Control and tracking optimization method for destination-oriented intention target

Abstract

The invention discloses a control and tracking optimization method for a destination guiding intention target. The control and tracking optimization method comprises the following steps of: 1, appointing an expected arrival place and expected arrival time of a controlled object; 2, constructing a destination limited model by using a constraint equation; 3, determining the motion type and motion parameters of the controlled object, and constructing a motion state space equation; and 4, according to the constructed motion state space equation, the controlled object autonomously plans and moves to the target location along the optimal path. According to the method, the accuracy of the controlled object arriving at the destination can be remarkably improved, related intention information can be effectively obtained, key auxiliary information is provided for upper-layer decision making, and potential threats can be accurately and timely coped with.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent control, and particularly relates to a control and tracking optimization method for destination-oriented intention targets. Background Art

[0002] In many application fields of the physical world, purpose-driven behaviors and end-point oriented movements are common phenomena. For example, in space technology, the rocket recovery process is a typical example. The rocket can fly precisely towards the predetermined landing point. In the field of unmanned aerial vehicle technology, the autonomous path planning ability enables unmanned aerial vehicles to adopt specific maneuvering strategies and accurately reach the target location. The end points pointed to by these specific behaviors essentially reveal the movement intentions of the targets. In a collaborative scenario, this intention information can be used as valuable control information input to effectively guide the movement of cooperative targets. In a non-cooperative scenario, it can be used as predictive information to help us infer the movement state of non-cooperative targets. As time goes by, the uncertainty of the target movement state will gradually decrease.

[0003] In a collaborative scenario, taking the rocket recovery mission as an example, the key to success focuses on whether the rocket can land at the predetermined position with an accurate speed of less than 2 m / s and a reasonable attitude at the specified moment. To achieve this goal, the control algorithm of the rocket plays a crucial role. The algorithm needs to continuously and real-time analyze and output accurate information to finely adjust the attitude and power output of the engine. Specifically, by integrating the movement intention information of the target (such as the expected landing point, speed requirements, etc.), the algorithm can predictively adjust the flight trajectory of the rocket to ensure that it can still fly precisely along the predetermined path in a complex flight environment and land smoothly at the required low speed in the final stage. This control method embedded with intention information significantly improves the accuracy and efficiency of rocket movement, making the recovery mission safer and more reliable.

[0004] In a non-cooperative scenario, when facing a swarm of unmanned aerial vehicles that track and strike or reconnoiter our important facilities, the situation becomes more complex. Since we cannot predict the specific action plans and actual arrival times of each unmanned aerial vehicle, we must rely on real-time data analysis and prediction models to deal with it. The specific approach is to combine the measured data of sensors such as radar and optics received, as well as prior knowledge such as the known end position, and use advanced algorithms to dynamically estimate the potential arrival time of the unmanned aerial vehicle. This process not only relies on precise data fusion technology but also uses machine learning algorithms to continuously optimize the prediction model, so as to more accurately estimate the current movement state of the unmanned aerial vehicle and predict its possible future action path in advance. Such prediction ability is crucial for formulating effective defense strategies and timely avoiding potential threats.

[0005] The methods commonly used in the prior art mainly rely on traditional tracking and prediction algorithms. However, when faced with complex, ever-changing, and highly dynamic environments, these methods often suffer from problems such as insufficient accuracy and response lag. Especially in non-cooperative scenarios, due to the lack of sufficient information input and efficient algorithm support, it is difficult to accurately predict and promptly respond to the behavior of unmanned aerial vehicle (UAV) swarms. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide, in view of the deficiencies in the above-mentioned prior art, an optimization method for controlling and tracking destination-oriented intent targets, which can significantly improve the accuracy of the controlled object reaching the destination, effectively obtain relevant intent information, provide key auxiliary information for upper-layer decision-making, and help accurately and promptly respond to potential threats.

[0007] To solve the above technical problem, the technical solution adopted by the present invention is: an optimization method for controlling and tracking destination-oriented intent targets, comprising the following steps:

[0008] Step 1: Specify the desired arrival location and expected arrival time of the controlled object;

[0009] Step 2: Construct a destination-constrained model using constraint equations;

[0010] Step 3: Identify the motion type and motion parameters of the controlled object, and construct a motion state space equation;

[0011] Step 4: Based on the constructed motion state space equation, the controlled object autonomously plans and moves along the optimal path towards the target location.

