Multi-UAV Cooperative Tracking Ground Multi-target State Fusion Estimation

Multi-UAV cooperative tracking multi-target is one of the important tasks of multiUAV system.It has higher tracking and positioning accuracy and stronger robustness than single-machine tracking mode.It has wide application in military and civilian fields..In this paper,the key problems in multi-UAV collaborative tracking multi-target tasks are studied.The multi-UAV tracking strategy based on target clustering and the target state fusion estimation method with or without environmental information are studied.The main research work and innovations are as follows:(1)The system architecture,application scenarios and task flow of multi-UAV cooperative tracking multi-target are analyzed systematically.A hierarchical distributed system architecture is designed,and the task scenarios applicable to the current architecture are given.It is elaborated and the corresponding mathematical model is built for the core module.(2)The system architecture,application scenarios and task flow of multi-UAV cooperative tracking multi-target are analyzed systematically.A hierarchical distributed system architecture is designed.Under this architecture,the UAV combines the advantages of independent decision-making and superior intervention.The multi-UAV tracking multi-target task scenario applicable to the current architecture is given.The specific task flow is elaborated,and the UAV platform model,sensor model,target motion model and communication topology are constructed for the core module.(3)Aiming at the problem of ground motion target state fusion estimation under target dynamic maneuver and inter-machine communication delay,an adaptive multimodel unscented Kalman particle filter fusion method without environment information and variable structure interactive multi-model particles with environmental information are proposed.Filtering method.In the scenario without environmental information,in order to effectively improve the accuracy of state estimation,an improved turning model is designed to correct the turning rate according to the measurement information in real time,in view of the uncertainty of the turning rate when the target turns;The multi-model selection problem is proposed,and the adaptive multi-model method is proposed to make the target model more closely match the current target motion state.Considering that the real situation of the target tracking is performed under the nonlinear non-Gaussian system,the unscented Kalman particle filtering method is adopted.The target state is filtered and then all the target state estimation information is fused by the federated filtering method.In the scenario with environmental information such as road restriction information,the variable structure interactive multi-model particle filter method based on road restriction can effectively improve the accuracy of target location.The numerical simulation experiment is designed.Compared with the results of other filtering methods,it is proved that the proposed method has better state estimation performance and can effectively improve the target positioning accuracy.Finally,under the condition of environmental information such as no road in the field,the flight experiment is designed for the adaptive multi-model unscented Kalman particle filter fusion method.It is proved that the algorithm has the advantages of high precision and good stability in practical applications.