Unmanned aerial vehicle beam tracking method based on analog beam

Abstract

The invention discloses an unmanned aerial vehicle wave beam tracking method based on an analog wave beam. The method comprises the following steps that S1, a base station end carries out wave beam scanning; s2, performing angle estimation based on AI; s3, positioning the target to obtain coordinate information of the target; s4, establishing an observation equation for the target, and performing tracking filtering by using extended Kalman filtering; s5, adjusting the beam center to the direction of the target; s6, the base station communicates with the unmanned aerial vehicle through the wave beam, and a communication link is established; and S7, monitoring the SNR of the target. The method is suitable for arrays in any form, uniform linear arrays and sparse arrays and can also be expanded to various area arrays, beam scanning of the method can be based on digital beams or analog beams, and a low-cost beam scanning scheme can be achieved.

Description

Technical Field

[0001] The present invention relates to the field of communication perception, and in particular, to a method for tracking an unmanned aerial vehicle (UAV) beam based on analog beamforming. Technical Background

[0002] In the current context of the integration of communication and perception, the technology for detecting and tracking UAVs has become one of the hot research directions in the field of communication and perception integration. For the communication and perception scenario, there are two major scenarios for UAV tracking. One is the "illegal flight" scenario. The second type is legal and autonomous UAVs, which are used to monitor the environment around the base station or to achieve communication with the base station. For either type of UAV, it is necessary to perform beam tracking on it through a sensing system.

[0003] Traditional UAV beam tracking is usually based on codebook-based beam scanning. The number of such codebooks is fixed, and these beams need to cover all the sensing ranges. When the UAV or other various targets are exactly located at the center of its beam, of course, it can detect them and accurately obtain their angle information. However, when the target deviates from the beam center, using such beam detection methods will result in a large angle measurement error, thus severely reducing the positioning accuracy of the sensing system for the target, as Figure 1 shown. Figure 1 Fig. shows the beam scanning schematic diagram of the base station for the UAV. Among them, when the UAV is exactly located at the center of the corresponding beam of the codebook, the base station can obtain the accurate angle information of the UAV, thus achieving precise positioning of the UAV. However, when the UAV deviates from the center of the beam, its accurate angle cannot be directly obtained based on the beam scanning, resulting in a large error in its positioning.

[0004] Therefore, the present invention provides a method for tracking a UAV by analog beam scanning in a communication and perception integrated scenario. The method provided by the present invention can also give the accurate angle information of the target when the target is not located at the center of the analog beam, thus achieving accurate detection and positioning in the entire airspace. The scenario targeted by the present invention is that a communication link needs to be established between the base station and the UAV, and the beam scanning network function is to assist in establishing a communication link between the base station and the UAV. Summary of the Invention

[0005] The present invention provides a method for tracking a UAV beam based on analog beam scanning in a communication and perception scenario to solve the problems existing in the existing methods. The core idea and the problem to be solved by the present invention are the problem of large angle measurement error when the target UAV is not at the center of the analog beam. The present invention introduces the idea of artificial intelligence (AI), preprocesses the information obtained by the analog beam, and uses the corresponding features obtained to calculate the accurate angle information of the target.

[0006] The present invention adopts the following technical solutions:

[0007] A UAV beam tracking method based on analog beams, comprising the following steps:

[0008] S1: The base station performs beam scanning;

[0009] S2: Angle estimation is performed based on AI;

[0010] S3: Locate the target to obtain the coordinate information of the target;

[0011] S4: Establish an observation equation for the target and perform tracking filtering using extended Kalman filtering;

[0012] S5: Adjust the beam center to the direction where the target is located;

[0013] S6: The base station communicates with the UAV using the beam to establish a communication link;

[0014] S7: Monitor the target SNR.

[0015] Further, the implementation method of beam scanning in step S1 is digital beam or analog beam.

[0016] Further, the beam scanning range performs full-airspace scanning, and the scanning range and beam scanning interval are determined according to the array size, etc. For example, the scanning beam can be set to [-60, -50, -40, -30, -20, -10, 0, 10, 20, 30, 40, 50, 60] degrees.

[0017] Further, the angle estimation algorithm based on AI in step S2 includes: First, obtain a training sample set, that is, obtain the amplitude vectors of the scanning beams at different angles, train the AI model with these training sets, and then apply the beam scanning results obtained each time to the obtained model to calculate the angle corresponding to the current scanning result.

