Research On Dynamic Decision Method In Multiple Extended Targets Tracking Optimization

With the deep development of information fusion theory and the wide application of high-resolution sensors,multiple extended targets tracking(METT)has been applied in many fields such as military,industry,agriculture,and gradually become one of the research hotspots in the academic community.The so-called extended target refers to a target with multiple scattering sources observed by high-resolution sensors.By processing the relevant information of all scattering sources,more comprehensive features of the target can be estimated.Due to the requirements of practical engineering for high-precision target comprehensive feature estimation and other special tasks,optimizing and improving the overall performance of METT systems becomes particularly important.During the high-precision operation of the METT system,the sensor control process plays a key role.By making dynamic decisions on the sensor system,flexibly scheduling limited sensor resources and giving full play to their optimal performance,the quality of the sensor observation process is improved,thereby improving the input information of the METT system.In view of this,during the METT process,this dissertation has deep studied the dynamic decision-making methods for optimizing the real-time working state of sensors,achieving the overall optimization process of the METT system through reasonable decision-making of sensor control strategies,and improving the estimation accuracy of multiple extended target comprehensive features.The main work is as follows:1)In order to optimize the performance of multiple extended targets tracking system,this dissertation proposes a sensor control decision method based on Gaussian Wasserstein distance minimization using multiple extended targets posterior information.First,the algorithm calculates the comprehensive prediction information of multiple extended targets using the Poisson Multi-Bernoulli Mixture(PMBM)filter and a Gaussian inverse Wishart filter,and then establishes a discrete decision space based on the current state of the sensor.On each element of the decision space,the prediction ideal measurement set(PIMS)is constructed from the comprehensive prediction information of multiple extended targets,thereby performing a pseudo update process and extracting pseudo posterior information for each target,based on this,decision criteria are constructed and numerical calculations are performed to determine the optimal sensor action.Finally,the optimized posterior information of multiple extended targets can be obtained by adjusting the working state of the sensor to obtain the optimal observation process and updating it optimally.The simulation results show that the sensor control strategy proposed in this dissertation is effective in improving the accuracy of multiple extended target comprehensive feature estimation.2)In order to reduce the probability of missing the optimal decision solution in the process of multiple extended targets optimal tracking in the discrete decision space,this dissertation proposes a sensor control decision-making method based on particle swarm optimization algorithm to complement the completeness of sensor control decision-making processes in discrete decision space.First,the algorithm calculates the prediction information of the comprehensive characteristics of multiple extended targets through the prediction process of a PMBM filter and a Gaussian inverse Wishart filter,and then uses this as a priori information to solve the optimal observation state of the sensor based on the criterion of maximizing the proximity of each target through a particle swarm optimization algorithm.Next,high-quality observation information of each target is captured by high-resolution sensors operating in the optimal state,Finally,the optimized multiple extended target synthesis features are obtained through the update process of PMBM filter and Gaussian inverse Wishart filter.Simulation experiments have compared the optimization effects of multiple extended target comprehensive feature estimation in discrete and continuous decision spaces,and the results show that the multiple extended target comprehensive feature estimation optimization strategy in continuous decision spaces has better estimation accuracy.3)In order to reduce the threat of the enemy target to the unmanned aerial vehicle(UAV)during the reconnaissance mission,this dissertation proposes a path planning strategy to effectively avoid the threat.First,the airborne high-resolution sensor is used to accurately and robustly estimate the state of each target,in which the PMBM filter is used to estimate the motion state of each target,and the Gaussian inverse Wishart filter is used to estimate the extended information of each target,and then the Technology for Order Preference by Similarity to an Ideal Solution(TOPSIS)is used to quantify the target threat degree and divide the threat target according to the three-way decision rule,Finally,combining the target threat degree and target tracking quality,a joint evaluation index is established to determine the optimal path of UAV.The simulation experiment compares the threat avoidance of UAV under different paths,and the results show that the path planning strategy proposed in this dissertation can reduce the target threat degree of UAV under the condition of ensuring good target comprehensive state estimation accuracy.