Multi-unmanned aerial vehicle tracking multi-target optimization method and system based on chaotic adaptive variation particle swarm

Abstract

The invention provides a multi-unmanned aerial vehicle tracking multi-target optimization method and system based on a chaotic adaptive variation particle swarm. The method comprises the steps of obtaining target position information and unmanned aerial vehicle state information; predicting a future trajectory of the target; the unmanned aerial vehicle performs probability locking on the tracking target; optimizing a control sequence of the unmanned aerial vehicle by using a chaos adaptive variation particle swarm optimization algorithm; enabling the multi-unmanned aerial vehicle control sequence to reach Nash equilibrium through multiple iterations, and obtaining an optimal control sequence; and controlling the multiple unmanned aerial vehicles to track multiple targets by using the optimal control sequence. According to the method, the problems of performance degradation when multiple unmanned aerial vehicles track random moving targets in a dynamic environment, collision avoidance in a distributed scene and tracking target number balance are effectively solved; nash optimization and chaos adaptive variation particle swarm optimization are combined, the tracking performance and cooperation efficiency of multiple unmanned aerial vehicles can be improved in a complex dynamic scene, the method is suitable for tracking tasks of targets in different motion modes, and the adaptability and robustness of an algorithm in a dynamic environment are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of inorganic multi-target tracking, and in particular to a multi-UAV multi-target tracking optimization method and system based on a chaotic adaptive mutation particle swarm. BACKGROUND

[0002] With the rapid development of UAV technology, the scenario of multiple UAVs cooperating to perform tasks is increasing, and the demand for multi-UAV tracking of multiple targets is increasing. In a complex dynamic environment, the target trajectory often exhibits randomness and uncertainty, and the multi-UAV system needs to achieve efficient cooperation and collision avoidance, which makes dynamic path planning and task allocation of multi-UAVs a key problem to be solved.

[0003] The problem of multi-UAV multi-target tracking has become an important research direction in the field of intelligent unmanned systems. With the progress of communication and sensing technology, information exchange and collaborative decision-making can be achieved among UAVs, and the entire tracking system constitutes a complex multi-agent system. Traditional static path planning and task allocation schemes are difficult to adapt to dynamically changing target trajectories and complex environmental constraints. For example, in scenarios where the target suddenly changes direction and the number of UAVs and targets dynamically adjusts, it is difficult to achieve efficient tracking and safe cooperation.

[0004] Current research on multi-UAV multi-target tracking methods is mostly based on intelligent optimization algorithms. In addition to traditional meta-heuristic algorithms, hybrid algorithms combining dynamic optimization and game theory provide new ideas for solving this problem and show effectiveness in dealing with dynamic environments and distributed decision-making. Although existing optimization methods achieve multi-target tracking by adjusting the control strategy of UAVs, they have improved the tracking performance of the system to some extent. However, there are still problems, such as the performance degradation of previous optimization solutions when facing dramatic changes in target trajectories; in distributed scenarios, it is difficult to achieve globally optimal collision avoidance due to the local observation capability of UAVs; and in task allocation, it is difficult to balance the number of UAVs tracking each target, leading to insufficient tracking accuracy for some targets or excessive system energy consumption. These problems make it difficult for traditional algorithms to generate tracking strategies that can be used in different dynamic scenarios and task sizes. SUMMARY

[0005] To overcome the deficiencies of the prior art, the present application provides a multi-UAV multi-target tracking optimization method and system based on a chaotic adaptive mutation particle swarm, which effectively solves the performance degradation of multi-UAV tracking of moving targets in dynamic environments, collision avoidance in distributed scenarios, and the balance of the number of tracking targets.

[0006] The technical solution of the present application is a multi-UAV multi-target tracking optimization method based on a chaotic adaptive mutation particle swarm, comprising the following steps:

[0007] S1), obtaining target position information and unmanned aerial vehicle state information;

[0008] S2), predicting target future trajectory by using extended Kalman filter algorithm;

[0009] S3), unmanned aerial vehicle probabilistic lock tracking target;

[0010] S4), using chaotic adaptive mutation particle swarm optimization algorithm to optimize unmanned aerial vehicle control sequence;

[0011] S5), through multiple iterations to make multi-unmanned aerial vehicle control sequence to Nash equilibrium, and the current obtained unmanned aerial vehicle control sequence as the optimal control sequence;

[0012] S6), using optimal control sequence to control multi-unmanned aerial vehicle tracking multi-target.

