Perception-based Communication Beam Tracking Method

Abstract

The present invention discloses a perception-based communication beam tracking method, which includes the following steps: at the transmitting base station side, design the beamforming vector for the next moment based on the position of the unmanned aerial vehicle (UAV) fed back by the data processing center; on the one hand, the beam transmitted by the base station is received by the UAV to achieve the transmission of communication information, and on the other hand, the beam is reflected by the UAV to the distributed receiving base station, and the receiving base station sends the collected echo signals to the data processing center; the data processing center models the state transition model and the observation model of the UAV, takes the collected complex echo signals as observation data, applies the unscented Kalman filtering algorithm to estimate and predict the position information of the UAV at the next moment, and finally feeds the predicted position information back to the transmitting base station. The present invention reduces the pilot overhead, improves the communication rate of useful information between the base station and the UAV, and at the same time improves the tracking performance of the UAV.

Description

Technical Field

[0001] The present invention belongs to the field of communication engineering, and particularly relates to a communication beam tracking method based on sensing, and more particularly to a communication beam tracking method under a large-scale MIMO-OFDM (Multi Input Multi Output-Orthogonal Frequency Division Multiplexing) system based on sensing. Background Art

[0002] Consider a single base station-single unmanned aerial vehicle (UAV) communication system. In order to ensure a stable communication link with the UAV, the base station needs to obtain the channel information between it and the UAV. In traditional solutions, it is generally achieved through beam training and beam tracking.

[0003] The process of the traditional feedback-based beam tracking scheme is as follows: First, the base station performs beam training, that is, cyclically selects all possible beam directions from a preset codebook for transmission, and takes the signal direction with the strongest path gain as the direction of the UAV. Then, the base station designs a corresponding beamforming vector based on this direction and sends an OFDM symbol block to the UAV to complete the first communication. Subsequently, in order to determine the channel information between the base station and the UAV at the next moment, the base station sends a pilot signal to the UAV. After the UAV completes channel estimation, it feeds back the channel information to the base station through the uplink. The base station redesigns a new beamforming vector and starts the second communication with the UAV. The subsequent tracking process follows the same pattern.

[0004] In the above tracking scheme, the most important step is to transmit pilot symbols for channel estimation. According to different channel conditions, the arrangement of pilot symbols is also different. In a communication scenario where the delay spread is greater than the length of the cyclic prefix (CP), block pilots are generally used. Block pilots periodically use all subcarriers of the entire OFDM symbol as pilots. To ensure the effect of channel estimation, the pilot symbol period S t needs to be kept consistent with the coherence time, that is (f Doppler is the Doppler frequency).

[0005] Analyzing the above feedback-based beam tracking scheme, it can be seen that before each beam transmission, the transmitting base station needs to send a pilot signal to the UAV. After the UAV completes channel estimation, it feeds back to the transmitting base station. This has two problems. First, since the period of the transmitted signal is very short during the tracking process, frequently sending pilots without carrying communication information is a large overhead, which will reduce the transmission rate of useful information. Second, during the tracking process, only a small number of OFDM symbols are sent as pilots each time. When the receiving base station performs a matched filtering operation after receiving the pilot signal, the matched filtering gain will be very low, thus affecting the performance of channel estimation. Summary of the Invention

[0006] Object of the Invention: Aiming at the defects existing in the prior art, the present invention proposes a communication beam tracking method under a perception-based large-scale MIMO-OFDM system, which can reduce the pilot overhead, improve the communication rate of useful information in the system, and at the same time improve the tracking performance of the transmitting base station for the UAV.

[0007] Technical Solution: The communication beam tracking method and performance analysis of a perception-based large-scale MIMO-OFDM system include the following steps:

[0008] (1) At the integrated transmitting base station, receive the predicted position information of the UAV fed back from the data processing center, design a beamforming vector for the UAV according to the predicted position information, and transmit the corresponding beam to complete communication with the UAV;

[0009] (2) The beam transmitted by the transmitting base station is reflected by the surface of the UAV and reaches the distributed receiving base station. The distributed receiving base station sends the observed echo signal to the data processing center for unified processing;

[0010] (3) The data processing center first models the state transition model and observation model of the UAV, and performs real-number conversion and straightening processing on the echo signal. Next, according to the UKF (Unscented Kalman Filter) algorithm and the processed echo data, it tracks the position information of the UAV at the current moment and predicts the position information of the UAV at the next moment, and feeds the predicted result back to the transmitting base station;

[0011] (4) Repeat steps (1), (2), and (3). While the transmitting base station tracks the trajectory of the UAV, it completes communication with the UAV.

