Interception method based on target motion trail prediction

Abstract

The invention discloses a target motion trajectory prediction-based interception method, which comprises the following steps of: firstly, establishing an aircraft trajectory library under different conditions according to a motion relationship adopted by an attacking aircraft in different motion modes of an aerial carrier, realizing a multi-step prediction framework of an aircraft target trajectory by adopting an encoder-decoder structure, and finally, carrying out target trajectory prediction. And designing an interception guidance law by using the predicted target trajectory information. According to the method, the predicted target trajectory information is used for interception guidance, and head-on small miss distance rendezvous of the interception projectile and the target is achieved.

Description

Technical Field

[0001] The invention belongs to the technical field of target detection, and in particular relates to an interception method based on target motion trajectory prediction. Background Art

[0002] At present, most of the interception technology research based on trajectory prediction adopts traditional methods, such as linearization model and extended Kalman filter theory (EKF). Traditional methods have problems such as complex dynamic models, poor prediction accuracy, and limited application scenarios. With the development of computer technology, deep learning has been widely studied and applied in image classification, speech recognition, target detection and other fields with its powerful feature extraction ability, efficient classification and regression ability, and large-scale data processing ability. In addition, more and more researchers are applying deep learning to trajectory prediction, using special networks to learn historical trajectories, infer motion models, and then predict future trajectories.

[0003] Sun Xueqi and Zhang Rui (Research on Anti-HTV2 Mid-Range Guidance Based on Trajectory Tracking Prediction, Modern Defense Technology, 2022, 50(04), 62-72) used an adaptive grid interactive multi-model filter to obtain target motion information, used the least squares method based on the forgetting factor to predict the target trajectory, obtained the motion state at the target prediction handover point, and used it as the mid-range guidance constraint to obtain the mid-range guidance instruction. The algorithm predicts the target trajectory based on the target kinematic model, uses the designed filter to track the target, and uses the least squares method to calculate the model coefficient based on the target tracking result. The method is relatively complicated, and the calculated model coefficient is quite different from the actual one, resulting in poor prediction results. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention provides an interception method based on target motion trajectory prediction. First, according to the motion relationship adopted by the attacking aircraft under different motion modes of the carrier aircraft, an aircraft trajectory library under different conditions is established, and an encoder-decoder structure is used to implement a multi-step prediction framework of the aircraft target trajectory. Finally, the predicted target trajectory information is used to design an interception guidance law. The present invention uses the predicted target trajectory information for interception guidance, and realizes a head-on rendezvous with the target with a small miss amount.

[0005] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0006] Step 1: Input the aircraft trajectory data. The aircraft trajectory contains a total of 6-dimensional data information (x, y, z, dx, dy, dz), where (x, y, z) is the three-dimensional spatial position coordinates of the aircraft in the geographic coordinate system, and (dx, dy, dz) is the unit time change of the aircraft position coordinates along the three axes in the geographic coordinate system;

[0007] The flight trajectory data is processed by removing outliers and normalizing the data; the Min-Max normalization method is used to normalize the original data, and the formula is as follows:

[0008]

[0009] Among them, D normalization represents the normalized data, D represents the data to be normalized, and D min Indicates the minimum value of all data, D max Indicates the maximum value among all data;

[0010] Step 2: Use the multi-step trajectory prediction framework of the aircraft target with an encoder-decoder structure to predict the trajectory of the incoming aircraft target. The specific process is as follows:

[0011] Step 2-1: Using the flight trajectory data processed in step 1, the trajectory point data is obtained by interval sampling. One trajectory point data is obtained every a1 points, and the data step length for trajectory prediction is a2s;

[0012] Step 2-2: Build an aircraft target trajectory prediction model based on the "encoder-decoder" structure, which consists of three parts: an encoder based on a GRU network, a decoder based on a GRU network, and a linear layer for representing the output; the GRU encoder is used to learn the hidden features of the input data and convert it into a feature vector; the GRU decoder predicts the representation of the future time series based on the latent space variables, and finally uses the linear layer to represent the decoded prediction value as an observable target trajectory prediction value;

[0013] Step 2-3: Use the error between the predicted value and the true value as the evaluation index, calculated as the average Euclidean distance difference between the predicted endpoint position and the true endpoint position;

