CN113095504A - Target track prediction system and prediction method - Google Patents

Abstract

A target trajectory prediction system includes a target maneuvering feature model set module, the target maneuvering feature set includes typical maneuvering local models of several targets; the forward trajectory prediction module defines a performance index function and establishes a target trajectory prediction model, according to each typical maneuvering local model. The matching degree between the model and the actual flight trajectory of the target determines the parameters of the target trajectory prediction model, and the numerical integration method is used to predict the forward trajectory; the backward feedback correction module establishes a global importance map model, and establishes a forward prediction according to the global importance map model. The relationship between the trajectory and the important flight nodes is determined, and the target maneuvering intention information is determined. After the parameters of the target trajectory prediction model are modified by the posterior conditional probability method, the trajectory prediction is performed again. The invention breaks through the open-loop prediction limitation of the traditional trajectory prediction method, integrates the prediction trajectory feedback correction based on game confrontation into the trajectory prediction process, realizes the trajectory closed-loop prediction, and improves the trajectory prediction accuracy.

Description

A target trajectory prediction system and prediction method

technical field

The invention belongs to the technical field of target trajectory prediction, and in particular relates to a trajectory prediction system and a prediction method of a flying target.

Background technique

Target trajectory prediction is a process of estimating the future motion state or reachable range of the target according to the target motion state parameters obtained by the detection system. Through effective trajectory prediction, the understanding of the subsequent state of the target can be improved, and the basis for related mission planning and management can be provided. It has broad application prospects in the fields of aviation control and aerospace target interception.

At present, trajectory prediction methods are mainly divided into two categories: analytical method and numerical integration method. Analytical method is a relatively intuitive trajectory prediction method. Its main principle is to simplify on the condition of a specific motion pattern or motion relationship. However, when the target motion pattern is complex and changeable, it is often difficult to obtain an analytical solution. Unknown non-cooperative targets are predicted, the method is more difficult to adapt. The numerical integration method obtains the target predicted trajectory by forward integration of the dynamic model. Similarly, when the target movement mode changes, especially when the trajectory is predicted for the non-cooperative target whose maneuvering mode is unknown, due to the prediction of the numerical integration method. The process is an open-loop process, and there is no external information to correct the predicted trajectory, which will lead to the continuous accumulation of errors in the prediction process and rapid divergence, which is difficult to meet the actual trajectory prediction requirements of non-cooperative targets.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a target trajectory prediction system and prediction method based on game confrontation which can improve the trajectory prediction accuracy.

In order to achieve the above object, the present invention adopts the following technical solutions:

A target trajectory prediction system, comprising: a target maneuvering feature model set module for establishing a target maneuvering feature model set, the target maneuvering feature set including typical maneuvering local models of several targets; a target maneuvering feature set for predicting the forward trajectory of the target A forward trajectory prediction module, the forward trajectory prediction module defines a performance index function and establishes a target trajectory prediction model, and calculates the matching degree between each typical maneuvering local model in the target maneuvering feature model set and the target actual flight trajectory through the performance index function, Determine the parameters of the target trajectory prediction model according to the matching degree between each typical maneuvering local model and the actual flight trajectory of the target, substitute the obtained parameters into the target trajectory prediction model, and use the numerical integration method to predict the target's forward trajectory;

A backward feedback correction module for correcting the target trajectory, the backward feedback correction module establishes a global importance map model according to the important flight nodes, and then establishes the association between the forward predicted trajectory and the flight important nodes according to the global importance map model relationship, determine the target maneuver intention information, use the target maneuver intention information, use the posterior conditional probability method to modify the parameters of the target trajectory prediction model, and then perform trajectory prediction according to the target trajectory prediction model after parameter correction, and obtain the modified trajectory prediction. result.

Further, when establishing the target maneuvering feature set, the target maneuvering feature model set module first determines the maneuvering feature model representing the target maneuvering mode, and then decomposes the target maneuvering state to establish a typical maneuvering partial model of the target, each typical maneuvering partial model. The model constitutes the target maneuvering feature model set.

Further, the maneuvering feature model includes a CA model, a CV model, a typical trajectory data model, and a dynamic model.