[0012] Further, step (1) includes the following steps:

[0013] The data processing center estimates the position information of the UAV at time t-1 Predict the position of the UAV at time t And according to and and the position of the transmitting base station, the predicted launch angle of the UAV at time t is calculated The corresponding beamforming vector at time t can be expressed as where k represents the serial number of the UAV, t represents different estimation times, represents the estimated position of the k-th UAV at time t-1, represents the predicted position of the k-th UAV at time t, and respectively represent the abscissa and ordinate of the predicted position of the UAV at time t, and the abscissa and ordinate of the predicted speed, represents the predicted launch angle of the UAV at time t.

[0014] Furthermore, the step (2) includes the following steps:

[0015] After the beam designed in step (1) is transmitted, it is transmitted to each receiving base station in the distributed receiving base station through the surface of the UAV. Each receiving base station is independent. For the r-th receiving base station, the expression of the received echo signal after matched filtering where K represents the number of UAVs in the monitoring area of the integrated communication and sensing system, Q represents the number of OFDM symbols included in each OFDM symbol block transmitted by the transmitter, M t represents the number of transmitting antennas of the transmitting base station, M r represents the number of receiving antennas of the receiving base station, f r,k represents the Doppler frequency shift of the k-th UAV relative to the transmitting base station and the r-th receiving base station, τ r,k represents the time delay for the transmitted signal to reach the r-th receiving base station through the k-th UAV, τ r,k =(||p t -p k || 2 +||p r -p k || 2 ) / c, represents the attenuation coefficient from the transmitting base station through the k-th UAV to the r-th receiving base station, θ r,k represents the receiving angle of the r-th receiving base station for the k-th UAV, β k represents the transmitting angle of the transmitting base station for the k-th UAV, represents the normalized transmitting steering vector for the k-th UAV, denotes the normalized received steering vector for the k-th UAV and the r-th receiving base station; where, ||x|| 2 denotes the 2-norm of vector x.

[0016] Furthermore, the step (3) includes the following steps:

[0017] 1) Determine the system state model and system observation model of the UAV:

[0018] The data processing center represents the state of the UAV as two-dimensional position coordinates (p k,x [t], p k,y [t]) and velocity coordinates (v k,x [t], v k,y [t]). A constant velocity model is adopted for the UAV. The observation model is based on the echo feedback from the receiving base station group, considering the UAV time delay τ r,k , Doppler frequency shift f r,k factor to model the echo model;

[0019] 2) Initialize the state information and estimated variance matrix of the UAV:

[0020] Before starting to track the UAV, use an estimation algorithm (such as MUSIC (Multiple Signal Classification), OMP (Orthogonal Matching Pursuit)) to preliminarily estimate the position of the UAV, and the estimation result is s k,0 , and the data processing center initializes accordingly: In addition, the estimated variance matrix needs to satisfy the condition of a positive definite matrix: M k [-1|-1] = M k,0 , M k,0 is a positive definite matrix; where and M k [-1|-1] represent the initialized position and variance matrix respectively;

[0021] 3) One-step prediction of the UAV state and estimated variance matrix:

[0022] Perform UT (Unscented Transformation) on to obtain the first set of Sigma sampling points and calculate the weights of each sampling point according to the sampling strategy; finally, combine all sampling points into a one-step prediction state s k [t|t - 1], and calculate the prediction variance matrix M k [t|t - 1];

[0023] 4) One-step prediction of the observed echo:

[0024] According to s in 3) k [t|t - 1] and M k [t|t - 1], perform the UT transformation to generate the second set of Sigma sampling points, and combine the second set of Sigma sampling points into the one-step prediction y of the observed echo through weights k [t|t - 1], and calculate the prediction variance matrix M of the observed echo accordingly k,zz [t - 1|t] and the prediction covariance matrix M between the state and the observed echo k,sz [t - 1|t];

[0025] 5) Update the UAV state and the estimated variance matrix:

[0026] According to the one-step prediction results s in 3) and 4) k [t|t - 1], M k [t|t - 1], y k [t|t - 1], M k,zz [t - 1|t] and M k,sz [t - 1|t], calculate the Kalman gain K k [t], update the UAV state information at the current moment and calculate the estimated variance matrix M k [t|t].