[0012] In the above-mentioned optimization method for controlling and tracking destination-oriented intent targets, after Step 4, there is further included Step 5: Based on the motion state space equation constructed in Step 3, perform non-cooperative target tracking and intent information estimation.

[0013] In the above-mentioned optimization method for controlling and tracking destination-oriented intent targets, the specification of the desired arrival location and expected arrival time of the controlled object in Step 1 is carried out through constraint conditions, and the constraint conditions include end constraints or process constraints, and the forms of the constraint conditions are linear equality constraints, linear inequality constraints, non-linear equality constraints, or non-linear inequality constraints.

[0014] In the above-mentioned optimization method for controlling and tracking destination-oriented intent targets, when the constraint condition in Step 1 is an end constraint, and the form of the constraint condition is a linear equality constraint, it is represented by a constraint function as:

[0015]

[0016] Wherein, is a constraint function, indicating that at the arrival time t D the controlled object must reach the desired arrival location b; represents the state of the controlled object at time t, and (x, y, z) represents the position, represents the three velocity components; x(t D ) represents the state of the controlled object at the arrival time t D ; A is a coefficient matrix with full row rank.

[0017] For the above control and tracking optimization method for destination-oriented intent targets, the specific process of using the constraint equation to construct the destination-constrained model in step 2 is as follows:

[0018] Step 201: Select the stochastic differential equation:

[0019]

[0020] is the constraint equation; where, x o (t) represents the relaxed state at a certain moment, represents the differential of x o (t), represents a Gaussian white noise process that is not related to the initial state x o (t0), represents a Gaussian process, q is the intensity matrix, δ(·) is the Dirac delta function, t represents time, t′ represents a moment, B(t) is the drift function in the Wiener process (also known as Brownian motion), Γ(t) is the jump function in the Wiener process (also known as Brownian motion), and both B(t) and Γ(t) are known functions;

[0021] Step 202: Using Ito's lemma, the solution of the above differential equation is obtained as:

[0022]

[0023] where, represents the relaxed state at the (k - 1)th moment; represents the state transition matrix, which is used to describe the deterministic state evolution law from time t k-1 to time t k ; t k represents the kth moment, t k-1 represents the (k - 1)th moment; represents the cumulative noise from time t k-1 to time t k ; τ represents the integration vector of the integrand;

[0024] Step 203: Establish the extended state matrix Under this stochastic differential equation, the relaxed constraint equation is:

[0025]

[0026] Among them, A = [0 A], which is orthogonally decomposed into two uncorrelated terms by Gram - Schmidt orthogonal decomposition, and we get:

[0027]

[0028] Among them,

[0029]

[0030] represents the state transition function from time t k to the end time t D ;

[0031] Substitute it into the after orthogonal decomposition, and expand this extended matrix, we get:

[0032]

[0033] Substitute into the above formula, we get:

[0034]

[0035] In this relaxed state, replace the relaxed constraint y with the strong constraint b as:

[0036]

[0037] Among them,

[0038]

[0039] is the covariance of the set noise , and the set noise is a Gaussian distribution;

[0040] At this time, the construction of the destination - restricted model is completed.

[0041] In the above - mentioned control and tracking optimization method for destination - oriented intent targets, the motion types of the controlled object described in step three include the CV model, CA model, CT model, and NCV model.

[0042] In the above - mentioned control and tracking optimization method for destination - oriented intent targets, the motion type of the controlled object described in step three is the NCV model, and the motion parameters described in step three are:

[0043]

[0044] Among them, t D is the arrival time, T is the sampling interval, and q is the acceleration; Q k-1 is the covariance of the cumulative noise ω k-1 at time k - 1, is the covariance of the cumulative noise D from time k - 1 to the end time t ;

[0045] At this time, the construction of the state - space equation is completed.

[0046] In the above - mentioned control and tracking optimization method for destination - oriented intent targets, when performing non - cooperative target tracking and intent information estimation based on the motion state - space equation constructed in step three, the state of the non - cooperative target is estimated by an estimation algorithm, and its arrival position and arrival time are predicted.

