RIS / IRS-user channel. The beam alignment between the AP and the RIS / IRS can be determined according to the previous channel state information, so the uniform linear array of the AP can be equivalent to a single antenna. We assume that there is a LoS channel between the user and each RIS / IRS, and the transmitted signal of the user reaching the AP through the reflection channels of the RIS / IRS set can be written as

[0068]

[0069] where, is the signal transmitted by the user with transmission power P, is additive white Gaussian noise (AWGN), h i represents the complex gain of the user-RIS / IRS i LoS channel, g i represents the complex gain of the effective channel from the i-th RIS / IRS to the AP, represents the diagonal reflection matrix of the i-th RIS / IRS, where is the phase shift of the nth reflection unit of the ith RIS / IRS. The superscript H represents the conjugate transpose operation, the superscript r represents receiving, and the superscript t represents transmitting.

[0070] (2) Hash beam and codebook generation: The single beam search mode of the prior art scans the entire space, which will cause a large delay. The present invention adopts a multi-arm beam scanning method, which can greatly reduce the number of scans. In this scenario, it is necessary to generate a different multi-arm beam scanning codebook for each RIS / IRS:

[0071] (2.1) Two-dimensional hash beam codebook: In the horizontal / vertical direction, the azimuth / elevation angle of the beam space is discretized into N directions, and the set is represented as [1, 2, ..., N]. The hash function is used to calculate the hash value of the keyword 1, 2, ..., N, and the storage location of the keyword, i.e., the bucket, can be found by using the hash value as an index. Figure 2 , the three hash functions shown in the figure evenly hash the directions 1 to 16 in the space into four buckets, so the hash result of the first hash function is that the first multi-arm beam consists of directions 1, 4, 11, 15, the second multi-arm beam consists of directions 2, 6, 12, 14, the third multi-arm beam consists of directions 3, 8, 10, 16, and the fourth multi-arm beam consists of directions 5, 7, 9, 13. For each RIS / IRS, L independent hash functions are selected from the hash function family, and each hash function is used to evenly hash the keywords 1, 2, ..., N into B buckets, so that a total of BL buckets can be obtained, and each bucket represents the sub-beam direction composition of a multi-arm beam.

[0072] Figure 3 Shown is Figure 2 The multi-arm beam image generated by the first round of hashing in , where the same line type represents a multi-arm beam. The multi-arm beam codebook C corresponding to each RIS / IRS i It can be constructed from the DFT codebook: where the multi-arm beam codebook C i Each row of represents a multi-arm beam, denoted as C i =[c i，1 , ..., c i，B | T Each multi-arm beam consists of R = N / B sub-beams, corresponding to the directions represented by the R keywords contained in the bucket. The codebook for each sub-beam is derived from the DFT codebook obtained by intercepting M adjacent elements from the rows in the corresponding direction, and can be expressed as where b(r) ∈ [1, 2,..., N] represents the keyword corresponding to the r-th sub-beam included in the b-th multi-arm beam, MN represents the number of antenna elements, represents the DFT codebook of the b(r)-th row.

[0073] (2.2) 3D Hash Beam Codebook: In practical situations, scanning needs to be performed in three-dimensional space directions. First, the azimuth and elevation angles of the beam space are discretized into N1 and N2 directions respectively. Therefore, there are a total of N1·N2 directions [φ, θ] composed of azimuth and elevation angles. For each RIS / IRS, L mutually independent hash functions are selected from the hash function family. Each hash function evenly hashes the keywords 1, 2,..., N1·N2 into B buckets, and a total of BL bucket groups can be obtained. Each bucket represents the sub-beam direction composition of a multi-arm beam. The corresponding multi-arm beam codebooks C1, C2,..., C I are still intercepted and constructed from the DFT codebook as shown in Figure 4 which shows a schematic diagram of a 3D hash beam for a RIS / IRS hashed with one hash function in the 16×16 directions of 3D space, evenly hashed into B = 32 buckets, that is, each multi-arm beam contains 8 sub-beams. In the figure, the ranges of the φ-axis and θ-axis are from 0 to 180 degrees, representing the azimuth and elevation angles of the scanning beam in 3D space.