[0027] Furthermore, the step (4) includes the following steps:

[0028] During the tracking process, the transmitting base station will transmit beams according to the predicted state information of the UAV to complete communication with the UAV. For K UAVs, the sum of the communication rates between them and the base station can be expressed as where SNR k represents the signal-to-noise ratio of the signal received by the k-th UAV.

[0029] Beneficial effects: The present invention is applicable to the communication beam tracking scenario in the 5G large-scale MIMO-OFDM system scenario. Compared with the traditional feedback-based beam tracking scheme, it avoids the overhead of transmitting pilots before each tracking, uses more communication resources to send useful information, and improves the communication rate of useful information; it takes the entire OFDM symbol block as the observed signal to participate in the tracking process, improves the matched filtering gain of the received signal, and enhances the tracking performance of the UAV. Compared with the traditional tracking algorithm based on EKF (Extended Kalman Filter), applying the UKF algorithm can avoid calculating the computationally complex and time-consuming Jacobian matrix, has better real-time performance, and its performance is superior to the EKF algorithm. Description of the Drawings

[0030] Figure 1 is the integrated communication and sensing system model diagram for real-time communication and tracking of the present invention;

[0031] Figure 2 is the schematic diagram of beam tracking based on communication feedback of the present invention;

[0032] Figure 3 is the block pilot arrangement diagram of the present invention;

[0033] Figure 4 is the schematic diagram of beam tracking based on integrated signals of the present invention;

[0034] Figure 5 is the comparison diagram of tracking performance based on UKF and feedback of the present invention;

[0035] Figure 6 is the comparison diagram of communication performance based on UKF and feedback of the present invention. Specific Embodiments

[0036] The present invention will be further clarified below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0037] The present invention proposes a communication beam tracking method based on sensing, which introduces radar into traditional pure communication-based beam tracking and applies the UKF algorithm, not only reducing the pilot overhead, improving the transmission rate of useful information, but also enhancing the sensing accuracy and tracking performance. The specific solution proposed by the present invention is as follows:

[0038] 1: System Model:

[0039] Consider the integrated communication and sensing system model for real-time communication and tracking as shown in Figure 1 Set there is 1 transmitting base station and R distributed receiving base stations in the system. Each transmitting base station is equipped with M t antennas, and each receiving base station is equipped with M r antennas. Assume there are K drones, and each drone is configured with a single antenna. Here, the MIMO-OFDM system adopts frequency-division multiple access, and each subcarrier independently designs the beamforming vector.

[0040] (1) Communication Signal Model

[0041] Assume that the transmitted communication baseband signal contains Q OFDM symbols, each OFDM symbol contains L subcarriers, and w k,l represents the beamforming vector of the l-th subcarrier for the k-th drone, and δk,l denotes the subcarrier selection factor, i.e., whether the \(l\)-th subcarrier is allocated to the \(k\)-th UAV, and \(\Delta f\) represents the interval between subcarriers. denotes the communication information sent to the \(k\)-th UAV on the \(l\)-th subcarrier of the \(q\)-th OFDM symbol, \(T\) o denotes the symbol length of the OFDM symbol, and \(T\) o = \(T + T\) cp , \(T\) represents the effective length, \(T\) cp denotes the CP length. Then, the baseband signal of the \(q\)-th OFDM symbol can be written as where, when \(t\in[0, T\) o , \(\xi(t)=1\), otherwise \(\xi(t)=0\). Therefore, the transmitted signal in the entire OFDM symbol block can be expressed as Before entering the channel, an up-conversion operation is also required, and the signal after up-conversion can be expressed as where \(f\) c represents the carrier frequency, and \(Re(x)\) represents the operation of taking the real part of \(x\). Assume that the state information of the \(k\)-th UAV can be expressed as \(s\) k [t]=[p k,x [t], p k,y [t], v k,x [t], v k,y [t]] T , where \(p\) k,x [t] and \(p\) k,y [t] represent the horizontal and vertical coordinates of the \(k\)-th UAV respectively, and \(v\) k,x [t] and \(v\) k,y [t] represent the horizontal and vertical velocities of the \(k\)-th UAV respectively. According to the previous assumption, the UAV adopts a constant velocity model. Then, the state transition model of the \(k\)-th UAV can be expressed as:

[0042]

[0043] where, \(u\) k,x [t] and \(u\) k,y [t] represent the perturbation noises of the horizontal and vertical velocities of the \(k\)-th UAV respectively, following a Gaussian distribution with zero mean and variance of . At time \(t\), the integrated communication and sensing system uses \(Q\) OFDM symbols to predict the position and velocity information of \(K\) targets at time \(t + 1\).

