An adaptive millimeter wave beam search method

Abstract

The present invention discloses an adaptive millimeter wave beam search method. The steps of the present invention are as follows: 1. beam search scenario, preset hierarchical beam and narrow beam search codebooks; 2. overall process of using hierarchical beam search codebooks and narrow beam search codebooks to perform adaptive beam search; 3. when using hierarchical beam search codebooks for beam search, a generalized likelihood ratio test is used to determine whether to switch the beam search codebook and whether to stop the hierarchical search of the current level and enter the next level; 4. when the hierarchical search starts a new layer of search, an angle interval to be searched is selected; 5. when the conditions for switching from the hierarchical beam search codebook to the narrow beam search codebook are met, a narrow beam search codebook is used for beam search; 6. when all beam searches are stopped, the final beamforming direction is selected. The present invention can realize the switching between the hierarchical beam codebook and the narrow beam codebook, thereby reducing the time overhead of the beam search while ensuring the accuracy of the beam alignment.

Description

Technical Field

[0001] The present invention belongs to the field of millimeter wave communication technology, and proposes an adaptive millimeter wave beam search method, which is a method that can adaptively decide to continuously execute a multi-precision layered beam search codebook for beam search or switch to a narrow beam search codebook for beam search, thereby adapting to different signal-to-noise ratio conditions and reducing beam search time overhead. Background Art

[0002] Millimeter wave communication is considered one of the key technologies for achieving high-speed data transmission. In millimeter wave communication systems, multiple antennas are usually used in conjunction with beamforming technology to improve the directional gain of the signal, thereby compensating for the path loss during transmission. In millimeter wave communication, the beam direction of the transmitter and receiver should be aligned with the main path direction of the communication channel as much as possible to achieve beam alignment, so as to obtain maximum gain.

[0003] The beam search method based on spatial scanning is one of the important means to achieve beam alignment. According to the different beam search codebooks, the beam search method based on spatial scanning can be divided into two categories. One is the exhaustive search (Exhaustive Search), which mainly uses narrow beams to traverse and search each beam pair to find the transceiver beam that best matches the main path direction of the channel. Because the exhaustive search needs to traverse all beam pairs, the beam scanning time is long and consumes more resources. The other is the hierarchical search (Hierarchical Search), which uses hierarchical multi-precision beam search codebooks of different widths to perform beam search layer by layer. First, a wider beam search codebook is used to determine the local angle interval of the main path of the channel, and then the beam width is reduced in the determined local angle interval to continue the search; by continuously reducing the local angle interval and the beam width used, the best narrow beam matching the communication channel is found. The hierarchical search method performs well under high signal-to-noise ratio conditions, but under low signal-to-noise ratio conditions, it usually takes longer beam search time than the exhaustive search method to achieve similar beam alignment accuracy.

[0004] In summary, the two existing classic millimeter wave beam search methods can only achieve good beam alignment performance under the applicable signal-to-noise ratio conditions, and cannot adapt to the complex and changeable mobile millimeter wave communication scenarios. In order to solve this problem, it is necessary to design a new type of beam search method that can not only adaptively select the appropriate beam search codebook according to different communication scenarios, but also adjust the beam search time to achieve satisfactory beam search results with minimal resource consumption. Summary of the invention

[0005] The present invention discloses an adaptive millimeter wave beam search method for complex millimeter wave communication systems. The method disclosed by the present invention uses a generalized likelihood ratio test to determine the unknown channel signal-to-noise ratio in real time, and can realize the switching between the layered search beam codebook and the narrow beam codebook, so that the appropriate beam search codebook can be adaptively selected under high and low signal-to-noise ratios, thereby reducing the beam search time overhead while ensuring the accuracy of beam alignment.

[0006] The present invention provides an adaptive millimeter wave beam search method, comprising the following steps:

[0007] Step 1: beam search scenario, preset multi-precision hierarchical beam search codebook and narrow beam search codebook;

[0008] Step 2: The overall process of performing adaptive beam search using a hierarchical beam search codebook and a narrow beam search codebook;

[0009] Step 3: When using the hierarchical beam search codebook for beam search, a generalized likelihood ratio test is used to determine whether to switch the beam search codebook and whether to stop the hierarchical search at the current level and enter the next level;

[0010] Step 4: When starting a new layer of search, select the angle interval to be searched;

[0011] Step 5: When the condition for switching from the hierarchical beam search codebook to the narrow beam search codebook is met, the narrow beam search codebook is used to perform beam search;

[0012] Step 6: When all beam searches are stopped, select the final beamforming direction.

[0013] Preferably, the beam search scenario described in step 1 includes a transmitter and a receiver, wherein the transmitter continuously transmits a pilot signal to the entire angular space covered by it, and the receiver uses a preset beam search codebook to perform beam scanning to find the optimal beamforming direction.

[0014] Preferably, the beam search codebook in step 1 is divided into two types, one is a hierarchical beam search codebook W, and the other is a narrow beam search codebook

[0015] Preferably, the hierarchical beam search codebook in step 1 has a tree structure, such as Figure 1 As shown, the coverage area of ​​a wider beam in the upper layer is jointly covered by multiple narrower beams in the lower layer. The layered beam search codebook used in the present invention has a total of K layers, of which the kth layer has a total of 2 k codewords, the codeword set of the kth layer is represented by W k ,and represents the i-th codeword in the k-th layer, N R =2 K Indicates the number of antennas equipped by the receiver. Codeword w (k,i) The angle range covered is Indicates that It is the result of taking the cosine of the coverage angle. In the layered beam search codebook, the union of the coverage ranges of all codewords in any layer should completely cover the entire antenna array angle domain; and the coverage range of a codeword in any layer can be completely represented by the union of the coverage ranges of the two codewords in the corresponding next layer.