The present invention also provides a target trajectory prediction method, comprising the following steps:

S1. Establish a target maneuvering feature model set, first determine the maneuvering feature model of the target, then decompose the target maneuvering state, establish a typical maneuvering local model of the target based on the maneuvering feature model, and the typical maneuvering local model constitutes the target maneuvering feature model set;

S2, target forward trajectory prediction, define the performance index function and establish the target trajectory prediction model, and evaluate the matching degree of each typical maneuvering local model in the target maneuvering feature model set with the target actual flight trajectory through the performance index function. The matching degree between the model and the actual flight trajectory of the target determines the parameters of the target trajectory prediction model, and the numerical integration method is used to predict the forward trajectory of the target based on the target trajectory prediction model;

S3, target trajectory correction, first determine the important flight nodes, then establish a global importance map model according to the flight important nodes, and then establish the relationship between the forward predicted trajectory and the flight important nodes according to the global importance map model, and determine the target maneuvering intention information, Then, the parameters of the target trajectory prediction model are modified by the posterior conditional probability method using the target maneuver intention information, and then the target trajectory is predicted according to the target trajectory prediction model after parameter correction, and the modified trajectory prediction result is obtained.

Further, in step S2, a weighted least squares method is used to solve the parameters of the target trajectory prediction model.

式中的J_i(k)表示目标机动特征模型集中第i个典型机动局部模型的性能指标值，M为目标机动特性模型集中典型机动局部模型的数量。Further, in step S2, the tracking error weighting method is used to define the performance index function, and the matching degree between the typical maneuvering local model and the actual flight trajectory of the target is equal to the weight of the typical maneuvering local model.

In the formula, J _i (k) represents the performance index value of the ith typical maneuvering local model in the target maneuvering characteristic model set, and M is the number of typical maneuvering partial models in the target maneuvering characteristic model set.

式中的p(x_k-1＝j)为k-1时刻第j个典型机动局部模型的先验概率，p(x_k＝i,x_k-1＝j)为典型机动局部模型对应的预测轨迹后验概率。Further, in step S3, the Bayesian inference method is used to determine the target maneuvering intention information, and the target maneuvering intention information is:

where p(x _k-1 =j) is the prior probability of the j-th typical maneuvering local model at time k-1, p(x _k =i, x _k-1 =j) is the corresponding value of the typical maneuvering local model Predicted trajectory posterior probability.

其中，

为目标机动特征模型集中的典型机动局部模型的参数。Further, the method for modifying the parameters of the target trajectory prediction model in step S3 is as follows: using the target maneuvering intention information p(x _k =i|x _k-1 =j) to correct the target prediction state, there are: p(x _k | x _1:k-1 )＝∫p(x _k |x _k-1 )p(x _k-1 |x _1:k-1 )dx _k-1 , and then according to the correction result p(x _k |x _{1 : k-1} ) Modify the parameters of the target trajectory prediction model, including:

in,

The parameters of the typical maneuvering local model in the target maneuvering feature model set.

As can be seen from the above technical solutions, the present invention establishes a target maneuvering feature model set composed of typical maneuvering local models, and the forward trajectory prediction module, on the basis of constraining the trajectory prediction based on the target motion feature, introduces the game-based confrontation through the backward feedback correction module. The target maneuvering intention evaluation is based on the target maneuvering intention, and the predicted trajectory is corrected backward, forming a closed-loop trajectory prediction system that combines forward prediction under the constraint of motion characteristics and prediction trajectory feedback correction under the game maneuvering intention evaluation, breaking through the traditional open-loop trajectory prediction method. Due to the limitations of prediction, the prediction trajectory feedback correction based on game confrontation is integrated into the trajectory prediction process to achieve closed-loop trajectory prediction and improve the trajectory prediction accuracy. Moreover, in the closed-loop prediction process of the present invention, target maneuvering characteristics and game confrontation intention are used to constrain the trajectory prediction range, and since the cooperative target has richer maneuvering intention information, it has a better prediction effect. The present invention is suitable for both cooperative target trajectory prediction and non-cooperative target trajectory prediction, the prediction system has good compatibility, and can be compatible with different types of data filtering and processing algorithms in each module.