[0044] (2) Sensing-based beam tracking

[0045] As Figure 2 shown, for the feedback-based beam tracking method, the transmitting base station needs to re-transmit pilots for channel estimation before each communication. For a frequency-selective channel, generally select asFigure 3 The block pilots shown as the pilot arrangement method lead to serious pilot overhead due to frequent communication processes. As for Figure 4 the perception-based beam tracking method shown, the entire OFDM data block is used as the observation data to track the UAV, and pilots are no longer required. Under the assumption of narrowband signals, for the r-th receiving base station, the received echo signal after down-conversion can be expressed as where f r,k represents the Doppler frequency shift of the k-th UAV relative to the transmitting base station and the r-th receiving base station, and can be expressed as τ r,k represents the time delay for the transmitted signal to reach the r-th receiving base station via the k-th UAV, and can be expressed as represents the attenuation coefficient from the transmitting base station to the r-th receiving base station via the k-th UAV, represents the receiving angle of the r-th receiving base station with respect to the k-th UAV, represents the transmitting angle of the transmitting base station with respect to the k-th UAV, represents the normalized transmitting steering vector for the k-th UAV, represents the normalized receiving steering vector for the k-th UAV and the r-th receiving base station.

[0046] Assume where ( represents rounding x downwards), considering the case where the delay spread is greater than the CP length, the signal after matched filtering can be expressed as:

[0047]

[0048] where, p k,l represents the power of the l-th subcarrier sent to the k-th UAV. In digital beamforming, the shape, direction, and energy of the beam can all be designed at the transmitting base station through the beamforming vector. To obtain the maximum signal-to-noise ratio at the receiving base station, the transmitting base station adopts and by using a large-scale MIMO array at the transmitting base station, the steering vectors can be approximately orthogonal in different directions, that is, for β k ≠β k , so the echo signals of different UAVs can be isolated at the receiving base station. In addition, for the unknown variables in formula (2) It is estimated after predicting the position information of the k-th drone at the next moment in the data processing center. Therefore, for the k-th drone, a four-dimensional (i.e., receiving base station dimension, receiving antenna dimension, OFDM symbol dimension, subcarrier dimension) signal can be obtained in the data processing center. In order to be able to apply this four-dimensional complex signal as observed data in the UKF process, the signal also needs to be converted from a complex signal to a real signal. The real part and the imaginary part of each complex number in the four-dimensional complex signal are taken and straightened respectively. Therefore, a real signal vector can be obtained for the k-th drone where y′ k represents the useful signal, and η k represents the noise signal. Finally, the data processing center applies the y at time t k to the UKF algorithm process to predict the launch angle at time t + 1 design the corresponding beamforming vector and the transmitting base station completes beam transmission accordingly.

[0049] For K drones, the overall transmission rate R all can be expressed as: where L represents the number of subcarriers, represents the channel information between the transmitting base station and the k-th drone, and f′ k represents the Doppler frequency shift of the k-th drone, represents the variable to be estimated, and τ′ k represents the time delay of the k-th drone, represents the communication signal noise received by the k-th drone.

[0050] 2: UKF-based beam tracking:

[0051] (1) Model the state transition equation and the observation equation of the drone. According to the previous description, the state transition equation and the observation equation of the k-th drone can be abstracted respectively. The state transition equation can be expressed as s k [t] = Fs k [t - 1] + ω k [t], and the observation equation can be expressed as y k [t] = h(s k [t]) + c k [t], where s k [t] represents the state of the k-th drone at time t, s k [t - 1] represents the state of the k-th drone at time t - 1, and F can be expressed as:

[0052]

[0053] yk The specific form of each term in [t] can be written as:

[0054]

[0055] ω k [t] and c k [t] both represent Gaussian distributions with zero mean, and their variances can be expressed as and

[0056] (2) Initialize the UAV state variables and the estimated variance matrix. That is M k [-1|-1] = M k,0 . Note that M k needs to be used for the matrix decomposition operation in the subsequent one-step prediction, so it needs to meet the positive definite condition.