[0074] The advantage of this joint hashing scanning scheme is that, compared with scanning the horizontal azimuth and vertical directions separately, it will not cause the superposition of scanning errors introduced in two dimensions, and can ensure higher accuracy under the condition of the same number of scanning times.

[0075] (3) Pilot Signal Transmission: In the uplink, the user continuously sends pilot signals to the AP during the beam scanning phase. During each training symbol duration, each RIS / IRS sequentially selects a row from its respective codebook, that is, uses a multi-arm beam to receive the pilot signal sent by the user and reflects it, and finally superimposes it at the AP. We have selected L mutually independent hash functions for each RIS / IRS, and a total of L rounds of hashing operations have been performed. Since the RIS / IRS can achieve simultaneous scanning, the total scanning time is Q = BL time slots. Denote the signal received by the AP from the user in Q time slots as

[0076]

[0077] q = lB + b, where y(q) represents the signal received by the AP in the b-th time slot of the l-th round of hashing.

[0078] (4) Demultiplexing the superimposed signals: Since the AP receives the signals superimposed after being reflected by multiple RIS / IRS reflection paths from the user, in order to identify the alignment directions of different RIS / IRSs with respect to the user, the prerequisite condition that no two or more RIS / IRSs scan the user simultaneously needs to be satisfied. This prerequisite assumption can be met by setting the number of bins B: In each time slot, due to the randomness of the hash function, the probability that a RIS / IRS scans the user is Due to the independence of the hash functions from each other, the probability that two RIS / IRSs scan the same user As long as the probability value of Pr is small enough, it can be ensured that y(q) contains at most the reflected signal of one RIS / IRS. When B = 8, this requirement can be basically met, that is

[0079]

[0080] where γ i represents the direction from the i-th RIS / IRS to the user, D i (q) represents the set of directions included in the q-th multi-arm beam of the i-th RIS / IRS, s i (q) represents the actual signal reflected from the user through the i-th RIS / IRS, and n(q) represents the noise in the q-th time slot. In addition, due to the limited transmit power of the user and different attenuations of each RIS / IRS reflection path, we assume that the signal-to-noise ratios (SNRs) of the signals reflected from different RIS / IRSs are as follows: SNR1 > SNR2 > …… > SNR I , according to the SNR of the received signal, the signals reflected by different RIS / IRSs can be separated from the Q signals: The largest L SNRs correspond to the reflection path of RIS / IRS1, the second largest L SNRs correspond to RIS / IRS 2, and so on until the SNRs of the remaining time slots are close to 0, or after cycling I rounds. Here, the phase information or channel state information of the received signal is not used, only based on the SNR value of the received signal.

[0081] (5) Decision and voting mechanism: If any L multi-arm beams of the i-th RIS / IRS are voted, due to the randomness of the hash function hashing, only when voting on the multi-arm beam containing the user's signal can the highest number of votes η i and the best alignment direction γ i be obtained, and the probability that the number of votes in the remaining directions is higher It is almost zero, representing an impossible event. Therefore, only by voting based on the additional useful information obtained through demultiplexing can we obtain the best alignment direction of the RIS / IRS assisting the user's communication with the user. Specifically, according to the calculated signal-to-noise ratio of the received signal, we first find the largest L value from it, and the corresponding time slot sequence number is defined as The same decision and voting operations are performed for each RIS / IRS, voting on the sub-beams in the multi-arm beam of the time slot The direction with the highest number of votes in the sub-beams of the multi-arm beam is the best alignment direction of the corresponding RIS / IRS i with respect to the user. Then, the above voting method is repeated for the time slots with the second-largest L signal-to-noise ratios until the alignment directions of all RIS / IRSs are calculated.