[0016] Preferably, the narrow beam search codebook in step 1 is a single layer, and the narrow beams jointly cover the entire angular space. The present invention uses all codewords W in the last layer of the layered beam search codebook. K As a narrow beam search codebook, it is expressed as

[0017] Preferably, the beam search described in step 2 starts from a hierarchical beam search codebook to perform beam scanning.

[0018] Preferably, the overall process of beam search described in step 2 is as follows: Figure 2 As shown. Specifically, when the hierarchical beam search codebook is used to perform beam scanning, the receiving end determines whether the conditions for switching to the narrow beam search codebook for beam search are met after completing a cycle of pilot signal acquisition each time. If the conditions for switching to the narrow beam search codebook are met, the hierarchical beam search codebook is stopped for beam search, and the narrow beam search codebook is used for beam search instead until the beam search stops. If the conditions for switching to the narrow beam search codebook are not met and the current hierarchical search level is not the last level, it is determined whether to enter the next level of hierarchical search; if the conditions for entering the next level of hierarchical search are met, the next level of beam search codebook is used to perform beam search. If the conditions for switching to the narrow beam search codebook are not met and the current hierarchical search level is the last level, it is determined whether the stop conditions for beam search are met. If none of the above conditions are met, the hierarchical search of the current level continues.

[0019] Preferably, the beam search process described in step 2 selects the final beamforming direction when the beam search stop condition is met.

[0020] Preferably, the hierarchical search receiving end in step 2 completes a cycle of pilot signal acquisition, which means that the receiving end sequentially uses all the candidate beams of the current layer to acquire the pilot signal transmitted by the receiver once.

[0021] Preferably, the switching from the layered beam search codebook to the narrow beam search codebook in step 2 is determined by detecting the signal-to-noise ratio of the unknown channel. The reason is that the layered beam search codebook is suitable for scenarios with high signal-to-noise ratios, while the narrow beam search codebook is suitable for scenarios with low signal-to-noise ratios. The beam search codebook switching mechanism proposed in the present invention can perfectly make up for the defects caused by using a single beam search codebook. Under different signal-to-noise ratio conditions, the most suitable beam search codebook can be used for beam search, thereby improving the accuracy of beam search and reducing resource consumption.

[0022] Preferably, when performing beam search using the hierarchical beam search codebook in step 3, the transmitter uses a beam covering the angle range of interest. Transmit signal, N T Indicates the number of antennas equipped by the transmitter. The receiver collects the pilot signal transmitted by the transmitter in a cyclic scanning manner. In each cyclic scanning, the receiver uses the codeword w (k,i) and w (k,i+1) Perform beamforming and receive the pilot signal sent by the transmitter. Assume that the time for a single beam to collect signals in each cyclic scan at the receiving end is n0 ≥ 1 symbol period. After m cyclic scans, the receiving end uses the jth codeword w of the kth layer, j∈{i,i+1} (k,j) The signal received by the corresponding beam can be expressed as:

[0023] y (k,j) =h (k,j) s (k,j) +z (k,j) (1)

[0024] in represents the equivalent channel between the transmitter and the receiver, represents the channel matrix between the transmitter and the receiver, represents the pilot signal sent and P T Indicates the transmit power, represents circularly symmetric Gaussian noise. The signal after matched filtering can be expressed as

[0025]

[0026] in Represents circularly symmetric Gaussian noise, and the signal-to-noise ratio of the channel can be obtained as

[0027] Preferably, in order to detect the current channel signal-to-noise ratio in step 3, a channel signal-to-noise ratio target value β is set in advance, and it is set that when the channel signal-to-noise ratio is greater than the target value, it is considered to be a high channel signal-to-noise ratio, otherwise it is a low channel signal-to-noise ratio. The above setting can be written as the following assumption:

[0028]

[0029] However, η in the above formula (k,j) It cannot be directly calculated by receiving the signal, so the above hypothesis cannot be directly tested. The present invention uses a generalized likelihood ratio test to perform hypothesis testing, and the test statistic can be obtained:

[0030]

[0031] Where γ∈(0,∞) represents the threshold value of hypothesis testing, and Respectively expressed in and Likelihood function under the condition. After calculation and derivation, formula (4) can be rewritten as:

[0032]

[0033] In formula (5), the present invention assumes that the noise variance σ 2 is the only known channel information, and r (k,j) , m, and n0 can be obtained through measurement, so the test statistic G(r (k,j) ). Therefore, the switching criteria of the beam search codebook are as follows:

[0034]

[0035] After each cyclic scan at the receiving end, a channel signal-to-noise ratio test is performed. (k,j) ) satisfies the condition of formula (6), the beam search codebook is switched and the narrow beam search codebook is used for beam search. If the detection result does not satisfy the condition of formula (6), the hierarchical beam search codebook is continued to be used for beam search.

[0036] Preferably, in step 3, when performing hierarchical search, a generalized likelihood ratio test is used to detect whether the signal-to-noise ratio of the cumulative received signal meets the preset target value. When the current level is the kth level, it can be seen from the output signal formula (2) of the matched filter that the signal-to-noise ratio of the received signal is Let the preset target value of the signal-to-noise ratio of the cumulative received signal be ρ, then according to the generalized likelihood ratio test, the test statistic that can be obtained is:

[0037]

[0038] After each cyclic beam scan, if the following conditions are met:

[0039]

[0040] Then stop the beam search of the current level (kth level), otherwise continue the beam search of the current level, where γ' is the set likelihood threshold value.

[0041] Preferably, when the hierarchical search in step 4 enters a new level of search, if k=1, that is, when the hierarchical search is initially entered, the angle interval to be searched is determined as The beam to be searched is w (1,1) and w (1,2) The corresponding beam.