[0047] In the above - mentioned control and tracking optimization method for destination - oriented intent targets, the estimation algorithm is the maximum a posteriori probability (MAP) estimation algorithm. When using the MAP estimation algorithm to estimate the state of the non - cooperative target, predict its arrival position and arrival time, the prior information of the non - cooperative target is fused with the current observation data to construct a posterior probability distribution model. By maximizing the posterior probability density, the target state estimation value that maximizes the posterior probability is obtained, which is expressed by the formula:

[0048]

[0049] The first equation represents the MAP estimate of the destination at time k (arrival location), and the MAP estimate will select the value that maximizes the conditional probability where Z k is the observation data or data set up to time k;

[0050] The second equation represents the MAP estimate of the arrival time t given the estimated destination k and the measurement data Z D , which combines the predicted destination and the available data to maximize the conditional probability of the arrival time;

[0051] The third equation represents the minimum mean - square error (MMSE) estimate of x of the state at time k given the estimated value of the destination , the estimated value of the arrival time k and the measurement data Z k , and the MMSE estimate is x kConditioned on available data and MAP estimates.

[0052] Compared with the prior art, the present invention has the following advantages:

[0053] 1. In terms of algorithm design, the present invention innovatively integrates the destination intention information of the controlled body directly into the modeling stage, which is in contrast to the post-processing method; this approach intuitively displays the state evolution, improves algorithm efficiency, and significantly reduces errors, thereby improving the accuracy of motion modeling; this accuracy ensures the efficiency and smoothness of the motion trajectory, provides a reliable model basis for subsequent control, solves the shortcomings of the prior art in post-processing intention information, and brings new breakthroughs to the field of motion control.

[0054] 2. In step three of the present invention, the motion type and motion parameters of the controlled object are clarified, and the motion state space equation is constructed. When applied in the rocket recovery scenario, the present invention has significant advantages and allows specific speed requirements to be set in the intention information. This feature makes the speed and attitude control during the rocket recovery process more precise, ensuring that the rocket lands at an appropriate speed at the predetermined time and place, improving the recovery reliability and safety, reducing risks and costs, and providing guarantees for the reuse and long-term operation of the rocket.

[0055] 3. In the cooperative target control application scenario of step 4, the present invention demonstrates strong flexibility, can construct a diversified product matrix according to different constraints, and provide rich strategy options; for example, in a drone attack mission, the present invention can design a control strategy based on the mission and environment to meet a variety of combat needs; this flexibility gives the present invention broad application prospects in the field of cooperative target control.

[0056] 4. In step five, the present invention can combine advanced models and filtering technologies to accurately estimate the arrival time and location of non-cooperative targets, support the acquisition of situational intelligence, and enhance the capabilities of the surveillance and early warning system; at the same time, it can provide key auxiliary information for upper-level decision-making, helping to accurately and timely respond to potential threats.

[0057] 5. The present invention can achieve accurate tracking and efficient prediction of targets such as drones by deeply integrating the target's motion intention information, real-time measurement data and advanced machine learning algorithms; it not only significantly improves the prediction accuracy and response speed, but also has strong adaptive capabilities, and can be flexibly adjusted according to different scenarios and needs, providing strong support for high-level automated decision-making such as situation awareness, conflict avoidance, opportunity discovery and optimal resource allocation.

[0058] In summary, the present invention can significantly improve the accuracy of the controlled object in reaching the destination, and can also effectively obtain relevant intention information, provide key auxiliary information for upper-level decision-making, and help to accurately and timely respond to potential threats.

[0059] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Description of the Drawings

[0060] Figure 1 is a flowchart of the method of the present invention;

[0061] Figure 2 is a schematic diagram of the rocket recovery scenario in the embodiment of the present invention;

[0062] Figure 3 is a schematic diagram of the scenario of a drone attacking a fixed target under 500 Monte Carlo experiments in the embodiment of the present invention;

[0063] Figure 4 is a schematic diagram of the non - cooperative target intention recognition in the embodiment of the present invention;

[0064] Figure 5 is a schematic diagram of the non - cooperative target situation assessment result in the embodiment of the present invention. Figure 6 is a schematic diagram of the non - cooperative target situation assessment result in Embodiment 2 of the present invention. Detailed Embodiment

[0065] Embodiment 1

[0066] As Figure 1 shown, the control and tracking optimization method for destination - oriented intention targets of the present invention includes the following steps:

[0067] Step 1: Specify the expected arrival location and expected arrival time of the controlled object;

[0068] In this embodiment, in Step 1, the specification of the expected arrival location and expected arrival time of the controlled object is carried out through constraint conditions. The constraint conditions include end - point constraints or process constraints, or other feature constraints. The form of the constraint conditions is linear equality constraints, linear inequality constraints, non - linear equality constraints, or non - linear inequality constraints, depending on the actual situation.