[0014] Step 3: Intercept based on the result of trajectory prediction. The position of the target trajectory point predicted at a3s is used as the virtual static target point. The interceptor missile takes the virtual static target point as the interception target before a3s, and takes the incoming aircraft as the target for interception after a3s. The specific process is as follows:

[0015] Step 3-1: Using the trajectory position result of the incoming aircraft predicted in step 2, the interceptor missile launched by the carrier aircraft sets the proportional guidance command according to the predicted position and flies to the predicted position; the proportional guidance law requires that the command acceleration signal is proportional to the line of sight change rate. This ratio is equal to the navigation ratio multiplied by the relative speed between the missile and the target. The guidance command is written as:

[0016]

[0017] Among them is Aircraft overload command, V r is the target approach speed of the aircraft, K is the proportional coefficient of the proportional guidance law, q ε represents the plumb line of sight angular velocity, q β represents the horizontal line of sight angular velocity, Indicates normal overload, Indicates lateral overload;

[0018] Step 3-2: After the interceptor missile reaches the predicted position, it will conduct a proportional guidance interception with the incoming aircraft as the target;

[0019] A neural network is used to learn the historical data of highly maneuverable nonlinear aircraft and predict its future trajectory. The location of the predicted target trajectory point is then used as a virtual stationary target point for guided interception based on the predicted virtual stationary target. After reaching the predicted virtual target, the incoming aircraft is used as the target for guided interception.

[0020] Preferably, a1=10, a3=0.1, a3=20.

[0021] Preferably, the evaluation index is an endpoint displacement error index FDE.

[0022] Preferably, in step 2-3, the historical trajectory data of the aircraft in the previous 16 seconds is used as input to predict the result of the position in the 20th second.

[0023] The beneficial effects of the present invention are as follows:

[0024] The present invention utilizes deep learning technology and an aircraft trajectory prediction model based on an "encoding-decoding" structure to achieve higher target trajectory prediction accuracy; the predicted target trajectory information is used for interception guidance to achieve a head-on rendezvous between the interceptor missile and the target with a small miss amount. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart of the method of the present invention.

[0026] Figure 2 It is a schematic diagram of XY plane trajectories of two guidance modes according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0028] The present invention uses deep learning technology to predict the motion trajectory of the target, and uses the predicted target trajectory to design an interception guidance law. In view of the target trajectory prediction problem, an encoder-decoder structure is used to predict the trajectory of the incoming target. Then, the predicted target trajectory information is used to design an interception guidance law so that the interceptor missile launched by the aerial platform will intersect the incoming aircraft head-on with a small miss margin.

[0029] The present invention is realized by the following technical solutions: an interception method based on target motion trajectory prediction, according to the motion relationship adopted by the attacking aircraft under different motion modes of the carrier aircraft, an aircraft trajectory library under different conditions is established, an encoder-decoder structure is used to realize the multi-step prediction framework of the aircraft target trajectory, and finally, the interception guidance law is designed using the predicted target trajectory information. The details are as follows:

[0030] The present invention utilizes the characteristic that the trajectory of an aircraft is constrained by the carrier platform to establish an aircraft maneuvering trajectory database, and the content mainly includes carrier flight simulation and aircraft flight simulation. First, a motion model of the carrier is established, and the flight trajectory data of the carrier is obtained through simulation. Then, a motion model of the incoming aircraft is established. From the perspective of the incoming aircraft, the carrier is its target. For each trajectory in the carrier flight trajectory data set, the aircraft uses a certain guidance method to simulate and guide the carrier, and obtains the maneuvering trajectory of the incoming aircraft under different maneuvering conditions, so as to establish an aircraft maneuvering trajectory database.

[0031] The present invention proposes a multi-step prediction framework for aircraft trajectories based on an encoder-decoder structure of an attention mechanism. The entire framework consists of four parts: an encoder based on a gated recurrent unit (GRU) network, a decoder based on a GRU network, and a linear layer for characterizing the output. Among them, the GRU encoder is used to learn the hidden features of the input data and convert it into a feature vector. The GRU decoder predicts the representation of future time series based on the latent space variables, and finally uses a linear layer to represent the decoded prediction value as an observable target trajectory prediction value. GRU is also a model for processing sequence data. Its structure is similar to that of the LSTM network. It is a variant of the LSTM network. Compared with LSTM, its structure is simpler, the calculation process is simpler, the parameters are fewer, the convergence is faster, and it is more suitable for situations where the data set is limited.