[0042] Preferably, when the hierarchical search in step 4 enters the k+1th layer (k>=1), the matched filter signal r accumulated by the receiving end when the kth layer beam search stops is (k,i) and r (k,i+1) , determine the beam search angle interval of the k+1th layer. Specifically, there are three possibilities for the beam search angle interval of the k+1th layer to be determined, such as Figure 3 The specific rules for determining it are as follows: When the estimated value of the channel main path angle is When selecting CV w(k,i) As the beam search angle interval of the k+1th layer; when the estimated value of the channel main path angle When selecting CV w(k,i+1) As the beam search angle interval of the k+1th layer; when the estimated value of the channel main path angle When selecting CV w(k,*) As the beam search angle interval of the k+1th layer. (k,i) and θ (k,i+1) Respectively represent the angle interval and The central angle of the channel, the estimated value of the main path angle From formula (9), we get:

[0043]

[0044] in represents the value range of θ, express and The angle of the junction.

[0045] in

[0046]

[0047] represents the steering vector at the receiving end, g * (k,i) (θ) represents g (k,i) The complex conjugate of (θ), represents the beamforming gain at the receiving end. Therefore, the angle interval to be searched at the k+1th layer can be summarized as:

[0048]

[0049] when Just choose As the beam search angle interval of the k+1th layer, the beams corresponding to all codewords in the interval are the beams to be searched in the k+1<Kth layer. is the center of the interval, the width and The equal angle interval is used as the beam search angle interval of the k+1th layer, and the beams corresponding to all codewords in the interval are the beams to be searched of the k+1<Kth layer.

[0050] Preferably, when performing beam search using the narrow beam search codebook in step 5, the active beam set is first initialized. And start to perform cyclic beam scanning.

[0051] Preferably, step 5 performs a cyclic scan on all beams in the active beam set, sending n0 pilot signals each time. After completing the mth cyclic scan, beam w i Signal received from the first scan It can be expressed as:

[0052] y i =h i s+z i (11)

[0053] in, is the equivalent channel between the sender and the receiver, is the pilot signal, and Where P T is the antenna transmission power, The variance is σ 2 Independent circularly symmetric Gaussian variables. i After the corresponding beam is scanned m times, the accumulated pilot signal is mn0, and we can get:

[0054]

[0055] in, is the signal after matched filtering, T i The degree of freedom is 2 and the noncentrality parameter is Noncentral chi-square distribution of .

[0056] Preferably, after each cycle of scanning the active beams in step 5, the active beam set is updated according to the posterior probability of the comparison of all beam intensities. The specific rule is: when f(T i,T j )>Γ, beam i is considered stronger than beam j, and beam j is removed from the active beam set; where f(T i ,T j ) indicates that when the accumulated pilot signals corresponding to beam i and beam j are T i and T j When (see formula 12), the posterior probability that the strength of the equivalent channel of beam i is higher than that of beam j is f(T i ,T j )=P r {|h i |>|h j |}, Γ represents the posterior probability threshold. The above rules are applied to the pairwise comparison of all beams in the active beam set. Therefore, after each cyclic scan, the updated active beam set can be expressed as:

[0057]

[0058] Among them, the posterior probability function f(x,y) can be expressed as follows:

[0059]

[0060] Where L l (·) is a Laguerre polynomial of order l. The posterior probability can be approximately calculated by summing a finite number of terms l according to the analytical expression of formula (14); it can also be approximately calculated using the Monte Carlo method, and the calculation result is stored in a lookup table and called in real time.

[0061] Preferably, in step 5, after each update of the active beam set, it is determined whether the following three stop conditions are met. If any one of the conditions is met, the narrow beam search stops.

[0062] Stop condition 1. There is only one beam codeword left in the active beam set:

[0063]

[0064] Stop condition 2. There are two adjacent codewords left in the active beam set and the following conditions are met:

[0065]

[0066] Stop condition 3. Limitation of the number of pilot signals:

[0067]

[0068] Where N max It is the maximum number of pilot signals allowed in the entire beam search process.

[0069] Preferably, in step 6, when the layered beam search codebook is used to search the last layer K, the beamforming direction obtained is as follows:

[0070]

[0071] Among them, w (K,i) represents any codeword in the last layer Kth layer beam search codebook, ⊙ represents the Hadamard product, express and w (K,i) The difference in the main lobe center direction angle, d represents the antenna spacing, and λ represents the signal wavelength.

[0072] Preferably, in step 6, when the narrow beam search codebook is used for beam search, when stop condition 1 or condition 3 is met, the beamforming direction is obtained as:

[0073] w opt =w i* (19)

[0074] in Indicates the last remaining active beam set The codeword corresponding to the strongest beam in .

[0075] When stop condition 2 is met, the beamforming direction is obtained as:

[0076]

[0077] ⊙ represents the Hadamard product, express and w (K,i) The difference in the main lobe center direction angle, d represents the antenna spacing, and λ represents the signal wavelength.

[0078] Compared with the existing beam search technology, the present invention has the following beneficial effects:

[0079] The present invention creatively designs a beam search codebook switching mechanism, which can switch to the most appropriate beam search codebook in a timely manner according to the change of the channel signal-to-noise ratio, realize adaptive beam search, and reduce the resource consumption of beam search while ensuring the accuracy of beam search. This technical solution is applicable to various communication environments, and can ensure satisfactory beam search effects regardless of whether the channel signal-to-noise ratio is high or low. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 Schematic diagram of the hierarchical beam search codebook structure.

[0081] Figure 2 This is a flow chart of the adaptive beam search method of the present invention.

[0082] Figure 3 Schematic diagram of the traditional hierarchical search angle interval division method.

[0083] Figure 4 It is a schematic diagram of the hierarchical search angle interval division method of the present invention.

[0084] Figure 5 This is a comparison chart of the spectrum efficiency of the adaptive beam search method of the present invention.