Description of drawings

1 is a block diagram of an embodiment of the present invention;

Figure 2 is a schematic diagram of the relationship between the flight trajectory of the re-entry gliding target and its local model representation;

3 is a schematic diagram of the trajectory space coverage of the maneuvering feature model set of the re-entry gliding target;

4 is a schematic diagram of an importance map;

Figure 5 is a diagram showing the relationship between the predicted trajectory and important flight nodes;

Fig. 6 is the error change curve diagram that adopts the method of the present invention to predict;

Fig. 7 is a graph showing the variation curve of the membership degree of the model;

Fig. 8 is the change contrast curve diagram of the model inclination angle β;

Fig. 9 is the change contrast curve diagram of model _KL ;

Fig. 10 is a graph showing the comparison of changes of the model K _D.

The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

Detailed ways

The present invention will be described in detail below with reference to the accompanying drawings. When describing the embodiments of the present invention in detail, for the convenience of explanation, the drawings representing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not be limited here. The scope of protection of the present invention. It should be noted that the accompanying drawings are in a simplified form and all use inaccurate scales, and are only used to facilitate and clearly assist the purpose of explaining the embodiments of the present invention.

1 is a block diagram of the system of the present invention. The target trajectory prediction system of the present invention includes a target maneuver feature model set module, a forward trajectory prediction module based on maneuver feature constraints, and a backward feedback correction module based on game confrontation.

The target maneuvering feature model set module is used to establish a target maneuvering feature model set that describes the typical maneuvering style of the target. The target maneuvering space is covered by the construction of the target maneuvering feature model set, and it is used as the prior information of the target trajectory prediction process to constrain the target trajectory prediction process. , through the introduction of target prior information to improve the target trajectory prediction accuracy.

The working process of the target maneuvering feature model set module is as follows: First, determine the maneuvering feature model representing the target maneuvering mode. The more common maneuvering feature models include CA model (constant velocity model), CV model (constant acceleration model), typical trajectory data model , dynamic model, etc. In practical applications, what maneuvering feature model is used to characterize the maneuvering mode of the target can be selected according to the motion characteristics of the target.

After determining the maneuvering feature model representing the maneuvering mode of the target, decompose the maneuvering state of the target, such as maneuvering mode, flight state, maneuvering ability, etc., to establish a typical maneuvering local model of the target, and the target maneuvering feature model is composed of several typical maneuvering partial models obtained. set. For example, if CA model and CV model are used to represent the maneuvering mode of the target, the maneuvering state of the target can be decomposed according to the target capability range, and different local characteristic model parameters in the model can be determined, thereby establishing multiple typical maneuvering local models of the target. These typical maneuvering local models constitute the target maneuvering feature model set. For another example, the typical trajectory data model is used as the maneuvering feature model to represent the maneuvering mode of the target, and the maneuvering state of the target can be decomposed according to the typical maneuvering patterns under different conditions and states of the target, and the target sample trajectory can be generated by using the dynamic principle to establish the target sample trajectory. The typical maneuvering local model of the target, and then the target maneuvering feature model set is formed.

The forward trajectory prediction module is used to establish a target trajectory prediction model, and based on the target trajectory prediction model, the forward trajectory prediction of the target is performed. The probability distribution of the typical maneuvering local model of the target, based on the target trajectory prediction model, the numerical integration method is used to predict the forward trajectory. The trajectory tracking data of the target is the data obtained after being detected, tracked and processed by the target trajectory tracking system.

The working process of the forward trajectory prediction module is as follows: define the performance index function and establish the target trajectory prediction model. The performance index function is used to evaluate the matching degree (correlation relationship) between each typical maneuvering local model in the target maneuvering feature model set and the actual flight trajectory of the target. , establishes a motion feature association constraint mechanism for trajectory prediction; establishes a kinematics or dynamics model for trajectory prediction combined with the motion features of the target, which is used for target trajectory prediction. A variety of forms can be used to define the performance index function, and the more commonly used methods are the tracking error weighting method, the tracking data interactive multi-model filtering method, and so on.