[0013] As preferred, in step S2), the obtained target position information and speed information are taken as the input of the extended Kalman filter algorithm to predict the target motion trajectory in the future N time domain length.

[0014] As preferred, in step S3), before optimization, the unmanned aerial vehicle randomly decides whether to lock the target tracking of the last round by comparing the random number with the probability threshold, when the random number is less than the probability threshold, the unmanned aerial vehicle locks the tracking of the current target, which remains unchanged in the optimization process; on the contrary, the unmanned aerial vehicle does not lock the tracking target.

[0015] As preferred, in step S4), the chaotic adaptive mutation particle swarm optimization algorithm is used to optimize the control sequence of the unmanned aerial vehicle, which specifically includes the following steps:

[0016] S41), the tracking target and control sequence of the unmanned aerial vehicle are combined into a particle;

[0017] S42), after each iteration, the fitness value of each particle in the particle swarm is evaluated, and whether the optimal fitness value of the particle swarm is degraded is detected;

[0018] S43), if the optimal fitness value is detected to be degraded, an adaptive strategy is adopted for adjustment, otherwise no change is made and the particle swarm is continued to be iteratively optimized.

[0019] As preferred, in step S42), after each iteration, the fitness value of each particle in the particle swarm is evaluated, and whether the optimal fitness value of the particle swarm is degraded is detected, which specifically includes the following steps:

[0020] S421), set the statistical period and the degradation threshold as the basic parameters for detection;

[0021] S422), in the initialization phase, collecting the optimal particle fitness value of each generation in the first statistical period, calculating the average value of these values, and taking the average value as the benchmark performance value for subsequent comparison;

[0022] S423), every time the data of a statistical period is accumulated, the average fitness value of the current period is calculated and compared with the benchmark performance value;

[0023] If the current average value is better than or equal to the benchmark performance value, the benchmark performance value is updated and the monitoring continues.

[0024] If the current average value is worse than the benchmark performance value and reaches the degradation threshold, it is determined that a change is detected, and the detection state is reset to initialization to prepare for detection after adaptive adjustment.

[0025] Preferably, in step S43), if the optimal fitness value is detected to be degraded, an adaptive strategy is adopted for adjustment, otherwise no change is made and the particle swarm is iteratively optimized, specifically:

[0026] When the degradation is detected, the position and speed of the particle are reinitialized, and the historical optimal solution record is emptied; and the motion inertia of the particle formed in the old environment is broken through chaotic disturbance, so that the particle more uniformly covers the solution space; when the change is detected, the mutation probability is enhanced and a fixed refractory period is maintained.

[0027] In the reinitialization phase after detecting the change, the inertia weight w is increased to enhance the global exploration ability of the particle swarm; when the system completes the initialization, the inertia weight w returns to the original value, at which time the search behavior of the particle swarm is biased towards local fine search.

[0028] Preferably, in step S5), the evaluation criteria of the multi-unmanned aerial vehicle system are set based on Nash equilibrium, and when the control sequence obtained by iteration satisfies Nash equilibrium, the iteration is ended, and the control sequence obtained at present is the final result.

[0029] Preferably, the application also provides a multi-unmanned aerial vehicle tracking multi-target optimization system based on chaotic adaptive mutation particle swarm, comprising:

[0030] A data acquisition module is configured to acquire target position information and unmanned aerial vehicle state information.

[0031] A trajectory prediction module is configured to predict the future trajectory of the target by using an extended Kalman filter algorithm.

[0032] A tracking target probability locking module is configured to randomly decide whether to lock the target tracking of the last round by comparing a random number with a probability threshold.

[0033] A control sequence optimization module is configured to optimize the control sequence of the unmanned aerial vehicle by using a chaotic adaptive mutation particle swarm optimization algorithm.