[0085] Figure 6 It is a comparison diagram of the adaptive beam search method of the present invention in terms of beam search time. DETAILED DESCRIPTION

[0086] The technical content of the present invention is described in detail and specifically below with reference to the accompanying drawings.

[0087] Step 1: beam search scenario, multi-precision hierarchical beam search codebook and narrow beam search codebook used.

[0088] Consider a unilateral beam search scenario, including a transmitter and a receiver. The transmitter uses a fixed beam to send a pilot signal, and the receiver searches for the best beam direction through beam scanning. Let the receiver be equipped with N R =2 K A uniform linear array with antennas and half-wavelength spacing, with a multi-precision hierarchical beam search codebook W and a narrow beam search codebook Where K represents the number of layers in the hierarchical beam search codebook, and the codeword set of the kth layer is represented by W k , the i-th codeword in the k-th layer of the hierarchical beam search codebook is used indicates that, and w (k,i) The angle range covered is Indicates that is the result of taking the cosine of the covered angular range. All codewords W in the last layer of the hierarchical beam search codebook are k As a narrow beam search codebook, it is expressed as where w i is a codeword, and The sending end is equipped with N T A half-wavelength uniform linear array of antennas using a fixed codeword The corresponding beam transmits the pilot signal. Assuming that during the beam search, the channel is a constant block fading model, the channel matrix is

[0089] like Figure 1 The hierarchical beam search codebook structure diagram shown in FIG. 1 satisfies the following two criteria:

[0090] Criterion 1: The coverage range of a codeword in any layer of the hierarchical beam search codebook can be fully represented by the union of the coverage ranges of the two codewords in the corresponding next layer:

[0091]

[0092] Criterion 2: The union of all codeword coverages at any layer of the hierarchical beam search codebook should completely cover the angular domain of the entire antenna array:

[0093]

[0094] Step 2: The overall process of performing adaptive beam search using a hierarchical beam search codebook and a narrow beam search codebook.

[0095] The overall flow chart of the adaptive beam search proposed by the present invention is as follows: Figure 2 As shown. Specifically, when the hierarchical beam search codebook is used to perform beam scanning, the receiving end determines whether the conditions for switching to the narrow beam search codebook for beam search are met after completing a cycle of pilot signal acquisition each time. If the conditions for switching to the narrow beam search codebook are met, the hierarchical beam search codebook is stopped for search, and the narrow beam search codebook is used instead for beam search until the beam search stops. If the conditions for switching to the narrow beam search codebook are not met and the current hierarchical search level is not the last level, it is determined whether to enter the next level of the hierarchical search; if the conditions for entering the next layer of hierarchical search are met, the hierarchical beam search codebook of the next level is used to perform beam search. If the conditions for switching to the narrow beam search codebook are not met and the current hierarchical search level is the last level, it is determined whether the stop conditions for the beam search are met. If none of the above conditions are met, the hierarchical search of the current level continues.

[0096] Step 3: When using the hierarchical beam search codebook for beam search, a generalized likelihood ratio test is used to determine whether to switch the beam search codebook and whether to stop the hierarchical search at the current level and enter the next level;

[0097] In the hierarchical search method, the transmitter uses a beam covering the angle range of interest. Transmit signal, N T Indicates the number of antennas equipped by the transmitter. The receiver collects the pilot signal transmitted by the transmitter in a cyclic scanning manner. In each cyclic scanning, the receiver uses the codeword w (k,i) and w (k,i+1) Perform beamforming and receive the pilot signal sent by the transmitter. Assume that the time for a single beam to collect signals in each cyclic scan at the receiving end is n0 ≥ 1 symbol period. After m cyclic scans, the receiving end uses the jth codeword w of the kth layer, j∈{i,i+1} (k,j)The signal received by the corresponding beam can be expressed as:

[0098] y (k,j) =h (k,j) s (k,j) +z (k,j) (twenty three)

[0099] in represents the equivalent channel between the sender and the receiver, Contains all pilot signals scanned in m cycles, Contains circularly symmetric Gaussian noise for all m cyclic scans with variance σ 2 The output signal after matched filtering can be expressed as:

[0100]

[0101] Among them, P T represents the transmit power of the antenna, The variance is still σ 2 The channel signal-to-noise ratio can be expressed as

[0102] The switch from the hierarchical beam search codebook to the narrow beam search codebook is determined by detecting the channel signal-to-noise ratio.

[0103] In order to detect the current channel signal-to-noise ratio, a channel signal-to-noise ratio target value β is set in advance, and it is set that when the channel signal-to-noise ratio is greater than the target value, it is considered to be a high channel signal-to-noise ratio, otherwise it is a low channel signal-to-noise ratio. The above settings can be written as the following assumptions:

[0104]

[0105] However, η in the above formula (k,j) It cannot be directly calculated by receiving the signal, so the above hypothesis cannot be directly tested. The present invention uses a generalized likelihood ratio test to perform hypothesis testing, and the test statistic can be obtained:

[0106]

[0107] Where γ∈(0,∞) represents the threshold value of hypothesis testing, and Respectively expressed in and Likelihood function under the condition. After calculation and derivation, formula (26) can be rewritten as:

[0108]

[0109] In formula (27), the present invention assumes that the noise variance σ 2is the only known channel information, and r (k,j) , m, and n0 can be obtained through measurement, so the test statistic G(r (k,j) ). Therefore, the switching criteria of the beam search codebook are as follows:

[0110]

[0111] After each cyclic scan at the receiving end, a channel signal-to-noise ratio test is performed. (k,j) ) satisfies formula (28), the beam search codebook is switched and the narrow beam search codebook is used for beam search. If the detection result does not satisfy formula (28), the hierarchical beam search codebook is continued to be used for beam search.

[0112] When determining whether to stop the layered search at the current level, the signal-to-noise ratio of the accumulated received signal is detected to determine whether it meets a preset signal-to-noise ratio target value.