[0032] When the target is guided, if the interceptor missile keeps tracking and intercepting its maneuvering trajectory and also maneuvers accordingly, most of the energy of the interceptor missile may be consumed, so that the interception mission cannot be completed later. In this case, the present invention uses the predicted target trajectory information to obtain the approximate position of the future target after a period of time, and first uses this position as the interception virtual target point for interception movement, which will increase the probability of finding the target and save the energy consumed by interception.

[0033] Example:

[0034] The basic process of the interception method based on the prediction of the target trajectory of the aircraft is as follows: Figure 1 As shown, the specific implementation of the present invention is described by using an example, but the technical content of the present invention is not limited to the scope described. The specific implementation includes the following steps:

[0035] Step 1: Input the aircraft trajectory data. The aircraft trajectory contains a total of 6-dimensional data information (x, y, z, dx, dy, dz), where (x, y, z) is the three-dimensional spatial position coordinates of the aircraft in the geographic coordinate system, and (dx, dy, dz) is the unit time change of the aircraft position coordinates along the three axes in the geographic coordinate system.

[0036] Trajectory data processing, in order to reduce the computational complexity of the neural network, improve training efficiency and improve training results, the flight trajectory data is processed by removing outliers and normalizing the data. The Min-Max normalization method is used to normalize the original data, and the formula is as follows:

[0037]

[0038] Step 2: Use the multi-step trajectory prediction framework of the aircraft target with an encoder-decoder structure to predict the trajectory of the incoming aircraft target. The specific process is as follows:

[0039] (1) Using the trajectory data processed in step 1, the trajectory point data is obtained by interval sampling. One trajectory point data is taken every 10 points, and the data step length used for trajectory prediction is 0.1s.

[0040] (2) Build an aircraft target trajectory prediction model based on the "encoder-decoder" structure. The entire model consists of four parts: an encoder based on a GRU network, a decoder based on a GRU network, and a linear layer for representing the output. Among them, the GRU encoder is used to learn the hidden features of the input data and convert it into a feature vector. The GRU decoder predicts the representation of the future time series based on the latent space variables, and finally uses the linear layer to represent the decoded prediction value as an observable target trajectory prediction value.

[0041] (3) The error between the predicted value and the true value is used as the evaluation index, namely the final displacement error (FDE), which is calculated as the average Euclidean distance difference between the predicted final position and the true final position. The historical trajectory data of the aircraft in the first 16 seconds is used as input to predict the position at the 20th second.

[0042] Step 3: Intercept based on the result of trajectory prediction. The position of the target trajectory point predicted at the 20th second is used as the virtual static target point. The interceptor missile takes the virtual static target point as the interception target before the 20th second, and takes the incoming aircraft as the target for interception movement after the 20th second. The specific process is as follows:

[0043] (1) Using the trajectory position result of the incoming aircraft predicted in step 2 at the 20th second, the interceptor missile launched by the carrier aircraft sets the proportional guidance command according to the predicted position and flies to the predicted position. The proportional guidance law requires that the command acceleration signal is proportional to the rate of change of the line of sight. This ratio is equal to the navigation ratio multiplied by the relative speed between the missile and the target. The guidance command can be written as:

[0044]

[0045] Among them is Aircraft overload command, V r is the target approach speed of the aircraft.

[0046] (2) After the interceptor missile reaches the predicted position, it will then target the incoming aircraft for proportional guidance interception. Traditional proportional guidance has always targeted incoming aircraft, and has poor tracking effect on highly maneuverable and nonlinear incoming aircraft. It will also cause excessive overload at the end, which seriously limits the interceptor missile's guidance and interception effect. The interception guidance based on trajectory prediction uses a neural network to learn the historical data of highly maneuverable and nonlinear aircraft, predict its future trajectory, and then use the predicted target trajectory point as a virtual stationary target point. Guided interception is performed based on the predicted virtual stationary target. After reaching the predicted virtual target, the incoming aircraft is targeted for guided interception, which improves the tracking effect of the incoming aircraft and reduces the overload at the end of the interception.

[0047] In order to demonstrate the necessity and effectiveness of the interception algorithm based on trajectory prediction, the present invention designs a simulation verification test. Assuming that the carrier aircraft flies in a rolling maneuver, the simulation results of the interceptor missile using the interception guidance method based on trajectory prediction and the guidance method based on proportional guidance to intercept the incoming aircraft are shown in the following figure. Figure 2 shown.