[0034] The iteration result judgment module is used to determine whether the control sequence of multiple UAVs has reached Nash equilibrium after multiple iterations. If so, the current UAV control sequence is taken as the optimal control sequence; otherwise, the target tracking probability locking module is called.

[0035] The control module is used to control multiple UAVs to track multiple targets according to the optimal control sequence.

[0036] The beneficial effects of this invention are as follows:

[0037] 1. This invention effectively solves the problems of performance degradation when multiple UAVs track randomly moving targets in dynamic environments, collision avoidance in distributed scenarios, and balancing the number of tracked targets;

[0038] 2. This invention combines Nash optimization with chaotic adaptive mutation particle swarm optimization, which can improve the tracking performance and collaborative efficiency of multiple UAVs in complex dynamic scenarios. It is applicable to tracking tasks of targets with different motion modes and enhances the adaptability and robustness of the algorithm in dynamic environments. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0040] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0041] Example 1

[0042] like Figure 1 As shown, this embodiment provides a multi-UAV tracking multi-target optimization method based on chaotic adaptive mutated particle swarm optimization, including the following steps:

[0043] S1) Obtain target location information and UAV status information;

[0044] S2) Predict the future trajectory of the target using the extended Kalman filter algorithm;

[0045] This embodiment uses the acquired target position and velocity information as input to the extended Kalman filter algorithm to predict the target's trajectory over the next N time domains; specifically as follows:

[0046] First, the system's nonlinear state equations are used to predict the state at the next time step:

[0047]

[0048] In the formula, denotes the state of the target at the next time, i.e. the position information of the target; f denotes the transition relation from the state at the current time to the state at the next time, i.e. the motion model of the target; u k denotes the control vector of the target motion, i.e. the velocity information of the target;

[0049] Then, the covariance matrix of the predicted state is calculated:

[0050]

[0051] In the formula, F k is the Jacobian matrix of the state transition matrix, P k|k denotes the current covariance matrix; Q k denotes the process noise covariance matrix; T denotes the transpose operation; the Jacobian matrix F k of the state transition matrix is:

[0052]

[0053] When there is a new observation value z k+1 , first, the observation residual is calculated:

[0054]

[0055] In the formula, z denotes the observation residual; h denotes the mapping relation from the state to the observation value;

[0056] Then, the covariance of the observation residual is calculated:

[0057]

[0058] In the formula, S k+1 denotes the covariance of the observation residual; H k+1 denotes the Jacobian matrix of the observation matrix; R k+1 denotes the observation noise covariance matrix;

[0059] In the formula, H k+1 is the Jacobian matrix of the observation matrix:

[0060]

[0061] Then, the Kalman gain K k+1 is calculated:

[0062]

[0063] Finally, the state estimation value and the covariance matrix are updated using the observation data:

[0064]

[0065] I represents a unit matrix;

[0066] The future N time domains are predicted by repeating the filtering and predicting process N times under the premise that the control quantity u is kept unchanged, and the position information of the N time domains is finally extracted to form a trajectory.

[0067] S3), the unmanned aerial vehicle probabilistically locks and tracks the target;

[0068] Before optimization, the unmanned aerial vehicle randomly decides whether to lock and track the target of the previous round by comparing a random number with a probability threshold value, and when the random number is less than the probability threshold value, the unmanned aerial vehicle locks and tracks the current target, which remains unchanged during the optimization process; on the contrary, the unmanned aerial vehicle does not lock and track the target.

[0069] S4), the control sequence of the unmanned aerial vehicle is optimized by using a chaotic adaptive mutation particle swarm optimization algorithm; specifically including the following steps:

[0070] S41), the tracking target of the unmanned aerial vehicle and the control sequence are combined into a particle;

[0071] S42), after each iteration, the fitness value of each particle in the particle swarm is evaluated, and whether the optimal fitness value of the particle swarm is degraded is detected; specifically including the following steps:

[0072] S421), a statistical period and a degradation threshold value are set as basic parameters for detection;

[0073] S422), in the initialization stage, the optimal particle fitness values of each generation in the first statistical period are collected, and the average value of these values is calculated as the benchmark performance value for subsequent comparison;

[0074] S423), after accumulating data for one statistical period, the average fitness value of the current period is calculated and compared with the benchmark performance value;

[0075] If the current average value is better than or equal to the benchmark performance value, the benchmark performance value is updated and monitoring continues;

[0076] If the benchmark performance value is worse and reaches the degradation threshold value, it is determined that a change is detected; and the detection state is reset to initialization to prepare for adaptive adjustment and detection.