[0069] In this embodiment, the constraint condition in Step 1 is an end - point constraint, and the form of the constraint condition is a linear equality constraint, which is represented by a constraint function as:

[0070]

[0071] Among them, is the constraint function, indicating that at the arrival time t D the controlled object must reach the expected arrival location b; the arrival time t D is a continuous random variable with a known prior distribution; Denote the state of the controlled object at time \(t\), and \((x, y, z)\) represents the position. Denote the three velocity components; \(x(t\) D ) represents the state of the controlled object at the arrival time \(t\) D ; \(A\) is a coefficient matrix with full row rank.

[0072] In specific implementation, \(A\in R\) m×n (\(m\leq n\)) and \(b\in R\) m are known, \(R\) m×n represents a real matrix with \(m\times n\) rows, and \(R\) m represents a vector with \(m\) rows or \(m\) columns.

[0073] The above are the constraint conditions that the rocket to be recovered needs to meet taking rocket recovery as an example. The above is a linear equality constraint. Of course, this equation can be extended to non - linear and inequality forms according to specific situations.

[0074] The above destination constraint is not only imposed on the final state, but also affects the evolution of the entire state during the rocket flight; due to the interdependence of states at different times, therefore, a spatio - temporal correlation between vectors must be established. The Gaussian process (GP) prior is usually used to specify the correlation of the state vector over time. The time - varying dependence can be written as:

[0075]

[0076] where and \(K(t, t')=\text{cov}(x(t, t'))\) represent the mean function and covariance function of the rocket at time \(t\) respectively. Compared with the stochastic difference equation, the correlation defined by the above GP model is more general because it specifies the correlation between state vectors at any time.

[0077] Step 2: Construct a destination - restricted model using the constraint equation;

[0078] In this embodiment, the specific process of constructing the destination - restricted model using the constraint equation in Step 2 is as follows:

[0079] Step 201: Select the stochastic differential equation:

[0080]

[0081] as the constraint equation; where \(x\) o (t) represents the relaxed state at a certain moment, represents the differential of \(x\) o (t), represents a Gaussian white noise process that is uncorrelated with the initial state \(x\) o (t0), denotes a Gaussian process, \(q\) is the intensity matrix, \(\delta(\cdot)\) is the Dirac delta function, \(t\) represents time, \(t'\) represents a moment, \(B(t)\) is the drift function in the Wiener process (also known as Brownian motion), \(\Gamma(t)\) is the jump function in the Wiener process (also known as Brownian motion), and both \(B(t)\) and \(\Gamma(t)\) are known functions;

[0082] Step 202: Using Ito's lemma, the solution to the above differential equation is obtained as follows:

[0083]

[0084] where, denotes the relaxed state at time \(k - 1\); denotes the state transition matrix, which is used to describe the deterministic state evolution law from time \(t\) k-1 to time \(t\) k ; \(t\) k denotes time \(k\), \(t\) k-1 denotes time \(k - 1\); denotes from time \(t\) k-1 to time \(t\) k the cumulative noise; \(\tau\) represents the integration vector of the integrand;

[0085] Step 203: Establish the extended state matrix Under this stochastic differential equation (SDE), the relaxation constraint equation is:

[0086]

[0087] where \(A = [0\ A]\), which is orthogonally decomposed into two uncorrelated terms by Gram - Schmidt orthogonal decomposition, and we get:

[0088]

[0089] where,

[0090]

[0091]

[0092] denotes the state transition function from time \(t\) k to the end time \(t\) D ;

[0093] Substitute the orthogonally decomposed into it, and expand this extended matrix, we get:

[0094]

[0095] Substitute Substituting the above equation, we get:

[0096]

[0097] where, is a special representation form at the end time t D and still represents in the form of at other times;

[0098] In this relaxation state, replacing the relaxation constraint y with the strong constraint b gives:

[0099]

[0100] where,

[0101]

[0102] is the covariance of the set noise and the set noise is Gaussian distributed;

[0103] At this point, the construction of the destination - restricted model is completed.