[0048] The interception guidance method based on trajectory prediction is compared with the guidance method based on proportional guidance to intercept the incoming aircraft. The miss amount, intersection angle, maximum longitudinal overload and maximum lateral overload are used as evaluation indicators. The comparison results are shown in Table 1. The results show that the interception guidance based on target trajectory prediction has a smaller miss amount, a smaller intersection angle and a smaller actual overload.

[0049] Table 1 Performance comparison of two guidance methods

[0050] Based on trajectory prediction Proportional navigation Missing distance(m) 0.0280 57.6353 Intersection angle (°) 18.6468 20.1726 Maximum longitudinal overload 9.25 80 Maximum lateral overload 30 24.5

Claims

1. An interception method based on target motion trajectory prediction, characterized in that: The steps include: Step 1: Input the aircraft trajectory data. The aircraft trajectory contains a total of 6-dimensional data information (x, y, z, dx, dy, dz), where (x, y, z) is the three-dimensional spatial position coordinates of the aircraft in the geographic coordinate system, and (dx, dy, dz) is the unit time change of the aircraft position coordinates along the three axes in the geographic coordinate system; The flight trajectory data is processed by removing outliers and normalizing the data; the Min-Max normalization method is used to normalize the original data, and the formula is as follows: Among them, D normalization represents the normalized data, D represents the data to be normalized, and D min Indicates the minimum value of all data, D max Indicates the maximum value among all data; Step 2: Use the multi-step trajectory prediction framework of the aircraft target with an encoder-decoder structure to predict the trajectory of the incoming aircraft target. The specific process is as follows: Step 2-1: Using the flight trajectory data processed in step 1, the trajectory point data is obtained by interval sampling. One trajectory point data is obtained every a1 points, and the data step length for trajectory prediction is a2s; Step 2-2: Build an aircraft target trajectory prediction model based on the "encoder-decoder" structure, which consists of three parts: an encoder based on a GRU network, a decoder based on a GRU network, and a linear layer for representing the output; the GRU encoder is used to learn the hidden features of the input data and convert it into a feature vector; the GRU decoder predicts the representation of the future time series based on the latent space variables, and finally uses the linear layer to represent the decoded prediction value as an observable target trajectory prediction value; Step 2-3: Use the error between the predicted value and the true value as the evaluation index, calculated as the average Euclidean distance difference between the predicted endpoint position and the true endpoint position; Step 3: Intercept based on the result of trajectory prediction. The position of the target trajectory point predicted at a3s is used as the virtual static target point. The interceptor missile takes the virtual static target point as the interception target before a3s, and takes the incoming aircraft as the target for interception after a3s. The specific process is as follows: Step 3-1: Using the trajectory position result of the incoming aircraft predicted in step 2, the interceptor missile launched by the carrier aircraft sets the proportional guidance command according to the predicted position and flies to the predicted position; the proportional guidance law requires that the command acceleration signal is proportional to the line of sight change rate. This ratio is equal to the navigation ratio multiplied by the relative speed between the missile and the target. The guidance command is written as: Among them is Aircraft overload command, V r is the target approach speed of the aircraft, K is the proportional coefficient of the proportional guidance law, q ε represents the plumb line of sight angular velocity, q β represents the horizontal line of sight angular velocity, Indicates normal overload, Indicates lateral overload; Step 3-2: After the interceptor missile reaches the predicted position, it will conduct proportional guidance interception with the incoming aircraft as the target; A neural network is used to learn the historical data of highly maneuverable nonlinear aircraft and predict its future trajectory. The location of the predicted target trajectory point is then used as a virtual stationary target point for guided interception based on the predicted virtual stationary target. After reaching the predicted virtual target, the incoming aircraft is used as the target for guided interception.

2. The interception method based on target motion trajectory prediction according to claim 1 is characterized in that: Said a1=10, a3=0.1, a3=20.

3. The interception method based on target motion trajectory prediction according to claim 1 is characterized in that: The evaluation index is the endpoint displacement error index FDE.

4. The interception method based on target motion trajectory prediction according to claim 1 is characterized in that: In step 2-3, the historical trajectory data of the aircraft in the previous 16 seconds is used as input to predict the result of the position in the 20th second.

Images