[0077] S43), if the optimal fitness value is degraded, an adaptive strategy is used for adjustment, otherwise no change occurs and the particle swarm is iteratively optimized; specifically:

[0078] When the degradation is detected, the position and velocity of the reinitialization particle are detected, and the historical optimal solution record is emptied; and the motion inertia of the particle formed in the old environment is broken by chaos disturbance, so as to make them more evenly cover the solution space; when the change is detected, the mutation probability is enhanced and a fixed refractory period is maintained;

[0079] In the reinitialization stage after detecting the change, the inertia weight w is increased, and the global exploration ability of the particle swarm is enhanced; when the system completes initialization, the inertia weight W returns to the original value, and the search behavior of the particle swarm is biased to local fine search.

[0080] S5), the multi-unmanned aerial vehicle system reaches Nash equilibrium through multiple iterations, and the optimal control sequence of the unmanned aerial vehicle is obtained;

[0081] The embodiment sets the evaluation standard of the multi-unmanned aerial vehicle system based on Nash equilibrium, and when the control sequence obtained by iteration meets Nash equilibrium, the iteration is ended, and the control sequence obtained at present is taken as the final result, and if Nash equilibrium is not reached, the step S3) is returned.

[0082] S6), the optimal control sequence is used to control the multi-unmanned aerial vehicle to track the multi-target.

[0083] Embodiment 2

[0084] The embodiment provides a multi-unmanned aerial vehicle tracking multi-target optimization system based on chaos adaptive mutation particle swarm, comprising:

[0085] A data acquisition module is configured to acquire target position information and unmanned aerial vehicle state information;

[0086] A trajectory prediction module is configured to predict a future trajectory of the target by using an extended Kalman filter;

[0087] In the embodiment, the acquired target position information and speed information are taken as inputs of the extended Kalman filter algorithm to predict the future motion trajectory of the target with a time domain length of N; that is:

[0088] A tracking target probability locking module is configured to randomly determine whether to lock the target tracking of the last round by comparing a random number with a probability threshold; when the random number is less than the probability threshold, the unmanned aerial vehicle locks the tracking of the current target, and remains unchanged in the optimization process; otherwise, the unmanned aerial vehicle does not lock the tracking target.

[0089] A control sequence optimization module is configured to optimize the control sequence of the unmanned aerial vehicle by using a chaos adaptive mutation particle swarm optimization algorithm; specifically, the following steps are included:

[0090] S41), the tracking target and the control sequence of the unmanned aerial vehicle are combined into a particle;

[0091] S42), evaluating the fitness value of each particle in the particle swarm after each iteration, detecting whether the optimal fitness value of the particle swarm is degraded; specifically comprising the following steps:

[0092] S421), setting a statistical period and a degradation threshold as basic parameters for detection;

[0093] S422), in the initialization stage, collecting the optimal particle fitness value of each generation in the first statistical period, calculating the average value of these values, and taking the average value as the baseline performance value for subsequent comparison;

[0094] S423), every time the data of a statistical period is accumulated, the average fitness value of the current period is calculated and compared with the baseline performance value;

[0095] If the current average value is better than or equal to the baseline performance value, the baseline performance value is updated and the monitoring continues;

[0096] If it is worse than the baseline performance value and reaches the degradation threshold, it is determined that a change is detected; and the detection state is reset to initialization, preparing for detection after adaptive adjustment.