[0082] As Figure 6 , there are two RIS / IRSs in space assisting the communication between the user and the AP at the same time. The channel gain of the reflection path of RIS / IRS 1 is greater than that of the reflection path of RIS / IRS 2. The spatial position of the user is at direction 26 of RIS / IRS 1 and at direction 5 of RIS / IRS 2. The table in the figure shows that 32 directions in space are evenly hashed into 8 multi-arm beams, with two rounds of hashing. The left table represents RIS / IRS 1, and the right table represents RIS / IRS 2. Each row of the table represents a multi-arm beam, which consists of 4 sub-beam directions. The known information after demultiplexing shows that the received signal SNR in the 2nd and 9th time slots is the largest, marked in light gray, and the received signal SNR in the 7th and 14th time slots is the second-largest, marked in dark gray. Theoretically, the signals in the 2nd and 9th time slots are reflected through the strongest RIS / IRS reflection path and only contain the signals reflected by RIS / IRS 1; the signals in the 7th and 14th time slots are reflected through the second-strongest RIS / IRS reflection path and only contain the signals reflected by RIS / IRS 2. Figure 6 As shown in (a) of Figure 6 , one vote is cast for each sub-beam in the multi-arm beam marked in light gray. It can be obtained that the number of votes in the direction 26 in the table of RIS / IRS 1 is the highest, and the number of votes in any direction in RIS / IRS 2 is less than the number of votes in the direction 26 in RIS / IRS 1. Thus, it can be obtained that the channel gain of the reflection path of RIS / IRS 1 is the strongest, and the user is at the 26 direction of RIS / IRS 1. Continuing to vote on the multi-arm beam marked in dark gray, as Figure 6 shown in (b) of

[0083] , the number of votes in the direction 5 in the table of RIS / IRS 2 is the highest. Thus, it can be obtained that the channel gain of the reflection path of RIS / IRS 2 is the second-strongest after RIS / IRS 1, and the user is at the 5 direction of RIS / IRS 2, and the voting result conforms to the theoretical expectation.3. The specific implementation steps of the fast low-complexity beam scanning method of the present invention in the multi-RIS / IRS multi-user scenario are as follows:

[0084] (1) Multi-RIS / IRS multi-user communication scenario modeling: As Figure 1 shown, the present invention studies the uplink of a multi-RIS / IRS-assisted multi-user communication system, where I distributed RIS / IRSs are deployed to assist the transmission from K single-antenna user groups to an access point (AP) with N A antennas. Each RIS / IRS (denoted by ) consists of N i reflecting elements. To minimize the path loss between the RIS / IRS and the corresponding user, we assume that the RIS / IRS is deployed near the user cluster in the actual scenario. The direct link between the AP and the user is blocked by obstacles. The AP communicates with each RIS / IRS through the control channel to adjust the phase shift of the reflecting elements. Since the positions of the access point and the RIS / IRS are fixed, for simplicity, we assume that the beamforming of the transmitter and receiver of the access point-RIS / IRS channel is fixed to focus on studying the beam scanning problem of the RIS / IRS-user.

[0085] In the millimeter-wave band, the wireless channel follows the geometric channel model. For a uniform planar array (UPA) with N i = N h ×N v antennas, N h is the number of horizontal array elements, and N v is the number of vertical array elements; the horizontal array element spacing is d h , and the vertical array element spacing is d v . Assuming that the azimuth angle of the incident wave from user k arriving at the i-th RIS / IRS is φ i，k , and the elevation angle is θ i，k , the steering vector can be expressed as

[0086]

[0087] where λ represents the wavelength of the incident wave, and Since the positions of the AP and the RIS / IRS are fixed, while the positions of the users are mobile, the AP-RIS / IRS channel can be regarded as a quasi-static channel with respect to the RIS / IRS-user channel. The beam alignment between the AP and the RIS / IRS can be determined according to the previous channel state information. Therefore, the uniform linear array of the AP can be equivalent to a single antenna. In addition, we assume that each user has a LoS channel with only a part of the RIS / IRS due to distance and obstacles, denoted as Item. and the cardinality are both unknown. The transmitted signal from user k arriving at the AP through the reflection channels of the RIS / IRS set can be written as

[0088]

[0089] where is the signal transmitted by user k with transmission power P, is the additive white Gaussian noise (AWGN), h i，k represents the complex gain of the user k-RIS / IRS i LoS channel, and g i represents the complex gain of the effective channel from the i-th RIS / IRS to the AP. represents the diagonal reflection matrix of the i-th RIS / IRS, where is the phase shift of the n-th reflection element of the i-th RIS / IRS. The superscript H is the conjugate transpose operation, the superscript r represents reception, and the superscript t represents transmission.