[0113] From the output signal formula (24) of the matched filter, we can know that the output of the matched filter is r (k,j) The signal-to-noise ratio is After each cyclic scan of any beam in the hierarchical search, the matched filter output r is detected. (k,j) Does the signal-to-noise ratio meet Where ρ is the preset signal-to-noise ratio target value. As long as the detection result of any beam satisfies The hierarchical search stops the beam search at the current level and executes the beam search at the next level until the last level of the hierarchical beam search codebook is reached. Whether the conditions are met, the following assumptions can be made:

[0114]

[0115] because h in (k,j) is unknown, so a simple hypothesis test cannot be performed directly. Similarly, the generalized likelihood ratio test is used to test the above hypothesis. The test statistic can be written as:

[0116]

[0117] In formula (30), the present invention assumes that the noise variance σ 2 is the only known channel information, and r (k,j) It can be obtained by measurement, so the test statistic can be calculated directly at the receiving end Get the beam search stop condition of the kth layer:

[0118]

[0119] Where γ' is the set likelihood threshold value. If formula (31) is satisfied, the beam search at the current level (kth level) is stopped; otherwise, the beam search at the current level is continued.

[0120] Step 4: Layered Search When starting a new layer of search, select the angle range to be searched.

[0121] When the kth layer of the hierarchical search meets the beam search stop condition, it is necessary to determine the beam search angle interval CV of the k+1th layer k+1 ,

[0122] Traditional hierarchical search angle interval division method such as Figure 3 As shown, there are only two results for the beam search angle interval of the k+1th layer: or This angle division method may have certain problems in practical applications: when the main path angle of the channel is located at the junction of two angle intervals, the error probability of the layered search will be higher.

[0123] Therefore, the present invention proposes a beam search angle interval division method, such as Figure 4 As shown in the figure, the angle search interval of each layer is divided into three intervals. The core of this method is that when the main path angle of the channel is and When the intersection is near, select the angle interval centered on the intersection As the k+1th layer beam search angle interval. Therefore, after each layer of beam search stops, the layered search needs to determine which of the three interval centers the main path direction of the channel is closer to, thereby determining the search interval of the next layer.

[0124] In order to solve this problem, this embodiment derives the basis for angle interval judgment based on the single-path model. Assume that the millimeter wave single-path channel can be expressed as where u(θ) and represents the steering vector at the receiving and transmitting ends, α represents the unknown path gain, and θ is the unknown channel main path angle.

[0125] make and Therefore, after the receiving end has scanned m times, the matched filter output signal is:

[0126]

[0127] To derive the channel path angle θ, maximum likelihood detection is used to estimate an angle closest to θ.

[0128]

[0129] where Φ k represents the value range of θ, p(r (k,i) ,r (k,i+1) |τ,θ) is the likelihood function. It can be expressed as:

[0130]

[0131] Let J(θ,τ)=|r (k,i) -τg (k,i) (θ)| 2 +|r (k,i+1) -τg (k,i+1) (θ)| 2 , the above problem (33) can be equivalent to:

[0132]

[0133] The minimum value of τ can be obtained by derivation:

[0134]

[0135] where g * (k,i) (θ) represents g (k,i) The complex conjugate of (θ),

[0136] represents the beamforming gain at the receiving end, Substituting into formula (35), we can get that problem (35) is equivalent to:

[0137]

[0138] The present invention defines the value space of θ as Φ k Constructed into a set of three elements:

[0139]

[0140] where θ (k,i) and θ (k,i+1) Respectively represent the angle interval and The center angle,

[0141] represents the angle at the intersection of two angle intervals. The angle interval to be searched for the k+1<Kth layer is obtained as follows:

[0142]

[0143] when Just choose As the beam search interval of the k+1th layer, otherwise is the center of the interval, the width and The equal angle interval is used as the beam search interval of the k+1th layer.

[0144] Step 5: When the conditions for switching from the hierarchical beam search codebook to the narrow beam search codebook are met, the narrow beam search codebook is used to perform beam search. The specific steps are as follows:

[0145] 5.1 When using the narrow beam search codebook for beam search, first initialize the active beam set And start to perform cyclic beam scanning.

[0146] 5.2 Perform a cyclic scan on all beams in the active beam set.

[0147] Assume that n0 pilot signals are sent each time. After completing the mth cycle scan, beam w i Signal received from the first scan It can be expressed as:

[0148] y i =h i s+z i (40)

[0149] in, is the equivalent channel between the sender and the receiver, is the pilot signal, and Where P T The antenna transmits power, The variance is σ 2 Independent circularly symmetric Gaussian variables. i After the corresponding beam is scanned m times, the accumulated pilot signal is mn0, and we can get:

[0150]

[0151] in, is the signal after matched filtering, T i The degree of freedom is 2 and the noncentrality parameter is Noncentral chi-square distribution of .

[0152] 5.3 After each cyclic scanning of the active beams, the active beam set is updated based on the posterior probability of the pairwise comparison of all beam intensities.

[0153] The specific rules are: when f(T i ,T j )>Γ, beam i is considered stronger than beam j, and beam j is removed from the active beam set; where f(T i,T j ) indicates that when the accumulated pilot signals corresponding to beam i and beam j are T i and T j When (see formula 41), the posterior probability that the strength of the equivalent channel of beam i is higher than that of beam j is f(T i ,T j )=P r {|h i |>|h j |}, Γ represents the posterior probability threshold. The above rules are applied to the pairwise comparison of all beams in the active beam set. Therefore, after each cyclic scan, the updated active beam set can be expressed as:

[0154]

[0155] Among them, the posterior probability function f(x,y) can be expressed as follows:

[0156]

[0157] Where L l (·) is a Laguerre polynomial of order l. The posterior probability can be approximately calculated by summing a finite number of terms l according to the analytical expression of formula (13); it can also be approximately calculated using the Monte Carlo method, and the calculation result is stored in a lookup table and called in real time.