[0097] S43), if the optimal fitness value is detected to be degraded, an adaptive strategy is used for adjustment, otherwise no change occurs and the iteration optimization of the particle swarm continues; specifically:

[0098] When the degradation is detected, the position and speed of the particle are reinitialized, and the historical optimal solution record is emptied; and the motion inertia of the particle formed in the old environment is broken through chaotic disturbance, so as to make them more uniformly cover the solution space; when the change is detected, the mutation probability is enhanced and a fixed refractory period is maintained;

[0099] In the reinitialization stage after the change is detected, the inertia weight w is increased to enhance the global exploration ability of the particle swarm; when the system completes initialization, the inertia weight w returns to the original value, at which time the search behavior of the particle swarm is biased towards local fine search.

[0100] An iteration result judgment module is configured to judge whether the multiple UAV control sequences reach Nash equilibrium after multiple iterations, if yes, the current obtained UAV control sequence is taken as the optimal control sequence; otherwise, a tracking target probability locking module is called; the embodiment sets the evaluation standard of the multiple UAV system based on Nash equilibrium, and when the control sequence obtained through iteration meets Nash equilibrium, the iteration is ended, and the current control sequence is taken as the final result.

[0101] A control module is configured to control the multiple UAVs to track the multiple targets according to the optimal control sequence.

[0102] The embodiment controls the UAV to move according to the optimal control sequence as the final UAV moving strategy.

[0103] The foregoing embodiments and description of the application have been presented for the purposes of illustration and description. They are not intended to be exhaustive or to limit the application to the precise form disclosed. Many modifications and variations are possible in light of this disclosure. It was chosen and described in order to provide the best explanation of the principles of the application and its best mode of operation. Other embodiments will be apparent to those of ordinary skill in the art from this disclosure and have the full scope of the claims that follow.

Claims

1. A multi-UAV tracking multi-target optimization method based on chaotic adaptive mutated particle swarm optimization, characterized in that, Includes the following steps: S1) Obtain target location information and UAV status information; S2) Predict the future trajectory of the target using the extended Kalman filter algorithm; S3) Probabilistic target locking and tracking by drones; S4) Optimize the control sequence of the UAV using a chaotic adaptive mutant particle swarm optimization algorithm; S5) Through multiple iterations, the control sequences of multiple UAVs are made to reach Nash equilibrium, and the current control sequence of the UAVs is taken as the optimal control sequence. S6) Utilize optimal control sequences to control multiple UAVs to track multiple targets.

2. The multi-UAV tracking multi-target optimization method based on chaotic adaptive mutated particle swarm optimization according to claim 1, characterized in that: In step S2), the acquired target position and velocity information are used as inputs to the extended Kalman filter algorithm to predict the target's trajectory over the next N time domain lengths.

3. The multi-UAV tracking multi-target optimization method based on chaotic adaptive mutated particle swarm optimization according to claim 1, characterized in that: In step S3), before optimization, the UAV randomly decides whether to lock onto the target tracking of the previous round by comparing a random number with a probability threshold. When the random number is less than the probability threshold, the UAV locks onto the tracking of the current target and keeps it unchanged during the optimization process. Conversely, drones do not lock onto or track targets.

4. The multi-UAV tracking multi-target optimization method based on chaotic adaptive mutated particle swarm optimization according to claim 1, characterized in that: In step S4), the control sequence of the UAV is optimized using the chaotic adaptive mutation particle swarm optimization algorithm, which specifically includes the following steps: S41) Synthesize the UAV's tracking target and control sequence into particles; S42) After each iteration, evaluate the fitness value of each particle in the particle swarm and check whether the optimal fitness value of the particle swarm has degenerated. S43) If the optimal fitness value is detected to have degraded, an adaptive strategy is used to make adjustments; otherwise, no changes are made and iterative optimization of the particle swarm continues.

5. The multi-UAV tracking multi-target optimization method based on chaotic adaptive mutated particle swarm optimization according to claim 4, characterized in that: In step S42), after each iteration, the fitness value of each particle in the particle swarm is evaluated, and it is checked whether the optimal fitness value of the particle swarm has degenerated. This specifically includes the following steps: S421) Set the statistical period and degradation threshold as the basic parameters for detection; S422) In the initialization phase, collect the best particle fitness values ​​for each generation within the first statistical period, calculate the average of these values, and use it as the benchmark performance value for subsequent comparisons. S423) After accumulating data for a full statistical period, calculate the average fitness value for the current period and compare it with the baseline performance value; If the current average value is better than or equal to the baseline performance value, update the baseline performance value and continue monitoring; If the performance is worse than the baseline value and reaches the degradation threshold, a change is detected. It also resets the detection state to initialization to prepare for adaptive adjustment of the detection.