[0018] Further, the angle estimation algorithm based on AI includes the following steps:

[0019] S21: Obtain the received signal amplitude value of each beam and construct a sample set;

[0020] S22: Establish an AI model and train to obtain an AI model;

[0021] S23: Input the amplitude vector obtained from the current scan into the AI network to obtain the corresponding angle.

[0022] Further, the content of obtaining the received signal amplitude value of each beam and constructing a sample set in step S21 includes:

[0023] Let the beam search codebooks be [W 1 ,W 2 ,...,WM , where W k ∈ C N×1 , k = 1, 2, ..., M, N is the number of antennas. Let the beam steering vectors corresponding to each codebook be [a 1 , a 2 , ..., a M , then the received signal amplitude can be expressed as:

[0024]

[0025] Write the amplitude obtained from each codebook in vector form, that is:

[0026] A i0 = [A 1 , A 2 , ..., A M , i = 1, 2, 3..., I

[0027] where I is the total number of samples;

[0028] Normalize each amplitude vector, we have:

[0029]

[0030] Encode the label value corresponding to each sample according to one - hot encoding.

[0031] Furthermore, a fully - connected neural network model is adopted in step S22.

[0032] Furthermore, the fully - connected neural network includes an input layer, a hidden layer, and an output layer.

[0033] Furthermore, the coordinate information of the target in step S3 includes two - dimensional or three - dimensional information.

[0034] Furthermore, in step S7, monitor the target SNR. If the SNR is lower than the threshold SNR_THR, restart the beam scanning and update the UAV beam.

[0035] The scenario targeted by the present invention is the UAV target beam tracking under simulated beams. Since under simulated beams, usually through the codebook method, the phase of each antenna is adjusted to achieve the beam scanning function. And the number of codebooks is generally fixed and used to cover the entire airspace. The simulated beam architecture has only one digital link. Therefore, the signals of each antenna channel cannot be obtained, and only the total received signal value under a certain codebook can be obtained.

[0036] Therefore, under simulated beam scanning, it is impossible to obtain the angle of the target using all the antenna channel data, and only the beam angle corresponding to the codebook can be used as the estimated angle of the target. When the target is at the exact center of the beam corresponding to the codebook, an accurate angle value of the target can be given. Conversely, when the target deviates from the beam center, there will be an error in the angle information given according to the codebook direction, which leads to a positioning error of the target.

[0037] The present invention utilizes the characteristics of the amplitude information obtained by beam scanning, takes the amplitude vector obtained by each beam scanning as a feature, constructs a neural network with the amplitude vector as the feature input, and performs one-hot encoding on the angle information corresponding to each feature vector as the label information corresponding to the amplitude vector.

[0038] Thus, it is possible to give an accurate target angle under the condition of only obtaining the scanned amplitude vector, without the need to obtain the signals of each antenna. Therefore, the present invention can obtain more accurate angle information under simulated beams.

[0039] The features of the present invention are as follows:

[0040] 1. The present invention is based on a beam scanning and tracking system under simulated beams;

[0041] 2. The present invention can train a deep neural network based on the amplitude vector of simulated beam scanning, and the neural network gives the accurate angle of the target;

[0042] 3. The present invention does not need to obtain the signals of each antenna channel, only needs the amplitude vector composed of the amplitudes of multiple scans, uses these amplitude vectors to establish an AI network, and calculates the angle information corresponding to each amplitude vector based on this network;

[0043] 4. The present invention also monitors the SNR of the UAV target and decides whether to perform beam scanning again according to the SNR monitoring result;

[0044] 5. The present invention can also be applied to a digital beam network, that is, in a fully digital network, a beam scanning architecture can also be used to obtain the accurate angle of the target, which can reduce part of the computation;

[0045] 6. The beam obtained by the scanning of the present invention can be used for the communication link between the base station and the UAV, which can help to quickly establish and maintain a stable connection between the UAV and the base station.

[0046] Figure 2 The comparison of the amplitude vector results of the scanning beam is given under the condition of a 10-element uniform linear array when the target is at -25 degrees and 25 degrees. It can be seen from the results that the beam amplitude vectors corresponding to different angles have obvious differences, and the amplitude vector of the beam scanning can be used as a feature for network training and angle estimation.