[0158] 5.4 After each update of the active beam set, determine whether the following three stop conditions are met. If any of the conditions is met, the narrow beam search stops.

[0159] 1. There is only one beam codeword left in the active beam set:

[0160]

[0161] 2. There are two adjacent codewords left in the active beam set, and the following conditions are met:

[0162]

[0163] 3. Limitation on the number of pilot signals:

[0164]

[0165] Where N max It is the maximum number of pilot signals allowed in the entire beam search process.

[0166] Step 6: When all beam searches stop, select the final beamforming direction.

[0167] 6.1 When the layered beam search codebook is used to search the last layer K, there is no need to determine the next layer search interval, but the direction used for beamforming needs to be determined. is the direction of beamforming. The corresponding codeword w in the hierarchical beam search codebook can be directly used (K,j) Beamforming is performed when When there is no directly corresponding codeword in the hierarchical beam search codebook, beamforming can be achieved. It is necessary to rotate adjacent codewords by a certain angle so that they can achieve beamforming pointing. Let w opt represents the codeword for the final beamforming. The above rules can be summarized as:

[0168]

[0169] Among them, w (K,i) represents any codeword in the last layer Kth layer beam search codebook, ⊙ represents the Hadamard product, d is the antenna spacing, and λ is the signal wavelength.

[0170] 6.2 Use the narrow beam search codebook to perform beam search. After the stop condition is met, the final beamforming direction needs to be determined.

[0171] When beam search is stopped by condition 1 or condition 3, select w i* The beam codeword finally selected is:

[0172] w opt =w i* (48)

[0173] Using w i* Just perform beamforming.

[0174] When condition 2 is used as the stop condition, w can be i* Move half the beam width to get a beam codeword that is better than the two codewords in the active beam set. The method to get the best beam codeword is as follows:

[0175]

[0176] ⊙ represents the Hadamard product, d represents the antenna spacing, and λ represents the signal wavelength. opt Just perform beamforming.

[0177] Simulation results and analysis:

[0178] Example simulation parameter settings of the adaptive millimeter wave beam search method: the transmitter is equipped with a single antenna, (N T=1), using omnidirectional transmission to transmit pilot signals, w T = 1. The receiving end is equipped with 32 uniform linear antenna arrays (N R =32), the beam search codebook used is the DFT hierarchical beam search codebook with the number of layers K=log232=5. All codewords in the last layer of the hierarchical beam search codebook are regarded as narrow beam search codebooks. Set the channel signal-to-noise ratio target value β=0, the hypothesis test threshold γ=4, and the number of pilot signals n0=1 sent by a single beam scan at one time. When using the hierarchical beam search codebook for beam search, set the signal-to-noise ratio target value ρ=4 of the received signal and the hypothesis test threshold γ'=2; when using the narrow beam search codebook for beam search, the posterior probability threshold is set to Γ=0.95. The maximum number of pilot signals allowed in the entire beam search process is N max =2000.

[0179] Figure 5 , Figure 6 A comparison of the spectrum efficiency and beam search time of the adaptive beam search algorithm proposed in the present invention and three existing methods under the LOS channel is given, including an adaptive beam hierarchical search algorithm (AHS, this algorithm can be found in the following patent: Liu Chunshan, Li Song, Zhao Lou. An adaptive millimeter-wave beam hierarchical search method [P]. Zhejiang Province: CN113225116B, 2022-05-31.), an adaptive narrow beam search algorithm (IDBS, this algorithm can be found in the following paper: C. Liu, M. Li, L. Zhao, P. Whiting, SV Hanly and IB Collings,"Millimeter-Wave Beam Search With Iterative Deactivation and Beam Shifting," in IEEE Transactions on Wireless Communications, vol. 19, no. 8, pp. 5117-5131, Aug. 2020, doi: 10.1109 / TWC.2020.2989343.) and a traditional hierarchical search algorithm.

[0180] from Figure 5 and Figure 6 It can be seen that under LOS channel, the adaptive beam search algorithm proposed by the present invention can achieve good spectrum efficiency under all channel signal-to-noise ratio conditions, and can approach or even reach the optimal spectrum efficiency. In terms of search time, it can adaptively adjust the beam search time according to the change of channel signal-to-noise ratio, and has good adaptability, which is specifically manifested as follows: under low channel signal-to-noise ratio conditions, the beam search time is relatively long; under high channel signal-to-noise ratio conditions, the beam search time becomes shorter.

[0181] from Figure 5 and Figure 6 It can be seen that under the LOS channel, the adaptive beam search algorithm proposed in the present invention is compared with the beam search algorithm IDBS that only uses a narrow beam search codebook. Under the condition of low channel signal-to-noise ratio, the spectrum efficiency is similar and the beam search time is the same. Under the condition of high channel signal-to-noise ratio, the spectrum efficiency is similar, but the beam search time is much shorter than that of the IDBS algorithm, thereby reducing resource consumption.

[0182] It can be seen that under LOS channel, the adaptive beam search algorithm proposed in the present invention is compared with the beam search algorithm AHS which only uses the hierarchical beam search codebook. Under the condition of high channel signal-to-noise ratio, the spectrum efficiency is slightly higher than that of the AHS algorithm, and the beam search time is almost the same. Under the condition of low channel signal-to-noise ratio, the spectrum efficiency is also slightly higher than that of the AHS algorithm, but the beam search time is much shorter than that of the AHS algorithm, which greatly reduces the consumption of resources. Compared with the traditional hierarchical search algorithm, the adaptive beam search algorithm proposed in the present invention is far superior to the traditional hierarchical search algorithm in terms of both spectrum efficiency and beam search time.

