Modeling and Optimization for Multi-UAVs Cooperative Target Tracking
|School||National University of Defense Science and Technology|
|Course||Control Science and Engineering|
|Keywords||Unmanned Aerial Vehicles (UAVs) Target Tracking Active Sensing Fusion Estimation Trajectory Optimization Consensus Estimation Optimal Observation Configuration Optimal Observation Trajectory Standoff Tracking Persistent Tracking|
It has been full of theoretical and practical significance in both military and civilian areas, to effectively assign multiple unmanned aerial vehicles (UAVs) to perform target tracking mission within the complex environment. Under such a circumstance, this dissertation focuses on the multiple UAVs cooperative target tracking mission on the ground moving objects. Two key aspects of this mission are researched, in terms of target state fusion estimation and observations trajectory optimization. The main work and contributions are summarized as follows:(1) Factors that influence the multi-UAVs cooperative target tracking are analyzed, a solving framework to this problem is presented, and then the two key sub-problems are formulated respectively.Formulation on the problem of multi-UAVs cooperative target tracking is achieved by comprehensive analysis on this mission process. In detail, the basic characteristics in this problem are analyzed, and meanwhile, the factors including UAV platforms, airborne image sensors, targets and network communication are considered and formalized respectively.Afterwards, a solving framework based on active sensing is established for single UAV target tracking problem. Furthermore, this framework is extended for multi-UAVs mode by using a distributed archtecture with limited centralized control. Eventually, two key sub-problems are suggested and formalized, namely target state fusion estimation and observations trajectory optimization.(2) An Interacting Multiple Model Unscented Information Filter (IMM_UIF) based algorithm is proposed for multi-UAVs information fusion estimation, and meanwhile an Adaptive Consensus Distributed Unscented Information Filter (AC_DUIF) based algorithm is developed for multi-UAVs distributed information fusion estimation.A fusion estimation algorithm, IMM_UIF, is presented on the basis of unscented kalman filter,dstributed information filter and interacting multiple model. This IMM_UIF algorithm follows a hierarchically distributed sensor fusion architecture, and has good performance in estimating accuracy (unscented transform), robustness (interacting multiple model), and fusion ability (information). Simulation results show that the proposed method can effectively improve the target state estimation performance of the multiple UAVs.Furthermore, a novel algorithm, AC_DUIF, is proposed and developed for the consensus problem of multi-UAVs distributed estimation with switching network topologies. This proposed algorithm makes a beneficial combination of unscented information filter and adaptive consensus algorithm. Simulation results show that the proposed method can improve the accuracy of target position estimation of each sensor node in a distributed fashion, when the distributed consensus among network nodes and the adaptability of switching network topologies are assured.(3) Some analytical results of multi-UAVs optimal observation configuration and single UAV limited steps optimal observation trajectory are proved, in order to guarantee optimal observations trajectory for target state estimation. An optimal observation configuration based approximately optimal observations trajectory algorithm is also presented for the multiple UAVs.A performance evaluation framework of multi-UAVs observations configuration and trajectory with bearing and distance sensor is established, to maximize the determinant of the Fisher Information Matrix (FIM). Then, the necessary condition and the performance upper-bound of arbitrary UAVs optimal observation configuration are presented and proved. A kind of optimal observation configuration analytical results is suggested, after proving that the optimal observation configuration result is not unique. Specifically, all analytical results of optimal observation configuration for two or three UAVs are provided and highlighted via illustrative examples. And then, the analytical result of single step optimal trajectory for single UAV is exhibited as well. The performance upper-bound,as well as the sufficient and necessary condition of single UAV multiple steps optimal observation trajectory are presetned and illustrated.An approximately optimal algorithm is proposed for multi-UAVs observations trajectory solving, with considerations on the troubleness computing complexity. This algorithm is developed on the ground of the multi-UAVs optimal observations configuration results. Simulation results show that the proposed method can get a satisfactory performance of the target state estimation with less computing costs.(4) A Lyapunov Vector Field (LVC) based trajectory optimation algorithm is proposed for multi-UAVs standoff target tracking, while a Receding Horizon Control(RHC)based trajectory optimation algorithm is proposed for multi-UAVs persistent target tracking.Under the active sensing architecture, an autonomous solving framework is proposed for tactical mission oriented multiple UAVs target tracking. This framework is based on the target estimation and prediction. Then, two tactical target tracking missions are solved respectively.An improved central motion planning algorithm is presented to solve the problem of cooperative standoff target tracking. This algorithm introduces two Lyapunov functions to ensure the standoff distance and the relatively angle between UAVs and the target. Simulation results demonstrate that the performance of cooperative standoff target tracking is improved with regard to the fusion estimation and prediction of the target state. A receding horizon control scheme of the target tracking trajectory planning problem is presented for the problem of cooperative persistent target tracking in a cluttered environment. Since this control scheme considers the kinds of limits from the environment, an evolution algorithm is developed to slove the RHC problem. Simulation results demonstrate that the target can be sustainably tracked, with an improving accuracy of the target state estiamtion.