[0104] Step 3: Define the motion type and its motion parameters of the controlled object, and construct the motion state - space equation;

[0105] In this embodiment, the motion types of the controlled object described in Step 3 include the CV model, CA model, CT model, and NCV model (nearly constant - velocity model).

[0106] In this embodiment, the motion type of the controlled object described in Step 3 is the NCV model (nearly constant - velocity model), and the motion parameters described in Step 3 are:

[0107]

[0108]

[0109] where, t D is the arrival time, T is the sampling interval, q is the acceleration; Q k-1 is the covariance of the cumulative noise ω k-1 at time k - 1, is the covariance of the cumulative noise D from time k - 1 to the end time t ;

[0110] For the specific setting of the constraint equation, not only can it be constrained to fly to a specific position at the arrival time, but also its arrival speed can be constrained at this moment:

[0111]

[0112] At this point, the construction of the state-space equation is completed.

[0113] In the state-space equation constructed in Step 3, the functions and are full-rank, which ensures that the system can be effectively controlled by using this discrete equation as the control equation of the dynamic system; with appropriate inputs, the system state can be driven to transition to the target state; these control instructions are transmitted to the hardware execution unit of the controlled object through a data bus or a dedicated communication interface, and the execution unit converts these instructions into actual physical actions, thereby achieving the desired control objective.

[0114] In specific implementation, based on the motion state-space equation constructed in Step 3, which serves as the core of the control system, the controlled object can autonomously plan a path through this motion state-space equation. Specifically, the controlled object uses the motion state-space equation to predict its future state in real time and plans an optimal path from the current position to the target location according to the prediction results; this path not only considers the reachability of the target but also ensures the stability and efficiency of the motion by optimizing the smoothness of the path and the variation of the control input.

[0115] During the execution of autonomous planning, the controlled object must continuously update its state information (such as position, velocity, acceleration, etc.) and adjust the control input according to this information and the position of the target location. This process generally relies on the structure of a closed-loop control system, where sensors provide real-time measurements of the current state, the controller calculates the control input to be applied based on these measurements, and the actuator is responsible for implementing these inputs to adjust the motion trajectory of the system.

[0116] Step 4: According to the constructed motion state-space equation, the controlled object autonomously plans and moves along the optimal path towards the target location.

[0117] To verify the technical effects that the present invention can produce, taking the controlled object as a rocket as an example, assume that the rocket recovery process is to descend from a high altitude at the position [0, 8000, 10000], and the landing point is [1000, 0, 0]. Now it is required to complete the following during the recovery process: 1> Descend to the specified position height of 300 m, that is, perform attitude adjustment at the position [1000, 0, 300] within 40 s, and require the three velocity vectors to be [100 m / s, -8 m / s, -10 m / s] respectively; 2> Descend to the specified position within 10 s, and require the velocity to reach [0 m / s, 0 m / s, -0.05 m / s]; The schematic diagram of the rocket recovery scenario is as Figure 2 shown.

[0118] By observing Figure 2From the multiple sub - figures, we can clearly see key information such as the relative position, speed, and attitude of the rocket at different time points. Especially at the critical moment when there are only 10 seconds left until the landing target point, the rocket starts to adjust its attitude. The main objective of this adjustment is to ensure that the rocket can reach the predetermined landing position vertically and precisely at a stable speed of less than 2 m / s.

[0119] To further verify the technical effects that the present invention can produce, taking an unmanned aerial vehicle (UAV) as the controlled object as an example, the UAV starts from the starting position [0, 0] with the goal of attacking another designated position [1000, 1000]. Different from the equality constraints adopted in the rocket example, the constraint conditions in this example follow the normal distribution law, and are specifically expressed as Due to this difference, the derived state - space equation also shows different characteristics compared with the rocket example.

[0120] To visually display the movement path of the UAV under the given constraint conditions, we conducted 500 Monte Carlo experiments. Figure 3 A schematic diagram of the UAV attacking a fixed target scenario under 500 Monte Carlo experiments is detailedly recorded, where the ellipse composed of "☆" is the intended destination of the UAV. Through this series of simulation experiments, we can more clearly understand how the UAV plans its flight path under the normal - distribution constraint to efficiently reach the attack target.

[0121] Embodiment 2

[0122] The difference between this embodiment and Embodiment 1 is that after Step 4, it further includes Step 5: Based on the motion state - space equation constructed in Step 3, non - cooperative target tracking and intention information estimation are carried out.