[0183] In summary, the adaptive millimeter wave beam search method proposed in the present invention uses two codebooks, a multi-precision layered beam search codebook and a narrow beam search codebook, for beam search, and can switch the beam search codebook in real time according to the change of the channel signal-to-noise ratio, which can not only ensure the effect of the beam search, but also reduce the time of the beam search, and has good adaptability. Compared with the beam search method using a single beam search codebook, the method proposed in the present invention can be applied to more diverse communication scenarios. Whether it is a low channel signal-to-noise ratio or a high channel signal-to-noise ratio scenario, a most suitable beam search codebook can be used to perform beam search to achieve a satisfactory beam search effect.

Claims

1. An adaptive millimeter wave beam search method, characterized in that The following steps are involved: Step 1: Determine a beam search scenario, a preset multi-precision hierarchical beam search codebook, and a narrow beam search codebook; Step 2: The overall process of performing adaptive beam search using a hierarchical beam search codebook and a narrow beam search codebook; Step 3: When using the hierarchical beam search codebook for beam search, a generalized likelihood ratio test is used to determine whether to switch the beam search codebook and whether to stop the hierarchical search at the current level and enter the next level; Step 4: When starting a new layer of search, select the angle interval to be searched; Step 5: When the condition for switching from the hierarchical beam search codebook to the narrow beam search codebook is met, the narrow beam search codebook is used to perform beam search; Step 6: When all beam searches are stopped, select the final beamforming direction; The beam search scenario described in step 1 includes a transmitter and a receiver, wherein the transmitter continuously transmits a pilot signal to the entire angle space covered by it, and the receiver uses a preset beam search codebook to perform beam scanning to find the best beamforming direction; The beam search codebook in step 1 is divided into two types: one is the hierarchical beam search codebook W, and the other is the narrow beam search codebook The hierarchical beam search codebook described in step 1 has a tree structure, that is, the coverage area of ​​a wider beam in the upper layer is jointly covered by multiple narrower beams in the next layer; the hierarchical beam search codebook used has a total of K layers, of which the kth layer has a total of 2 k codewords, the codeword set of the kth layer is represented by W k ,and represents the i-th codeword in the k-th layer, N R =2 K Indicates the number of antennas equipped by the receiver; codeword w (k,i) The angle range covered is Indicates that It is the result after taking the cosine of the coverage angle; in the layered beam search codebook, the union of the coverage ranges of all codewords in any layer completely covers the entire antenna array angle domain; and the coverage range of the codewords in any layer is completely represented by the union of the coverage ranges of the two codewords in the corresponding next layer; The narrow beam search codebook described in step 1 is single-layer, and the entire angular space is covered by the narrow beams; Use the hierarchical beam search codebook to find all codewords W in the last layer K As a narrow beam search codebook, it is expressed as The beam search described in step 2 performs beam scanning starting from the hierarchical beam search codebook; and the overall beam search process is specifically implemented as follows: When the hierarchical beam search codebook is used to perform beam scanning, the receiver determines whether the conditions for switching to the narrow beam search codebook for beam search are met after completing a cycle of pilot signal acquisition each time; if the conditions for switching to the narrow beam search codebook are met, the hierarchical beam search codebook is stopped for beam search, and the narrow beam search codebook is used for beam search instead until the beam search stops; if the conditions for switching to the narrow beam search codebook are not met and the current hierarchical search level is not the last level, it is determined whether to enter the next level of hierarchical search; if the conditions for entering the next level of hierarchical search are met, the next level of beam search codebook is used to perform beam search; if the conditions for switching to the narrow beam search codebook are not met and the current hierarchical search level is the last level, it is determined whether the stop conditions for beam search are met; if none of the above conditions are met, the hierarchical search of the current level continues; The beam search process selects a final beamforming direction when the beam search stop condition is met; The hierarchical search receiver completes a cycle of pilot signal acquisition, which means that the receiver uses all the candidate beams of the current layer in turn to collect the pilot signal transmitted by the receiver once; The switching from the hierarchical beam search codebook to the narrow beam search codebook is determined by detecting the signal-to-noise ratio of the unknown channel; When performing beam search using the hierarchical beam search codebook as described in step 3, the transmitter uses a beam covering the angle interval of interest. Transmit signal, N T Indicates the number of antennas equipped by the transmitter; the receiver uses a cyclic scanning method to collect the pilot signal transmitted by the transmitter. In each cyclic scanning, the receiver uses the codeword w (k,i) and w (k,i+1) Perform beamforming and receive the pilot signal sent by the transmitter; suppose the time for a single beam to collect signals in each cyclic scan of the receiver is n0 ≥ 1 symbol period, then after m cyclic scans, the receiver uses the jth codeword w of the kth layer, j∈{i,i+1} (k,j) The signal received by the corresponding beam is expressed as: y (k,j) =h (k,j) s (k,j) +z (k,j) (1) in represents the equivalent channel between the transmitter and the receiver, represents the channel matrix between the transmitter and the receiver, represents the pilot signal sent and P T Indicates the transmit power, represents circularly symmetric Gaussian noise; the signal after matched filtering is expressed as: Among them, P T represents the transmit power of the antenna, represents circularly symmetric Gaussian noise, and the signal-to-noise ratio of the channel is In step 3, in order to detect the current channel signal-to-noise ratio, a channel signal-to-noise ratio target value β is set in advance, and it is set that when the channel signal-to-noise ratio is greater than the target value, it is considered to be a high channel signal-to-noise ratio, otherwise it is a low channel signal-to-noise ratio; the above setting can be written as the following assumption: However, η in the formula (k,j) It cannot be directly calculated from the received signal, so the