[0090] (2) Hash beam and codebook generation: The single-beam search mode of the prior art scans the entire space, which causes a large delay. The present invention adopts a multi-arm beam scanning method, which can greatly reduce the number of scans. In this scenario, different multi-arm beam scanning codebooks need to be generated for all RIS / IRSs:

[0091] (2.1) Two-dimensional hash beam codebook: In the horizontal / vertical direction, the azimuth / pitch angle of the beam space is discretized into N directions, and the set is represented as [1, 2,..., N]. The hash values of the keywords 1, 2,..., N are calculated using a hash function, and the storage location of the keyword, i.e., the bucket, can be found through the hash value. For example, Figure 2 , it is shown that 3 hash functions evenly hash the directions 1 to 16 in the space into 4 buckets. Therefore, the hashing result of the first hash function is that the first multi-arm beam consists of directions 1, 4, 11, and 15, the second multi-arm beam consists of directions 2, 6, 12, and 14, the third multi-arm beam consists of directions 3, 8, 10, and 16, and the fourth multi-arm beam consists of directions 5, 7, 9, and 13. For each RIS / IRS, L mutually independent hash functions are selected from the hash function family, and each hash function evenly hashes the keywords 1, 2,..., N into B buckets. A total of BL bucket groups can be obtained, and each bucket represents a multi-arm beam.

[0092] Figure 3 As shown in Figure 2 is the multi-arm beam image generated by the first-round hashing in iIt can be constructed from the DFT codebook: where the multi-arm beam codebook C i Each row of represents a multi-arm beam, denoted as C i =[c i，1 ,..., c i，B | T . Each multi-arm beam consists of R = N / B sub-beams, corresponding to the directions represented by the R keywords contained in the bucket. The codebook of each sub-beam is obtained by intercepting M adjacent elements from the row corresponding to the direction of the DFT codebook and can be expressed as where b(r) ∈ [1, 2,..., N] represents the keyword corresponding to the r-th sub-beam contained in the b-th multi-arm beam, MN represents the number of antenna elements, represents the b(r)-th row of the DFT codebook .

[0093] (2.2) 3D hash beam codebook: In practical situations, scanning in three-dimensional space directions is required. First, the azimuth angle and elevation angle in the beam space are discretized into N1 and N2 directions respectively. Therefore, there are N1·N2 directions [φ, θ] composed of the azimuth angle and elevation angle. For each RIS / IRS, L mutually independent hash functions are respectively selected from the family of hash functions. Each hash function evenly hashes the keywords 1, 2,..., N1·N2 into B buckets. A total of BL bucket groups can be obtained, and each bucket represents the sub-beam directions constituting a multi-arm beam. The corresponding multi-arm beam codebooks C1, C2,..., C I are still constructed by intercepting from the DFT codebook . As shown in Figure 4 , it shows a schematic diagram of the 3D hash beam for a RIS / IRS hashed with one hash function in the 3D space with N1×N2 = 16×16 directions, evenly hashed into B = 32 buckets, that is, each multi-arm beam contains 8 sub-beams. In the figure, the ranges of the φ-axis and θ-axis are from 0 to 180 degrees, representing the azimuth angle and elevation angle of the scanning beam in the 3D space.

[0094] The advantage of this joint hashing scanning scheme is that, compared with horizontal azimuth and vertical scanning respectively, it will not cause the superposition of scanning errors introduced in two dimensions, and can ensure higher accuracy under the condition of the same number of scanning times.

[0095] (3) Pilot signal transmission: In the uplink, each user continuously sends pilot signals to the AP during the beam scanning phase. During each training symbol duration, each RIS / IRS sequentially selects a row from its respective codebook, that is, uses a multi-arm beam to receive the pilot signals sent by each user and reflects them. Finally, they are superimposed at the AP. We selected L mutually independent hash functions for each RIS / IRS and performed L rounds of hash operations. Since the RIS / IRS can achieve simultaneous scanning, the total scanning time is Q = BL time slots. Denote the signal received by the AP from user k in Q time slots as

[0096]

[0097] q = lB + b, where y k (q) represents the signal received by the AP from user k in the b-th time slot of the l-th round of hashing.