Then use the performance index function to calculate the matching degree of each typical maneuvering local model in the target maneuvering feature model set and the actual flight trajectory of the target based on the similarity principle. Predict the parameters of the model, substitute the obtained parameters into the target trajectory prediction model, and then use the numerical integration method to predict the forward trajectory of the target. The weighted least squares method can be used to solve the parameters of the target trajectory prediction model.

The backward feedback correction module is used to correct the target trajectory predicted by the forward trajectory prediction module. The backward feedback correction module uses the trajectory prediction results of the forward trajectory prediction module to evaluate the target game maneuver intention by constructing a game maneuver intention evaluation importance map model, extracts the predicted trajectory correction information with the target game maneuver intention evaluation, and then calculates the target game maneuver intention. It introduces the target trajectory prediction model established by the target trajectory forward prediction module, and corrects the predicted trajectory.

The working process of the backward feedback correction module is as follows: Firstly, a global importance map model is established according to the important flight nodes, and the global importance map model can be established by the multi-model "decomposition-combination" method, that is, the importance of each node in the airspace where the target flight is located is determined first. degree, establish a local model centered on important flight nodes, establish a global description relationship through topology mapping, and obtain a global importance map model. The important flight nodes are set accordingly according to the application scenarios. For example, important political, military or economic places and facilities that have a greater impact on the target flight mission can be set as flight important nodes.

Then, according to the global importance map model, the relationship between the forward predicted trajectory and the important flight nodes is established, and the target maneuvering intention information is determined. More specifically, the global importance map model and the forward trajectory prediction area (target trajectory prediction spatial distribution ) to establish the relationship between the forward predicted trajectory and the important flight nodes to determine the target maneuver intention information. Assuming that the greater the correlation degree is, the greater the threat degree of the target, the Bayesian inference method can be used to determine the target maneuver intention information. The forward trajectory prediction area is determined by the trajectory predicted by the target trajectory prediction model.

Using the target maneuvering intention information, the parameters of the target trajectory prediction model are modified by the posterior conditional probability method, and then the trajectory prediction is carried out according to the target trajectory prediction model after parameter correction, and the modified trajectory prediction result is obtained.

The trajectory prediction process of the target trajectory prediction system of the present invention will be further described below with a specific embodiment. This embodiment is to predict the trajectory of a re-entry gliding target.

X_l为目标第l个LGR节点的状态量，U_l为目标第l个LGR节点的飞行控制量。用典型轨迹数据模型实现对目标在特定条件下轨迹的表征，如图2所示。Firstly, the target maneuvering feature model set of the re-entry gliding target is established. The target maneuvering feature model set module can obtain the typical motion trajectory of the target by using the kinematics principle according to the motion characteristics of the re-entry gliding target. Therefore, the typical trajectory data model is used as the re-entry gliding target. Enter the maneuvering characteristic model of the gliding target, select the state variable with N LGR nodes, and use the Radau pseudospectral method to discrete points to characterize the local models of different states at different stages, T={K ₁ ,K ₂ ,...,K _N }, where,

X _l is the state quantity of the lth LGR node of the target, and U _l is the flight control quantity of the lth LGR node of the target. A typical trajectory data model is used to characterize the trajectory of the target under specific conditions, as shown in Figure 2.

进而建立目标机动特征模型集Pc如下：

对覆盖约束空间的元素进行求解，即可得到覆盖目标轨迹空间的机动特征模型集。本实施例选择再入高度为72km，再入速度为4000m/s，终端高度约束为10km，当纵向飞行距离发生变化时，对不同纵向飞行距离约束进行离线轨迹优化构成覆盖飞行范围的轨迹集如图3所示，由典型机动局部模型构成的机动特征模型集合可覆盖目标不同机动模式下的飞行空域。Then, according to the different maneuvering modes and different flight states of the target, the dynamic principle is used to obtain multiple characteristic trajectory sample data, and the multi-model modeling method is used to decompose the maneuvering state of the target to obtain the typical maneuvering local model of the target and the typical maneuvering local model. A set of target maneuvering feature models is formed. In this embodiment, the state of each dimension in the d-dimensional constraint space is defined as p ^j , j=1, 2, .