[0057] (3) Perform a one-step prediction on the UAV state. First, perform the UT transformation. Since the state variable is P-dimensional, 2P + 1 Sigma points need to be generated. For p = 0, the Sigma point For p = 1, 2,..., P, For p = P + 1,..., 2P, And is the vector obtained by transposing the p-th row after the estimation error matrix decomposition operation, which can be expressed as Substitute each Sigma point into the state transition equation, that is, the transformed Sigma point Then the one-step prediction of the UAV state variable can be expressed as the weighted sum of all variables, that is where ω (p) represents the weight of the p-th Sigma point, and its value is determined according to the sampling strategy of the Sigma sampling points. In addition, the estimated variance matrix of the one-step prediction can be expressed as is the noise variance in the state transition equation. Among them, λ represents the hyperparameter set by humans.

[0058] (4) Perform a one-step prediction on the observed echo at the receiving base station. Perform the UT transformation in the same way as in step (3). It should be noted that the result s k [t|t - 1] and M k[t|t - 1], and substitute the generated Sigma points into the observation model to obtain the one-step predicted value y of the observed echo k [t|t - 1], and calculate the predicted variance matrix M of the observed echo k,zz [t - 1|t] and the predicted covariance matrix M between the UAV state and the observed echo k,sz [t - 1|t], and the calculation method is similar to that in step (3).

[0059] (5) Update the UAV state and the estimated variance matrix. First, update the Kalman gain K k [t], which can be expressed as K k [t]=M k,sz [t|t - 1](M k,zz [t|t - 1]) -1 , and based on the echo data y k [t] collected by the data processing center at time t from the distributed receivers after being real - valued and straightened, update the state variable at time t and the estimated variance matrix M k [t|t]=M k [t|t - 1]-K k [t]M k,zz [t|t - 1](K k [t]) T .

[0060] As Figure 5 shown, the abscissa represents the movement time of the UAV, and the ordinate represents the root mean square error of the estimated UAV position (abscissa) under two tracking algorithms. It can be seen from the figure that the dotted line represents beam tracking based on feedback, and the error range is within 10 -2 -10 2 , and the solid line represents beam tracking based on UKF, and the error range is within 10 -4 -10 0 . By analyzing the results in the figure, except for the fluctuation period at the beginning of filtering, as time goes by, the UAV approaches the base station, the signal - to - noise ratio at the receiving end increases, and the estimation performance of the two tracking algorithms gets better. Then, as the UAV moves away from the base station, the signal - to - noise ratio at the receiving end decreases, and the estimation performance of the two tracking algorithms gets worse. However, during the process, the RMSE (Root Mean Square Error) of the tracking method based on UKF is less than that of the beam tracking method based on feedback. And after moving away from the base station, the performance of the tracking method based on feedback drops too fast, while the tracking method based on UKF drops slowly.

[0061] Figure 6The communication performance between the base station and the UAV is compared under two methods: UKF-based beam tracking and feedback-based beam tracking. The abscissa is the movement time of the UAV, and the ordinate is the communication rate. The solid line represents the communication rate under UKF, and the dotted line represents the communication rate under feedback. Analyzing the results in the figure, in the first 6 seconds, the communication rate based on feedback can basically keep up with the communication rate based on UKF. After 7 seconds, the communication rate based on feedback begins to oscillate and decline, and drops sharply at 8.9 seconds. This is because the tracking performance of the feedback-based method is too poor at 8.9 seconds, making it difficult for the base station to transmit a beam aligned with the UAV, resulting in a sharp drop in the communication rate. The method based on UKF has better tracking performance, does not show a sharp drop in the communication rate, and can achieve a higher peak rate during the entire tracking process, with better overall communication quality.

[0062] The simulation proves that the communication beam tracking method for the perception-based large-scale MIMO-OFDM system proposed in the present invention can improve the communication rate of useful information while reducing the pilot overhead, and has better tracking performance than the traditional feedback-based beam tracking method, verifying the correctness of the theory.