[0098] (4) Demultiplexing the superimposed signal: According to the signals received in each time slot, the AP can separate the signals from different users based on existing modulation and demodulation techniques. For each user, the methods of demultiplexing and voting are the same and carried out simultaneously. Since the AP receives the signals superimposed after being reflected by multiple RIS / IRS reflection paths from user k, in order to identify the alignment directions of different RIS / IRS with respect to user k, the prerequisite condition that no two or more RIS / IRS scan user k simultaneously needs to be satisfied. This prerequisite assumption can be met by setting the number of buckets B: In each time slot, due to the randomness of the hash function, the probability that a RIS / IRS scans user k is Due to the independence of the hash functions from each other, the probability that two RIS / IRS scan the same user As long as the probability value of Pr is small enough, it can be ensured that y k (q) contains at most the reflected signal of one RIS / IRS. When B = 8, this requirement can be basically met, that is

[0099]

[0100] where γ i，k represents the direction from the i-th RIS / IRS to user k, D i (q) represents the set of directions included in the q-th multi-arm beam of the i-th RIS / IRS, s i，k (q) represents the actual signal reflected by user k through the i-th RIS / IRS, and n(q) represents the noise in the q-th time slot. In addition, due to the limited transmission power of the user and the different attenuations of each RIS / IRS reflection path, we assume that the signal-to-noise ratio (SNR) of the signals reflected from different RIS / IRS is as follows: Based on the SNR of the received signal, signals reflected by different RIS / IRSs can be separated from the Q signals: the largest L SNRs correspond to the reflection paths of RIS / IRS 1, the second largest L SNRs correspond to RIS / IRS 2, and so on until the SNRs of the remaining time slots approach 0, that is, the received signal only contains noise or interference signals. The phase information or channel state information of the received signal is not used here, and it is only based on the SNR value of the received signal.

[0101] (5) Decision and voting mechanism: If we vote on any L multi-arm beams of the i-th RIS / IRS, due to the randomness of the hash function hashing, only when voting on the multi-arm beams of the signal containing the user can we obtain the highest number of votes η i and the best alignment direction γ i , the probability that the number of votes of the remaining (j-th) RIS / IRS is higher almost approaches zero, which is an impossible event. Therefore, only by voting according to the information obtained by demultiplexing can we obtain the best alignment direction of the RIS / IRS assisting user communication to the user. Specifically, according to the calculated SNR of the received signal, we first find the largest L value from it, and the corresponding time slot sequence number is defined as The same decision and voting operations are performed for each RIS / IRS, and the sub-beams in the multi-arm beams of the time slot are voted on, and the direction with the highest number of votes is the best alignment direction of the corresponding RIS / IRS i to the user. Then, the above voting method is repeated for the time slots with the second largest L SNRs until the alignment directions of all RIS / IRSs are calculated.

[0102] Figure 7Shows the multi-arm beam voting mechanism in three-dimensional beam space. Each beam direction is jointly composed of azimuth and elevation angles [φ, θ]. There are two RIS / IRS in the space simultaneously assisting the communication between user k and the AP. The channel gain of the reflection path of RIS / IRS 1 is greater than that of the reflection path of RIS / IRS 2. The spatial position of the user is in the direction [3, 3] of RIS / IRS 1 and in the direction [2, 4] of RIS / IRS 2. The table in the figure represents evenly hashing 4×4 directions in the space into 4 multi-arm beams, and performing 4 rounds of hashing. The left table represents RIS / IRS 1, and the right table represents RIS / IRS 2. Each row of the table represents a multi-arm beam, which is composed of 4 sub-beam directions. The known information after demultiplexing has the maximum received signal SNR in the time slots marked in light gray, and the received signal SNR in the time slots marked in dark gray is the second largest. Theoretically, the signal in the light gray time slot is reflected via the strongest RIS / IRS reflection path and only contains the signal reflected by RIS / IRS 1; the signal in the dark gray time slot is reflected via the second strongest RIS / IRS reflection path and only contains the signal reflected by RIS / IRS 2. As Figure 7 shown in (a) therein, voting for each sub-beam in the multi-arm beam marked in light gray, it can be obtained that the direction [3, 3] in the table of RIS / IRS 1 has the highest number of votes, which is 4 votes, while the number of votes for the direction [2, 3] with the highest number of votes in RIS / IRS2 is less than the number of votes for the direction [3, 3] in RIS / IRS 1. Thus, it can be obtained that the channel gain of the reflection path of RIS / IRS 1 is the strongest, and user k is in the [3, 3] direction of RIS / IRS 1. Continuing to vote on the multi-arm beam marked in dark gray, as Figure 7 shown in (b) therein, the direction [2, 4] in the table of RIS / IRS 2 has the highest number of votes. Thus, it can be obtained that the channel gain of the reflection path of RIS / IRS 2 is the second strongest after RIS / IRS 1, and user k is located in the [2, 4] direction of RIS / IRS 2.