6. The multi-UAV tracking multi-target optimization method based on chaotic adaptive mutated particle swarm optimization according to claim 5, characterized in that: In step S43), if a degradation of the optimal fitness value is detected, an adaptive strategy is used for adjustment; otherwise, no change is made and iterative optimization of the particle swarm continues. Specifically: When degradation is detected, the position and velocity of the particles are reinitialized and the historical best solution record is cleared; and the motion inertia formed by the particles in the old environment is broken by chaotic perturbation, so that they cover the solution space more evenly; when change is detected, the mutation probability is increased and a fixed refractory period is maintained. During the re-initialization phase after detecting changes, the inertial weight w is increased to enhance the global exploration capability of the particle swarm. When the system completes initialization, the inertial weight w returns to its original value, at which point the particle swarm's search behavior tends towards localized refined search.

7. The multi-UAV tracking multi-target optimization method based on chaotic adaptive mutated particle swarm optimization according to claim 6, characterized in that: In step S5), the evaluation criteria for the multi-UAV system are set based on Nash equilibrium. If the control sequence obtained through iteration satisfies Nash equilibrium, the iteration ends and the currently obtained control sequence is taken as the final result; otherwise, the process returns to step S3.

8. A multi-UAV tracking multi-target optimization system based on chaotic adaptive mutant particle swarm optimization, characterized in that, The system is optimized using the method described in any one of claims 1-7, and the system comprises: The data acquisition module is used to acquire target location information and UAV status information; The trajectory prediction module is used to predict the future trajectory of the target using the extended Kalman filter algorithm; The target tracking probability locking module is used to randomly determine whether to lock the target tracking from the previous round by comparing a random number with a probability threshold. The control sequence optimization module is used to optimize the control sequence of the UAV using a chaotic adaptive mutation particle swarm optimization algorithm. The iteration result judgment module is used to determine whether the control sequence of multiple UAVs has reached Nash equilibrium after multiple iterations. If so, the current UAV control sequence is taken as the optimal control sequence; otherwise, the target tracking probability locking module is called. The control module is used to control multiple UAVs to track multiple targets according to the optimal control sequence.

9. A multi-UAV tracking multi-target optimization system based on chaotic adaptive mutated particle swarm optimization according to claim 8, characterized in that: The control sequence optimization module utilizes a chaotic adaptive mutation particle swarm optimization algorithm to optimize the UAV's control sequence, specifically including the following steps: S41) Synthesize the UAV's tracking target and control sequence into particles; S42) After each iteration, evaluate the fitness value of each particle in the particle swarm and check whether the optimal fitness value of the particle swarm has degenerated. S43) If the optimal fitness value is detected to have degraded, an adaptive strategy is used to make adjustments; otherwise, no changes are made and iterative optimization of the particle swarm continues.

10. A multi-UAV tracking multi-target optimization system based on chaotic adaptive mutated particle swarm optimization according to claim 9, characterized in that: After each iteration, the fitness value of each particle in the particle swarm is evaluated to check whether the optimal fitness value of the particle swarm has degenerated. This includes the following steps: S421) Set the statistical period and degradation threshold as the basic parameters for detection; S422) In the initialization phase, collect the best particle fitness values ​​for each generation within the first statistical period, calculate the average of these values, and use it as the benchmark performance value for subsequent comparisons. S423) After accumulating data for a full statistical period, calculate the average fitness value for the current period and compare it with the baseline performance value; If the current average value is better than or equal to the baseline performance value, update the baseline performance value and continue monitoring; If the performance is worse than the baseline value and reaches the degradation threshold, a change is detected. It also resets the detection state to initialization to prepare for adaptive adjustment of the detection.

Images