Claims

1. A perception-based communication beam tracking method, comprising the following steps: (1) At the integrated transmitting base station, receive the predicted position information of the unmanned aerial vehicle (UAV) fed back from the data processing center, design a beamforming vector for the UAV according to the predicted position information, and transmit a corresponding beam to complete communication with the UAV; (2) The beam transmitted by the transmitting base station is reflected by the surface of the UAV and then reaches the distributed receiving base station. The distributed receiving base station sends the observed echo signal to the data processing center for unified processing; (3) The data processing center first models the state transition model and the observation model of the UAV, and performs real-number conversion and straightening processing on the echo signal. Next, according to the UKF algorithm and the processed echo data, track the position information of the UAV at the current moment, predict the position information of the UAV at the next moment, and feed the predicted result back to the transmitting base station; (4) Repeat steps (1), (2), and (3). While the transmitting base station tracks the trajectory of the UAV, complete communication with the UAV. Step (1) includes the following steps: The data processing center estimates the position information of the UAV at time t-1 to predict the position of the UAV at time t And according to and as well as the position of the transmitting base station, calculate the predicted launch angle of the UAV at time t The corresponding beamforming vector at time t is expressed as where k represents the serial number of the UAV, t represents different estimation times, represents the estimated position of the k-th UAV at time t-1, represents the predicted position of the k-th UAV at time t, and respectively represent the abscissa and ordinate of the predicted position of the UAV at time t, and the abscissa and ordinate of the predicted speed, represents the predicted launch angle of the UAV at time t; The step (2) includes the following steps: After the designed beam is transmitted in step (1), it is transmitted to each receiving base station in the distributed receiving base station through the surface of the UAV. Each receiving base station is independent. For the r-th receiving base station, the expression of the received echo signal after matched filtering where K represents the number of UAVs in the monitoring area of the integrated communication and sensing system, Q represents the number of OFDM symbols included in each OFDM symbol block transmitted by the transmitter, M t represents the number of transmitting antennas of the transmitting base station, M r represents the number of receiving antennas of the receiving base station, f r,k represents the Doppler frequency shift of the k-th UAV relative to the transmitting base station and the r-th receiving base station, τ r,k represents the time delay for the transmitted signal to reach the r-th receiving base station through the k-th UAV, τ r,k =(||p t -p k || 2 +||p r -p k || 2 ) / c, represents the attenuation coefficient from the transmitting base station to the r-th receiving base station through the k-th UAV, θ r,k represents the receiving angle of the r-th receiving base station with respect to the k-th UAV, β k represents the transmitting angle of the transmitting base station with respect to the k-th UAV, represents the normalized transmitting steering vector for the k-th UAV, represents the normalized receiving steering vector for the k-th UAV and the r-th receiving base station; where, ||x|| 2 represents the 2-norm of the vector x; The step (3) includes the following steps: 1) Determine the system state model and the system observation model of the UAV; The data processing center represents the state of the UAV as two-dimensional position coordinates (p k,x [t], p k,y [t]) and velocity coordinates (v k,x [t], v k,y [t]). A constant velocity model is adopted for the UAV, and the observation model is based on the echoes fed back from the receiving base station group, considering the UAV time delay τ r,k , Doppler frequency shift f r,k factors to model the echo model; 2) Initialize the state information and the estimated variance matrix of the UAV; Before starting to track the UAV, use an estimation algorithm to make a preliminary estimate of the UAV's position, and the estimation result is s k,0 , and the data processing center initializes accordingly: In addition, the estimated variance matrix needs to satisfy the condition of a positive definite matrix: M k [-1|-1] = M k,0 , M k,0 is a positive definite matrix; where and M k [-1|-1] represent the initialized position and variance matrix respectively; 3) Perform a one-step prediction of the UAV state and the estimated variance matrix; Pair Perform unscented transformation to obtain the first set of Sigma sampling points And calculate the weights of each sampling point according to the sampling strategy; finally, combine all sampling points into the one-step prediction state s k [t|t-1], and calculate the prediction variance matrix M k [t|t-1]; 4) Perform a one-step prediction of the observed echo; According to s in 3) k [t|t - 1] and M k Perform unscented transformation on [t|t - 1] to generate a second set of Sigma sampling points, and combine the second set of Sigma sampling points into a one-step prediction y of the observed echo through weights k [t|t - 1], and calculate the predicted variance matrix M of the observed echo based on this k,zz [t - 1|t] and the predicted covariance matrix M between the state and the observed echo k,sz [t - 1|t]; 5) Update the UAV state and the estimated variance matrix; According to the one-step prediction results s in 3) and 4) k [t|t - 1], M k [t|t - 1], y k [t|t - 1], M k,zz [t - 1|t] and M k,sz [t - 1|t], calculate the Kalman gain K k [t], update the UAV state information at the current moment and calculate the estimated variance matrix M k [t|t].

2. The perception-based communication beam tracking method according to claim 1, wherein, step (4) includes the following steps: During the tracking process, the transmitting base station emits beams according to the predicted state information of the UAVs to complete communication with the UAVs. For K UAVs, the sum of the communication rates between them and the base station is expressed as where SNR k represents the signal-to-noise ratio of the signal received by the k-th UAV.

Images