Then, the target maneuvering feature model set Pc is established as follows:

By solving the elements covering the constraint space, the maneuvering feature model set covering the target trajectory space can be obtained. In this embodiment, the re-entry height is selected as 72 km, the re-entry speed is 4000 m/s, and the terminal height constraint is 10 km. When the longitudinal flight distance changes, offline trajectory optimization is performed for different longitudinal flight distance constraints to form a trajectory set covering the flight range. As shown in Figure 3, the maneuvering feature model set composed of typical maneuvering local models can cover the flight airspace of the target under different maneuvering modes.

式中的e_i(k)表示目标机动特征模型集中第i个典型机动局部模型在时刻k的轨迹偏差，M为目标机动特性模型集中典型机动局部模型的数量，exp表示指数运算，a₀和a₁分别为当前误差权重因子和历史误差权重因子，a₀和a₁均大于0，决定了当前误差和历史误差在性能指标中的权重，用于确定当前时刻误差和过去时刻误差组合对性能指标的相对重要程度，τ＞0为遗忘因子，表示性能指标的记忆效应，h为所选取的有限时域长度，

为第i个典型机动局部模型在时刻k的轨迹状态，y(k)为在时刻k目标的实际飞行轨迹状态，rd^l表示计算欧式距离；利用当前目标的轨迹跟踪数据与飞行过程历史信息构成性能指标，通过性能指标函数来评价机动特征模型集中的各典型机动局部模型与目标实际飞行轨迹(目标当前运动状态)的匹配度，性能指标值J_i(k)越大，说明第i个典型机动局部模型与实际目标状态的匹配度越低，基于性能指标值J_i(k)，用典型机动局部模型的权值

来表示典型机动局部模型与目标实际飞行轨迹的匹配度，w_i(k)为第i个典型机动局部模型与目标实际飞行轨迹的匹配度。After establishing the target maneuver characteristic model set, the forward trajectory prediction module performs forward trajectory prediction. The forward trajectory prediction module defines the performance index function and establishes the target trajectory prediction model. In this embodiment, the tracking error weighting method is used to define the performance index function, and the obtained performance index function is:

where e _i (k) represents the trajectory deviation of the ith typical maneuvering local model in the target maneuvering feature model set at time k, M is the number of typical maneuvering local models in the target maneuvering feature model set, exp represents the exponential operation, a ₀ and a ₁ is the weight factor of the current error and the weight factor of the historical error respectively, a ₀ and a ₁ are both greater than 0, which determine the weight of the current error and the historical error in the performance index, and are used to determine the combination of the current time error and the past time error on the performance The relative importance of the index, τ>0 is the forgetting factor, indicating the memory effect of the performance index, h is the selected finite time domain length,

is the trajectory state of the i-th typical maneuvering local model at time k, y(k) is the actual flight trajectory state of the target at time k, rd ^l represents the calculated Euclidean distance; the trajectory tracking data of the current target and the historical information of the flight process are used to form The performance index is used to evaluate the matching degree of each typical _maneuvering local model in the maneuver feature model set and the actual flight trajectory of the target (the current motion state of the target) through the performance index function. The lower the matching degree of the maneuvering local model and the actual target state is, based on the performance index value J _i (k), the weight of the typical maneuvering local model is used.

to represent the matching degree between the typical maneuvering local model and the actual flight trajectory of the target, w _i (k) is the matching degree between the ith typical maneuvering local model and the actual flight trajectory of the target.

式中的v为当地速度，γ为当地航迹倾角，χ为当地航迹偏角，r为目标地心距离，θ为纬度，

为经度，ρ为目标所在大气环境下的空气密度，g为目标所在位置的重力加速度，β为倾侧角，

s为受力参考面积，C_D和C_L分别为目标的阻力系数和升力系数，m为目标质量。This embodiment simplifies the motion model in the half-velocity coordinate system of the re-entry gliding target, and the established target trajectory prediction model is:

where v is the local speed, γ is the local track inclination, χ is the local track declination, r is the target geocentric distance, θ is the latitude,

is the longitude, ρ is the air density in the atmospheric environment where the target is located, g is the gravitational acceleration at the location of the target, β is the tilt angle,

_s is the force reference area, CD and _CL are the drag coefficient and lift coefficient of the target, respectively, and m is the target mass.