[0103] The functions and effects of the present invention are further illustrated and demonstrated through the following simulation experiments:

[0104] (1) Single RIS / IRS multi-user scenario

[0105] (1.1) Simulation conditions

[0106] Set the number of users to K = 5 and the number of RIS / IRS to I = 1. For simplicity, we consider a group of RIS / IRS deployed in the y-z plane and arranged parallel to each other on the y-axis. These RIS / IRS are equipped with N v = 32, N h= 32 reflection units, where the first RIS / IRS is located at the center of the coordinate axis, i.e., (0, 0, 0) meters (m). The users are distributed on the x-y plane, so the elevation angle from the RIS / IRS to each user Since the beamforming in the vertical direction of the RIS / IRS does not depend on the user's position, only the beam alignment between the horizontal array of the RIS / IRS and the users needs to be considered. The AP centered at (16, 16, 0) meters consists of N A = 64 antennas, and the antenna spacing

[0107] During the simulation, we compared the scanning scheme of the present invention with the exhaustive search method, hierarchical search method, and equidistant search method in the prior art in terms of two metrics: scanning accuracy and complexity.

[0108] (1.2) Simulation results

[0109] Figure 8 The influence of the signal-to-noise ratio on the beam recognition accuracy is plotted when the fixed number of directions N = 32. Under the same simulation scenario settings, we performed beam scanning using the exhaustive, hierarchical, equidistant, and hash multi-arm beam scanning methods respectively. It can be found that, first, as the signal-to-noise ratio increases, the influence of noise becomes smaller, and the accurate recognition rates of all four scanning methods gradually increase. When the SNR value is greater than 0 dB, the successful recognition rate tends to converge. Second, in this simulation scenario, as long as the number of multi-arm beams B is greater than 2, the proposed hash function-based multi-arm beam scanning has obvious advantages over the equidistant scanning method and can achieve an accurate recognition rate of about 99.5%. Even when the signal-to-noise ratio is quite small, the recognition rate of hash is significantly improved compared with the equidistant scanning and hierarchical scanning methods. This is expected because the independence of the hash function and the independence and randomness of the hash function reduce the leakage interference between sub-beams or different multi-arm beams.

[0110] Figure 9 Shows the influence of different numbers of directions on the scanning time of beam scanning under the conditions of fixed signal-to-noise ratio and accurate recognition rate in this simulation scenario. Among them, the hierarchical scanning, equidistant scanning, and hash multi-arm beam scanning are at the logarithmic level, and compared with the traditional exhaustive scanning, the training overhead can be significantly reduced. In addition, compared with the equidistant method, the hash-based method can further reduce the number of scans. Although the number of scans in the hierarchical scanning is the least in the figure, since the hash multi-arm beam scanning only makes a voting decision at the AP end after multiple rounds of scanning, while the hierarchical scanning needs to make a decision at each scan to determine the next level of sub-beams, the hierarchical scanning will cause higher latency in practical applications.

[0111] (2) Multi-RIS / IRS assisted scenario

[0112] (2.1) Simulation conditions

[0113] Assume the number of users is K = 3 and the number of RIS / IRS is I = 3. For simplicity, we consider a set of RIS / IRS deployed in the yz plane and arranged in parallel on the y axis. These RIS / IRS are equipped with N v =32, N h = 32 reflection units, where the first RIS / IRS is located at the center of the coordinate axis, i.e. (0, 0, 0) meters (m). Users are distributed on the xy plane, so the elevation angle from RIS / IRS to each user is Since the beamforming in the vertical direction of the RIS / IRS does not depend on the position of the user, only the beam alignment between the RIS / IRS horizontal array and the user needs to be considered. The AP centered at (16, 16, 0) meters is represented by N A = 64 antennas, antenna spacing

[0114] During the simulation, we compared the scanning scheme of the present invention with the exhaustive search method, hierarchical search method and equal-interval search method in the prior art in terms of scanning accuracy and complexity. The exhaustive search method uses a single-beam scanning method in which each RIS / IRS takes turns.