[0123] All the remaining steps are the same as those in Embodiment 1.

[0124] In this embodiment, when carrying out non - cooperative target tracking and intention information estimation based on the motion state - space equation constructed in Step 3 in Step 5, an estimation algorithm is used to estimate the state of the non - cooperative target, predict its arrival position and arrival time.

[0125] The motion state space equation constructed in Step 3 not only lays a solid foundation for the control of cooperative targets but also applies equally to the tracking tasks of non-cooperative targets. When facing non-cooperative targets, the key lies in how to accurately estimate the target's state, predict its arrival position and arrival time. In this process, estimation algorithms can be used, including the Minimum Mean Square Error Estimation Algorithm (MMSE algorithm), the Maximum A Posteriori Probability Estimation Algorithm (MAP algorithm), as well as Kalman filtering, particle filtering, etc., to perform precise state prediction and tracking of the target. The core idea of these algorithms is to continuously update the state estimation of the target through measurement results and the previous state space model, and combine it with the observed data to improve the accuracy and reliability of the estimation.

[0126] In this embodiment, the estimation algorithm is the Maximum A Posteriori Probability Estimation Algorithm (MAP algorithm). When using the Maximum A Posteriori Probability Estimation Algorithm (MAP algorithm) to estimate the state of a non-cooperative target, predict its arrival position and arrival time, the prior information of the non-cooperative target is fused with the current observed data to construct a posterior probability distribution model. By maximizing the posterior probability density, the target state estimation value that maximizes the posterior probability is obtained, which is expressed by the formula:

[0127]

[0128] The first equation represents the MAP estimate of the destination at time k (arrival location), and the MAP estimate will select the value that maximizes the conditional probability where Z k is the observed data or data set up to time k;

[0129] The second equation represents the MAP estimate of the arrival time t given the estimated destination k and the measurement data Z D , which combines the predicted destination and the available data to maximize the conditional probability of the arrival time;

[0130] The third equation represents the MMSE estimate of the x of the state at time k given the estimated value of the destination , the estimated value of the arrival time k and the measurement data Z k , and the MMSE estimate is the conditional probability of x k based on the existing data and the MAP estimate.

[0131] The MAP estimation method is used to find the most likely value of the parameter (such as the destination or arrival time) based on the observation results; the MMSE estimation method calculates the expected value of the state variable to minimize the mean square error of the estimate.

[0132] This process combines the historical information of the target and real-time observation data, thereby improving the tracking accuracy of non-cooperative targets.

[0133] To further verify the technical effects that the present invention can produce, target tracking was carried out for non-cooperative targets, and the Kalman filtering technology was integrated to realize the inference of the target intention, specifically covering two key elements: time of arrival and location of arrival; the schematic diagram of the non-cooperative target intention recognition obtained is as Figure 4 shown, and the schematic diagram of the non-cooperative target situation assessment result is as Figure 5 shown. By observing Figure 4 , we can find that in the initial stage, the estimation of the target intention may deviate from the actual situation. However, with the continuous incorporation and utilization of subsequent measurement information, our estimation of the target intention gradually becomes more accurate. Further, Figure 5 intuitively shows the situation assessment result constituted by the speed of the target, the current position, and the comprehensively quantified intention inference information at a specific moment.

[0134] As described above, it is only a preferred embodiment of the present invention and does not impose any limitation on the present invention. Any simple modification, change, and equivalent structural change made to the above embodiments according to the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A control and tracking optimization method for destination-oriented intent targets, characterized in that The method includes the following steps: Step 1, specifying the expected arrival location and expected arrival time of the controlled object; Step 2, constructing a destination-constrained model by using constraint equations; Step 3, clarifying the motion type and its motion parameters of the controlled object, and constructing a motion state space equation; Step 4, according to the constructed motion state space equation, the controlled object autonomously plans and moves along the optimal path to the target location.

2. The control and tracking optimization method for a destination-oriented intention target according to claim 1, wherein: After Step 4, it further includes Step 5, according to the motion state space equation constructed in Step 3, performing non-cooperative target tracking and intention information estimation.

3. A control and tracking optimization method for a destination-oriented intention target according to claim 1, characterized in that: In Step 1, the specification of the expected arrival location and expected arrival time of the controlled object is carried out through constraint conditions, and the constraint conditions include end constraints or process constraints, and the forms of the constraint conditions are linear equality constraints, linear inequality constraints, non-linear equality constraints or non-linear inequality constraints.