Finally, use the performance index function to calculate the matching degree _wi (k) of the typical maneuvering local model and the actual flight trajectory of the target, and use the weighted least squares method to fit the parameters of the target trajectory prediction model (K _D , K according to _wi (k) _L , β), the parameters obtained by fitting are substituted into the target trajectory prediction model, and the fourth-order "Runge-Kutta" numerical integration method is used to obtain the forward predicted trajectory of the target.

The backward feedback correction module is used to correct the trajectory predicted by the forward trajectory prediction module. First, the multi-model "decomposition-combination" method is used to establish the global importance map model. The multi-model "decomposition-combination" method is a known method and is not an innovation of the present invention, and will not be repeated here. In this embodiment, it is assumed that there are 4 important areas (flight important nodes) and they obey the Gaussian distribution. On the basis of calculating the local distribution of each important flight node according to the Gaussian distribution, the joint distribution probability function f _m can be used to form a global importance map model, as shown in Figure 4.

形成全局重要性地图模型，式中的x_m0为重要性地图中第m个飞行重要节点中心点位置，σ_m为第m个飞行重要节点覆盖范围，x表示全局重要性地图模型中其它点的位置。Secondly, the forward trajectory predicted by the target trajectory prediction model is mapped to the global importance map model, and the target maneuver intention information is determined through the relationship between the target forward trajectory prediction area and the global importance map model area under the feature constraints, as shown in the figure. 5 shown. The global importance map model is used to evaluate the target maneuvering intention. The important area (flight important node) in this embodiment obeys the Gaussian distribution, and the joint distribution probability function is used.

A global importance map model is formed, where x _m0 is the center point position of the m-th important flight node in the importance map, σ _m is the coverage of the m-th important flight node, and x represents the distance of other points in the global importance map model. Location.

式中的w_p1和w_p2分别为目标前向轨迹预测区域的两个空间分布的后验概率，x₁、x₂、x₃、x₄为飞行重要节点分布范围(见图5)；然后根据预测轨迹的后验概率，用贝叶斯推理方法确定各典型机动局部模型的条件概率，即各典型机动局部模型所描述的目标机动意图信息为：

式中的p(x_k-1＝j)为k-1时刻第j个典型机动局部模型(j＝1,2,…,M)的先验概率，典型机动局部模型的先验概率为典型机动局部模型与目标实际飞行轨迹的匹配度，在本实施例中典型机动局部模型的先验概率是典型机动局部模型的权值w_i(k)，p(x_k＝i,x_k-1＝j)为典型机动局部模型对应的预测轨迹后验概率，也就是预测轨迹的后验概率w_a；基于贝叶斯预测原理，利用目标机动意图信息p(x_k＝i|x_k-1＝j)对目标预测状态进行修正，则有：p(x_k|x_1:k-1)＝∫p(x_k|x_k-1)p(x_k-1|x_1:k-1)dx_k-1，其中p(x_k-1|x_1:k-1)为k-1时刻的条件概率，也就是k-1时刻典型机动局部模型的目标机动意图信息，再根据修正结果p(x_k|x_1:k-1)对前向轨迹预测模块建立的目标轨迹预测模型的参数进行修正，有：

其中，

为目标机动特征模型集中的典型机动局部模型的参数(本实施例中典型机动局部模型的参数包括K_D，K_L，β)，M为目标机动特征模型集中典型机动局部模型的数量，在此基础上，利用参数修正后的目标轨迹预测模型进行轨迹预测，得到修正后的轨迹预测结果。The spatial distribution of the target forward trajectory prediction area obtained according to the target trajectory prediction model in this embodiment is: f _p1 (x) and f _p2 (x), and the posterior probability _wa of the corresponding obtained predicted trajectory is:

In the formula, w _p1 and w _p2 are the posterior probabilities of the two spatial distributions of the target forward trajectory prediction area, respectively, and x ₁ , x ₂ , x ₃ , and x ₄ are the distribution ranges of important flight nodes (see Figure 5); then According to the posterior probability of the predicted trajectory, the Bayesian inference method is used to determine the conditional probability of each typical maneuvering local model, that is, the target maneuvering intention information described by each typical maneuvering local model is:

where p(x _k-1 =j) is the prior probability of the jth typical maneuvering local model (j=1,2,...,M) at time k-1, and the prior probability of the typical maneuvering partial model is The matching degree between the maneuvering local model and the actual flight trajectory of the target. In this embodiment, the prior probability of the typical maneuvering partial model is the weight w _i (k) of the typical maneuvering partial model, p(x _k =i, x _k-1 =j) is the posterior probability of the predicted trajectory corresponding to the typical maneuvering local model, that is, the posterior probability _wa of the predicted trajectory; based on the Bayesian prediction principle, the target maneuvering intention information p(x _k =i|x _k-1 =j) modify the target prediction state, then there are: p(x _k |x _1:k-1 )=∫p(x _k |x _k-1 )p(x _k-1 |x _1:k-1 )dx _k-1 , where p(x _k-1 |x _1:k-1 ) is the conditional probability at time k-1, that is, the target maneuver intention information of the typical maneuvering local model at time k-1, and then according to the correction result p(x _k |x _1:k-1 ) modifies the parameters of the target trajectory prediction model established by the forward trajectory prediction module, as follows:

in,

is the parameter of the typical maneuvering local model in the target maneuvering feature model set (in this embodiment, the parameters of the typical maneuvering partial model include K _D , K _L , β ), M is the number of typical maneuvering partial models in the target maneuvering feature model set, here On the basis, the target trajectory prediction model after parameter correction is used for trajectory prediction, and the corrected trajectory prediction result is obtained.

Fig. 6 to Fig. 10 are the effect diagrams of using the revised target trajectory prediction model to predict the trajectory of the re-entry gliding target. The change curve of the prediction error in Fig. 6 shows that the method of the present invention has better prediction in the 200-second target prediction. precision. The variation curve of the model membership relationship in Fig. 7 shows that the correlation of the local feature models converges during the prediction process. The comparison curves of the model parameters in Fig. 8 to Fig. 10 show that the prediction of the model parameters is convergent in the prediction process, and has good prediction performance at the same time.

The present invention also provides a target trajectory prediction method, comprising the following steps:

S1. Establish a target maneuvering feature model set;

Firstly, the maneuvering feature model of the target is determined, then the maneuvering state of the target is decomposed, and a typical maneuvering local model of the target is established based on the maneuvering feature model, and the typical maneuvering partial model constitutes the target maneuvering feature model set;

S2, target forward trajectory prediction;

Define the performance index function and establish the target trajectory prediction model. The performance index function is used to evaluate the matching degree of each typical maneuvering local model in the target maneuvering feature model set and the actual flight trajectory of the target. According to each typical maneuvering local model and the actual flight trajectory of the target The matching degree of the target trajectory prediction model determines the parameters of the target trajectory prediction model, and the obtained parameters are substituted into the target trajectory prediction model, and the numerical integration method is used to use the target trajectory prediction model to predict the forward trajectory of the target;

S3, target trajectory correction;

Firstly, a global importance map model of important flight nodes is established. According to the global importance map model, the relationship between the forward predicted trajectory and the flight important nodes in the global importance map model is established to determine the target maneuvering intention information; using the target maneuvering intention information, The parameters of the target trajectory prediction model are modified by the posterior conditional probability method, and then the target trajectory is predicted according to the target trajectory prediction model after parameter correction, and the modified trajectory prediction result is obtained.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention in any form. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Technical personnel, within the scope of the technical solution of the present invention, can make some changes or modifications to equivalent examples of equivalent changes by using the technical content disclosed above, but any content that does not depart from the technical solution of the present invention, according to the present invention. The technical essence of the invention Any simple modifications, equivalent changes and modifications made to the above embodiments still fall within the scope of the technical solutions of the present invention.