[0115] (2.2) Simulation results

[0116] Figure 10 The effect of the signal-to-noise ratio on the beam recognition accuracy when the number of directions N=32 is fixed is plotted. Under the same simulation scenario setting, we use exhaustive, hierarchical, equally spaced and hash multi-arm beam scanning for beam scanning respectively. It can be found that, first, as the signal-to-noise ratio increases, the influence of noise becomes smaller, and the accurate recognition rate of all four scanning methods gradually increases. When the SNR value is greater than 0dB, the successful recognition rate tends to converge. Secondly, in this simulation scenario, exhaustive scanning, hierarchical scanning and equally spaced scanning will all fail to varying degrees because they cannot demultiplex the superimposed signals. The hash multi-arm beam scanning method in the embodiment of the present invention is based on the randomness and independence of the hash function itself, and can realize the effective demultiplexing and decision voting. When the number of beams B≥4, the hash multi-arm beam scanning can achieve a relatively high accuracy; when the number of beams B=8, the hash multi-arm beam scanning can achieve an accurate recognition rate of about 97.5%, which is the highest accuracy except for exhaustive search. This is because when B = 8, it guarantees the probability that two RIS / IRS will scan the same user at the same time. Small enough to ensure the preconditions of the scanning method we proposed. Even when the signal-to-noise ratio is quite small, compared with the equal-interval scanning and hierarchical scanning methods, the recognition rate of hashing is significantly improved. This is expected because the independence of the hash function and the independence and randomness of the hash function reduce the leakage interference between sub-beams or different multi-arm beams.

[0117] Figure 11 Shows the influence of different numbers of directions on the scanning time of beam scanning under the conditions of fixed signal-to-noise ratio and accurate recognition rate in this simulation scenario. The exhaustive search method scans each RIS / IRS in turn, and the scanning time is proportional to the number of RIS / IRS, resulting in a very large scanning overhead. Hierarchical scanning, equal-interval scanning, and hashing multi-arm beam scanning are at the logarithmic level. Compared with the exhaustive search scan, the training overhead can be significantly reduced. Although the number of scans in hierarchical scanning is the least in the figure, because the hashing multi-arm beam scanning only makes a voting decision at the AP end after multiple rounds of scanning, while hierarchical scanning needs to make a decision at each scan to determine the next layer of subdivision beams, hierarchical scanning will cause higher latency in practical applications. In addition, compared with the equal-interval method, the hashing-based method can further reduce the number of scans, and it is independent of the number of RIS / IRS and the number of users. The number of scans Q = BL = O(BlogN).

[0118] From the above results, it can be seen that the present invention not only has high accuracy in beam scanning alignment, but also has the characteristics of small energy consumption overhead and low latency.

[0119] The above embodiments are used to explain the present invention, rather than to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims of the present invention fall within the protection scope of the present invention.

Claims

1. A fast and low-complexity beam scanning method, characterized in that, The method includes the following steps: Step 1, perform RIS / IRS-assisted communication scenario modeling; Step 2, generate hash multi-arm beams and codebooks: discretize the beam space into N directions, and for each RIS / IRS, hash it with L mutually independent randomly generated hash functions, where each hash function evenly hashes these N directions into B multi-arm beams, and construct the codebook corresponding to each RIS / IRS; The codebook corresponding to each RIS contains BL rows, and each row consists of R equally long segments, representing a multi-arm beam containing R = N / B sub-beams, and each segment corresponds to a sub-beam direction; Step 3, pilot signal transmission: in the uplink, each RIS / IRS uses a multi-arm beam to receive the pilot signal sent by each user and reflect it to the access point, and finally superimpose it at the access point; Step 4, the access point uses modulation and demodulation technology to separate the signals from different users received in each time slot, and then demultiplex the superimposed signals of the same user received; the demultiplexing includes: The first round of loop, the time slots corresponding to the L values with the largest soft decision signal-to-noise ratio, and the time slots contain the signals reflected by the user through the strongest RIS reflection path; The second round of loop, the time slots corresponding to the L values with the second largest soft decision signal-to-noise ratio, and the time slots contain the signals reflected by the user through the second strongest RIS / IRS reflection path; and so on until the loop end condition is reached; The loop condition is as follows: if the number of RIS / IRS is known, the number of loop rounds is equal to the number of RIS / IRS; if the number of RIS / IRS is unknown, the loop stops until the remaining time slots only contain noise and interference signals; Step 5, perform decision voting. The access point votes for each sub-beam direction in the corresponding multi-arm beam of each RIS / IRS according to the time slots containing the user signals obtained by demultiplexing in Step 4, and selects the RIS / IRS direction with the highest number of votes as the best alignment direction of the user.