[0047] Compared with the prior art, the present invention has the following technical advantages:

[0048] 1. The beam scanning of the present invention can be based on digital beams or analog beams, enabling a low-cost beam scanning solution;

[0049] 2. When the target is not at the exact center of the beam, the present invention can also give an accurate angle result of the target only through beam scanning, enabling high-precision beam tracking of the target and enhancing the reliability of the link;

[0050] 3. The present invention is applicable to any form of array, including uniform linear arrays, sparse arrays, and can also be extended to various planar arrays. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a beam scanning diagram of the base station beam on the UAV;

[0052] Figure 2 is a beam amplitude vector diagram at different target angles;

[0053] Figure 3 is a UAV beam tracking flowchart;

[0054] Figure 4 is a diagram for calculating the UAV angle based on the beam scanning result;

[0055] Figure 5 is a fully connected neural network diagram for training;

[0056] Figure 6 is a partial training sample vector diagram;

[0057] Figure 7 is a fully connected neural network loss convergence result diagram. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] The implementation process of the method of the present invention is as Figure 3 shown, Figure 3 which gives the UAV beam tracking process. First, the base station performs beam scanning. Here, the beam scanning can be analog beams or digital beams, and the implementation method is not limited. Then, angle estimation is performed. The present invention adopts an AI-based angle estimation algorithm, which can utilize the amplitude information of the beam scanning to calculate the angle information of the target. Finally, the target is tracked and the SNR of the target is monitored. When the SNR is lower than the given threshold SNR_THR, a new round of beam scanning is started.

[0059] The specific steps are described as follows:

[0060] S1: The base station side performs beam scanning.

[0061] The beam scanning implementation method can be digital beam or analog beam, and the present invention does not limit this. The beam scanning range can perform full airspace scanning. For example, the scanning beam can be set as [-60, -50, -40, -30, -20, -10, 0, 10, 20, 30, 40, 50, 60] degrees. The present invention does not need to obtain the received signal of each antenna during scanning, and only the received amplitude of each beam is required.

[0062] S2: Perform angle estimation based on AI, and the implementation process is shown in Figure 4 , Figure 4 which gives the flowchart of the angle estimation algorithm of the present invention. The algorithm first needs to obtain a training sample set, that is, it needs to obtain the amplitude vectors of the scanning beams at different angles. These training sets are used to train the AI model. Then, the beam scanning results obtained each time are applied to the obtained model to calculate the angle corresponding to the current scanning result.

[0063] S21: Obtain the received signal amplitude value of each beam and construct a sample set.

[0064] Let the beam search codebooks be [W 1 , W 2 ,..., W M , where W k ∈C N×1 , k = 1, 2,..., M, and N is the number of antennas. Let the beam steering vectors corresponding to each codebook be [a 1 , a 2 ,..., a M , then the received signal amplitude can be expressed as:

[0065]

[0066] Write the amplitude obtained from each codebook in vector form, that is:

[0067] A i0 = [A 1 , A 2 ,..., A M , i = 1, 2, 3..., I

[0068] where I is the total number of samples.

[0069] Normalize each amplitude vector, and we have:

[0070]

[0071] For each sample's corresponding label value, perform one-hot encoding, that is, only 1 value in each label is 1 and the rest are 0. The encoding rule is as follows: encode all the ranges of angles to be measured separately. For example, in the range of [-60, 60] degrees, if the interval is set to 1 degree, then there are 121 angles to be encoded, that is, the label length of each sample is 121. Set the bit corresponding to this angle to 1 and the rest to zero. For example, for the angle of -60 degrees, the corresponding label can be written as [1, 0, 0,..., 0].

[0072] S22: Establish an AI model and train it to obtain the AI model.

[0073] A fully connected neural network model can be adopted, as shown in Figure 5 , Figure 5 shows a fully connected neural network adopted by the present invention, which includes an input layer, a hidden layer, and an output layer. The input data of the input layer is the amplitude vector of beam scanning. The hidden layer processes the input data, and the output layer gives the classification label result. Angle information can be obtained from the label. It should be noted that only the basic model structure is shown in the figure, and the specific model parameters need to be adjusted according to specific parameters such as the specific antenna arrangement and the angle scanning range.