Claims (8)

1. A target trajectory prediction system, comprising:

the system comprises a target maneuvering characteristic model set module used for establishing a target maneuvering characteristic model set, wherein the target maneuvering characteristic set comprises typical maneuvering local models of a plurality of targets;

the forward track prediction module is used for performing forward track prediction on the target, defining a performance index function and establishing a target track prediction model, calculating the matching degree of each typical maneuvering local model in a target maneuvering characteristic model set and the target actual flight track through the performance index function, determining parameters of the target track prediction model according to the matching degree of each typical maneuvering local model and the target actual flight track, substituting the obtained parameters into the target track prediction model, and performing forward track prediction on the target by adopting a numerical integration method;

the backward feedback correction module is used for correcting the target track, a global importance map model is established according to flight important nodes, then an incidence relation between a forward prediction track and the flight important nodes is established according to the global importance map model, target maneuver intention information is determined, parameters of the target track prediction model are corrected by the target maneuver intention information through a posterior conditional probability method, track prediction is carried out according to the target track prediction model after the parameters are corrected, and a corrected track prediction result is obtained.

2. The target trajectory prediction system of claim 1, wherein: when the target maneuvering characteristic model set module establishes a target maneuvering characteristic set, firstly, a maneuvering characteristic model for representing a target maneuvering mode is determined, then, maneuvering state decomposition is carried out on a target, a typical maneuvering local model of the target is established, and the obtained typical maneuvering local model forms a target maneuvering characteristic model set.

3. The target trajectory prediction system of claim 2, wherein: the maneuvering characteristic model comprises a CA model, a CV model, a typical track data model and a dynamic model.

4. A target trajectory prediction method is characterized by comprising the following steps:

s1, establishing a target maneuvering characteristic model set, firstly determining a maneuvering characteristic model of a target, then carrying out maneuvering state decomposition on the target, establishing a typical maneuvering local model of the target based on the maneuvering characteristic model, and forming the target maneuvering characteristic model set by the typical maneuvering local model;

s2, predicting a target forward track, defining a performance index function, establishing a target track prediction model, evaluating the matching degree of each typical maneuvering local model in the target maneuvering characteristic model set and the target actual flight track through the performance index function, determining parameters of the target track prediction model according to the matching degree of each typical maneuvering local model and the target actual flight track, and predicting the forward track of the target based on the target track prediction model by adopting a numerical integration method;

s3, correcting the target track, namely firstly determining flight important nodes, then establishing a global importance map model according to the flight important nodes, then establishing an incidence relation between a forward predicted track and the flight important nodes according to the global importance map model, determining target maneuver intention information, correcting parameters of the target track prediction model by using the target maneuver intention information and adopting a posterior conditional probability method, and predicting the target track according to the target track prediction model after parameter correction to obtain a corrected track prediction result.

5. The target trajectory prediction method according to claim 4, characterized in that: in step S2, parameters of the target trajectory prediction model are solved by a weighted least squares method.

6. The target trajectory prediction method according to claim 4, characterized in that: in step S2, a tracking error weighting method is adopted to define a performance index function, and the matching degree of the typical maneuvering local model and the target actual flight path is the weight of the typical maneuvering local model

Wherein J is_i(k) And the performance index value of the ith typical maneuvering local model in the target maneuvering characteristic model set is represented, and M is the number of the typical maneuvering local models in the target maneuvering characteristic model set.

7. The target trajectory prediction method according to claim 6, characterized in that: in step S3, a Bayesian inference method is adopted to determine target maneuver intention information

In the formula, p (x)_k-1J) is the prior probability of the j-th typical mobile local model at time k-1, p (x)_k＝i,x_k-1J) is the predicted trajectory posterior probability corresponding to a typical maneuver local model.

8. The target trajectory prediction method according to claim 7, characterized in that: the method for correcting the parameters of the target trajectory prediction model in step S3 is as follows: eyes of userTarget movement intention information p (x)_k＝i|x_k-1J) correcting the target predicted state, there are: p (x)_k|x_1:k-1)＝∫p(x_k|x_k-1)p(x_k-1|x_1:k-1)dx_k-1Then based on the corrected result p (x)_k|x_1:k-1) Parameters of the target track prediction model are corrected, and the parameters comprise:

wherein,

parameters of a typical maneuvering local model in the target maneuvering characteristic model set.

Images