4. A control and tracking optimization method for a destination-oriented intention target according to claim 3, characterized in that: In Step 1, the constraint condition is an end constraint, and the form of the constraint condition is a linear equality constraint, which is represented by a constraint function as: Among them, is a constraint function, indicating that at the arrival time t D the controlled object must reach the desired arrival location b; represents the state of the controlled object at time t, (x, y, z) represents the position, represents three velocity components; x(t D ) represents the state of the controlled object at the arrival time t D ; A is a coefficient matrix with full row rank.

5. A control and tracking optimization method for a destination-oriented intention target according to claim 2, characterized in that: The specific process of constructing a destination-constrained model by using constraint equations in Step 2 is as follows: Step 201, selecting a stochastic differential equation: is a constraint equation; where, x o (t) represents the relaxed state at a certain moment, represents the differential of x o (t), represents a Gaussian white noise process that is not related to the initial state x o (t0), represents a Gaussian process, q is the intensity matrix, δ(·) is the Dirac delta function, t represents time, t′ represents a moment, B(t) is the drift function in the Wiener process, and Γ(t) is the jump function in the Wiener process; Step 202, using Ito's lemma, the solution of the above differential equation is obtained as: Among them, represents the relaxed state at time k - 1; represents the state transition matrix, which is used to describe the deterministic state evolution law from time t k-1 to time t k ; t k represents time k, t k-1 represents time k - 1; represents t k-1 from time to time t k cumulative noise; τ represents the integration vector of the integrand; Step 203: Establish an extended state matrix Under this stochastic differential equation, the relaxed constraint equation is as follows: where, A = [0 A], and through Gram-Schmidt orthogonal decomposition into two uncorrelated terms, we get: Among them, Σ D，k =(Σ k，D ) T Represents the description t k From the moment to the end moment t D The state transition function between; Substitute into the result of orthogonal decomposition and expand the extended matrix to obtain: Λ = A T (AA T ) -1 Substitute into the above formula, we get: In this relaxed state, replacing the relaxed constraint y with the strong constraint b gives: where, To aggregate noise of which the covariance, the aggregated noise is Gaussian distributed; At this time, the construction of the destination-constrained model is completed.

6. The control and tracking optimization method for a destination-oriented intention target according to claim 2, characterized in that: The motion types of the controlled object in Step 3 include the CV model, CA model, CT model and NCV model.

7. A method for optimizing control and tracking of a destination-oriented intention target according to claim 6, characterized in that: In Step 3, the motion type of the controlled object is the NCV model, and the motion parameters in Step 3 are: Among them, t D is the arrival time, T is the sampling interval, and q is the acceleration; Q k-1 is the covariance of the cumulative noise w k-1 at time k - 1, is the covariance of the cumulative noise D from time k - 1 to the end time t ; At this time, the construction of the state space equation is completed.

8. A method for optimizing control and tracking of a destination-oriented intention target according to claim 2, characterized in that: When performing non-cooperative target tracking and intention information estimation according to the motion state space equation constructed in Step 5 in Step 5, an estimation algorithm is used to estimate the state of the non-cooperative target, predict its arrival position and arrival time.

9. A control and tracking optimization method for a destination-oriented intention target according to claim 8, characterized in that: The estimation algorithm is the maximum a posteriori probability estimation algorithm. When using the maximum a posteriori probability estimation algorithm to estimate the state of the non-cooperative target, predict its arrival position and arrival time, the prior information of the non-cooperative target is fused with the current observation data to construct a posterior probability distribution model, and by maximizing the posterior probability density, the target state estimation value that maximizes the posterior probability is obtained, which is represented by the formula as: The first equation represents the MAP estimate of the destination at time k The MAP estimate will select a value that maximizes the conditional probability where Z k is the observed data or data set up to time k; The second equation represents the MAP estimate of the time of arrival t given the estimated destination and the measurement data Z k which combines the predicted destination and the available data to maximize the conditional probability of the time of arrival; D ​ The third equation represents the MMSE estimate of the state x at time k given the estimate, time of arrival estimate, and measurement data Z k and is conditioned on the available data and the MAP estimate. k The MMSE estimate is of x k conditioned on the available data and the MAP estimate.

Images