above hypothesis cannot be directly tested; using the generalized likelihood ratio test to test the hypothesis, the test statistic can be obtained: Where γ∈(0,∞) represents the threshold value of hypothesis testing, and Respectively expressed in and Likelihood function under the condition; after calculation and derivation, formula (4) is rewritten as: In formula (5), it is assumed that the noise variance σ 2 is the only known channel information, and r (k,j) , m, and n0 can be obtained through measurement, so the test statistic G(r (k,j) ); Therefore, the switching criteria of the beam search codebook are as follows: After each cyclic scan of the receiver, a channel signal-to-noise ratio test is performed; if the test statistic G(r (k,j) ) satisfies the condition of formula (6), the beam search codebook is switched and the narrow beam search codebook is used for beam search; if the detection result does not satisfy the condition of formula (6), the hierarchical beam search codebook is continued to be used for beam search; Step 3: When performing hierarchical search, a generalized likelihood ratio test is used to detect whether the signal-to-noise ratio of the cumulative received signal meets the preset target value. When the current level is the kth level, the output signal formula (2) of the matched filter shows that the signal-to-noise ratio of the received signal is Let the preset target value of the signal-to-noise ratio of the cumulative received signal be ρ, then according to the generalized likelihood ratio test, the test statistic that can be obtained is: After each cyclic beam scan, if the following conditions are met: Then stop the beam search of the current k-th level, otherwise continue the beam search of the current level, where γ' is the set likelihood threshold; When the hierarchical search enters a new level of search, if k=1, that is, when the hierarchical search is initially entered, the angle interval to be searched is determined to be The beam to be searched is w (1,1) and w (1,2) The corresponding beam; When the hierarchical search enters the k+1th layer and k≥1, the matched filter signal r accumulated by the receiver when the kth layer beam search stops is calculated. (k,i) and r (k,i+1) , determine the beam search angle interval of the k+1th layer; specifically: there are three possibilities for the beam search angle interval of the k+1th layer to be determined, and the specific rules for determining it are as follows: when the estimated value of the channel main path angle When selecting CV w(k,i) As the beam search angle interval of the k+1th layer; when the estimated value of the channel main path angle When selecting CV w(k,i+1) As the beam search angle interval of the k+1th layer; when the estimated value of the channel main path angle When selecting CV w(k,*) As the beam search angle interval of the k+1th layer; where θ (k,i) and θ (k,i+1) Respectively represent the angle interval and The central angle of the channel, the estimated value of the main path angle From formula (9), we get: in, represents the value range of θ, express and The angle of the junction; in represents the steering vector at the receiver, g * (k,i) (θ) represents g (k,i) The complex conjugate of (θ), represents the beamforming gain of the receiver; Therefore, the angle interval to be searched at the k+1th layer is summarized as: When select as the beam search angle interval for the (k + 1)-th layer. All the beams corresponding to the codewords within the interval are the beams to be searched for the (k + 1)-th layer; otherwise, use as the interval center, and the angle interval with the same width as is used as the beam search angle interval for the (k + 1)-th layer. All the beams corresponding to the codewords within the interval are the beams to be searched for the (k + 1)-th layer; Step 5: When using the narrow beam search codebook for beam search, first initialize the active beam set. And start to perform cyclic beam scanning; Step 5 cyclically scans all beams in the active beam set, sending n0 pilot signals each time; after completing the mth cyclic scan, beam w i Signal received from the first scan It can be expressed as: y i =h i s+z i (11) in, is the equivalent channel between the transmitter and the receiver, is the pilot signal, and Where P T is the antenna transmission power, The variance is σ 2 Independent circularly symmetric Gaussian variables; w i After the corresponding beam is scanned m times, the accumulated pilot signal is mn0, and we can get: in, is the signal after matched filtering, T i The degree of freedom is 2 and the noncentrality parameter is Non-central chi-square distribution of Step 5 After each cycle of scanning the active beams, the active beam set is updated according to the posterior probability of the comparison of all beam intensities. The specific rule is: when f(T i ,T j )>Γ, beam i is considered stronger than beam j, and beam j is removed from the active beam set; where f(T i ,T j ) indicates that when the accumulated pilot signals corresponding to beam i and beam j are T i and T j The posterior probability that the strength of the equivalent channel of beam i is higher than that of beam j is f(T i ,T j )=P r {|h i |>|h j |}, Γ represents the posterior probability threshold; the above rule is applied to the pairwise comparison of all beams in the active beam set. Therefore, after each cyclic scan, the updated active beam set can be expressed as: Among them, the posterior probability function f(x,y) can be expressed as follows: Where L l (·) is a Laguerre polynomial of order l; the posterior probability is approximated by summing a finite number of terms l according to the analytical expression of formula (14); Step 5: After each update of the active beam set, determine whether the following three stop conditions are met; if any of the conditions is met, the narrow beam search stops; Stop condition 1. There is only one beam codeword left in the active beam set: Stop condition 2. There are two adjacent codewords left in the active beam set and the following conditions are met: Stop condition 3. Limitation of the number of pilot signals: Where N max It is the maximum number of pilot signals allowed in the entire beam search process; Step 6: When the last layer K is searched using the layered beam search codebook, the beamforming direction is as follows: Among them, w (K,i) represents any codeword in the last layer Kth layer beam search codebook, ⊙ represents Hadamard, express and w (K,i) The difference in the main lobe center direction angle, d represents the antenna spacing, and λ represents the signal wavelength; Step 6: When the narrow beam search codebook is used for beam search, when stop condition 1 or stop condition 3 is met, the beamforming direction is obtained as follows: In opt =in i* (19) in Indicates the last remaining active beam set The codeword corresponding to the strongest beam in ; When stop condition 2 is met, the beamforming direction is obtained as: ⊙ represents the Hadamard product, express and w (K,i) The difference in the main lobe center direction angle, d represents the antenna spacing, and λ represents the signal wavelength.

Images