Research On Mission Planning For Unmanned Aerial Vehicles Based On Multi-Stage Path Prediction

The applications of unmanned aerial vehicles(UAVs) call for an increasing level of autonomy. In the future, the ultimate goal of UAVs is to achieve autonomous swarm.Mission planning including task assignment and path planning is the basis of achieving the autonomous system for UAVs. The trajectory of UAVs depends on the result of task assignment. The objective function of assignment is based on the range-to-go of UAVs to each task. Thus, the research on combination of task assignment and path planning is practically useful. In this dissertation, a multi-stage path prediction(MSPP) algorithm of the mission planning for UAVs is presented.Firstly, according to the role of path planning, the MSPP is implemented on a hierarchical fashion, which consists of four stages: path estimation, path planning, trajectory smoothing and trajectory generation for rendezvous. Every stage provides varying degrees of reference trajectory for UAVs. The forbidden area is modeled as polygon. In order to calculate rapidly, the local A* algorithm utilizes the vertexes in adjacent area of the UAVs as the search nodes to construct the search space. The detected vertexes of the polygon is continuously added to the search space of global A* algorithm as the mission is being carried out. A specific description of the proposed algorithm is as following:1. Path estimation: In every planning horizon, each UAV utilizes the local A* algorithm first to estimate the path to each task. Then, the estimated results sever as input of the task assignment.2. Path planning: The polygonal shortest path is calculated by global A* algorithm on the basis of the assignment result. The global A* algorithm consists of the actual shortest path in the detection range and the heuristic path to the interesting task.3. Trajectory smoothing: Considering the flight constraints, the shortest path is further smoothed to obtain the flyable path by using the cubic B-spline curve, which serves as a reference path of the UAVs.4. Trajectory generation for rendezvous: For the rendezvous task, cooperative path planning for rendezvous of UAVs is considered. At the end of B-spline curve, the Dubins path is addressed as reference path. Based on the difference of range-togo between UAVs, the wandering maneuver with shorter range-to-go and circle maneuver with longer range-to-go are designed.Furthermore, based on the path estimation of MSPP, the task assignment is developed in two aspects: centralized and decentralized manner. By modifying the particlestructure, improved PSO algorithm is presented in centralized assignment to accelerate the searching process. Each element of particle is a Boolean variable, which indicate if the task is assigned to corresponding UAV. Then the search space of improved PSO is reduced. Proposed algorithm can produce a global optimal solution with fast computation.On the other hand, considering the disadvantages of centralized approach, a decentralized task assignment based on cluster algorithm and path estimation of MSPP is introduced.In the literature of task assignment, in order to obtain a global optimal solution, all the tasks are assigned at every planning horizon. Even for the decentralized method, the assignment must be re-computed in the presence of pop-up tasks. The cluster method is utilized to assign the task combination with the higher value first. All tasks are taken into account at every planning horizon. At the initial planning horizon, each UAV is assigned a task from the task list. When there is a pop-up task, it will be added to the task list and considered to be assigned in the next planning horizon. This strategy can generate the quasi-optimal task assignment.Finally, the frameworks of centralized and decentralized mission planning are designed and analyzed in four scenarios: centralized mission planning based on MSPP with and without rendezvous, decentralized mission planning based on MSPP with and without pop-up tasks. From application condition, typical applicable mission scenario, simulation result and application characteristics aspects, the effectiveness and efficiency of proposed method are demonstrated.