2. A fast and low-complexity beam scanning method according to claim 1, wherein The RIS / IRS-assisted communication scenario is that RIS / IRS assists users and the access point in communication. The RIS / IRS consists of tunable reflection elements, and the access point communicates with the RIS / IRS through a control channel to adjust the phase shift of the reflection elements.

3. A fast and low-complexity beam scanning method according to claim 1, characterized in that The number L of the hash functions can be proven and verified by simulation, and the formula is as follows: L = O(logN).

4. A fast and low-complexity beam scanning method according to claim 1, characterized in that In Step 2, the codebook is a two-dimensional hash multi-arm beam codebook, including: In the horizontal / vertical direction, discretize the azimuth angle / elevation angle of the beam space into N directions, and the direction set is represented as [1, 2,..., ]; calculate the hash values of the keywords 1, 2,... using the hash function, and the storage location of the keyword, that is, the bucket, can be found through the hash value as an index; select L mutually independent hash functions from the hash function family, and use each hash function to evenly hash the keywords 1, 2,... into B buckets, and a total of BL groups of buckets are obtained, and each bucket represents a multi-arm beam.

5. A fast and low-complexity beam scanning method according to claim 1, characterized in that, In Step 2, the codebook is a three-dimensional hash multi-arm beam codebook, including: The azimuth and elevation angles in the beamspace are discretized into N1 and N2 directions respectively, and there are N1N2 directions [φ,θ] composed of azimuth and elevation angles; L mutually independent hash functions are selected from the hash function family, and each hash function evenly hashes the keywords 1, 2,..., 1N2 into B buckets, resulting in a total of BL bucket groups, and each bucket represents a multi-arm beam.

6. A fast and low-complexity beam scanning method according to claim 1, characterized in that The method of demultiplexing and voting for each user is the same and is carried out simultaneously.

7. A fast and low-complexity beam scanning method according to claim 1, characterized in that In the single RIS / IRS multi-user communication scenario, since there is only a single RIS / IRS assisting communication, for each user, the signal received by the access point is only reflected by a single RIS / IRS; according to the power of the received signal, it is determined whether the received signal in each time slot contains the useful signal from the user by means of hard decision or soft decision.

8. A fast and low-complexity beam scanning method according to claim 7, characterized in that, In the single RIS / IRS multi-user communication scenario, the hard decision is: for the received signal in each time slot, if the signal power value is greater than the preset threshold, it is determined that the user signal is included, otherwise it is determined that only noise and interference signals are included.

9. A fast and low-complexity beam scanning method according to claim 7, characterized in that In the single RIS / IRS multi-user communication scenario, the soft decision is: among the received signals in a total of BL time slots, the L time slots corresponding to the L largest signal power values contain the user signal, that is, the multi-arm beam of the RIS / IRS can scan to the user, and the remaining time slots only contain noise or interference signals.

10. A fast and low-complexity beam scanning method according to claim 1, characterized in that In the multi-RIS / IRS scenario, the voting process in step 5 is specifically as follows: According to the received signal-to-noise ratio, first find the largest L values from it, and the corresponding time slot sequence numbers are defined as The same decision and voting operations are performed for each RIS / IRS, and the sub-beams in the multi-arm beam of the time slot are voted on. The direction of the RIS / IRS with the highest number of votes is the best alignment direction of the user; then repeat the above voting method for the time slots with the second largest L signal-to-noise ratios until the alignment directions of all RIS / IRSs are calculated.

Images