[0074] For example, for a uniform linear array with 10 array elements, the scanning range is [-60, 60] degrees, and the scanning interval is 10 degrees. A feasible network structure is shown in Table 1.

[0075] Table 1 Network structure used for training

[0076] Number of layers Number of outputs Input layer 13 Hidden layer 1 32 Hidden layer 2 64 Hidden layer 3 512 Hidden layer 4 1024 Output layer 121

[0077] S23: Input the amplitude vector obtained from the current scan into the AI network to obtain the corresponding angle.

[0078] S3: Locate the target to obtain the coordinate information of the target, that is, two-dimensional or three-dimensional information (x, y) or (x, y, z).

[0079] S4: Establish an observation equation for the target and perform tracking filtering using the extended Kalman filter.

[0080] S5: Adjust the beam center to the direction where the target is located;

[0081] S6: The base station uses this beam to communicate with the UAV and establish a communication link;

[0082] S7: Monitor the SNR of the target. If the SNR is lower than the threshold SNR_THR, restart the beam scan and update the UAV beam.

[0083] Figure 6The partial sample vectors for training are given, that is, the amplitude vectors corresponding to partial angles. Figure 7 The convergence results of the training loss value of the fully connected neural network are given.

Claims

1. A UAV beam tracking method based on simulated beam, characterized in that: The following steps are involved: S1: The base station performs beam scanning; S2: Angle estimation based on AI; S3: locate the target and obtain the coordinate information of the target; S4: Establish observation equation for the target and use extended Kalman filter for tracking filtering; S5: Adjust the beam center to the direction of the target; S6: The base station communicates with the drone using the beam to establish a communication link; S7: Monitor the target SNR.

2. The UAV beam tracking method based on simulated beam according to claim 1, characterized in that: The beam scanning in step S1 is implemented by digital beam or analog beam.

3. The UAV beam tracking method based on simulated beam according to claim 2, characterized in that: The beam scanning range performs a full spatial scan, and the scanning range and beam scanning interval can be determined based on the array size. For example, the scanning beam can be set to [-60, -50, -40, -30, -20, -10, 0, 10, 20, 30, 40, 50, 60] degrees.

4. The UAV beam tracking method based on simulated beam according to claim 1, characterized in that: The AI-based angle estimation algorithm in step S2 includes: first, obtaining a training sample set, that is, obtaining the amplitude vector of the scanning beam at different angles, training the AI ​​model with these training sets, and then applying each beam scanning result obtained to the obtained model to calculate the angle corresponding to the current scanning result.

5. The UAV beam tracking method based on simulated beam according to claim 4, characterized in that: The AI-based angle estimation algorithm includes the following steps: S21: Obtain the received signal amplitude value of each beam and construct a sample set; S22: Establish an AI model and train it to obtain the AI ​​model; S23: Input the amplitude vector obtained by the current scan into the AI ​​network to obtain the corresponding angle.

6. The UAV beam tracking method based on simulated beam according to claim 5, characterized in that: In step S21, the received signal amplitude value of each beam is obtained, and the contents of constructing the sample set include: Assume that the beam search codebooks are [W1, W2, ..., W M ], where W k ∈C N×1 , k = 1, 2, ..., M, N is the number of antennas, and the beam steering vectors corresponding to each codebook are [a1, a2, ..., a M ], the received signal amplitude can be expressed as: The amplitude obtained from each codebook is written in vector form, namely: A i0 =[A1,A2,...,A M ],i=1,2,3...,I Where, I is the total number of samples; Normalize each magnitude vector to get: Corresponding label values ​​to each sample are encoded according to one-hot encoding.

7. The UAV beam tracking method based on simulated beam according to claim 5, characterized in that: In step S22, a fully connected neural network model is adopted.

8. The UAV beam tracking method based on simulated beam according to claim 7, characterized in that: The fully connected neural network includes an input layer, a hidden layer, and an output layer.

9. The UAV beam tracking method based on simulated beam according to claim 1, characterized in that: The coordinate information of the target in step S3 includes two-dimensional or three-dimensional information.

10. The UAV beam tracking method based on simulated beam according to claim 1, characterized in that: In step S7, the target SNR is monitored. If the SNR is lower than the threshold SNR_THR, the beam scan is restarted to update the UAV beam.

Images