Online Trajectory Generation with Rendezvous for UAVs Using Multistage Path Prediction

To improve the overall performance of mission planning for the unmanned aerial vehicles (UAVs), a multistage path prediction (MPP) for trajectory generation with rendezvous is addressed in this paper. The proposed real-time algorithm consists of four stages: path estimation, path planning, flyable trajectory generation, and trajectory modification for rendezvous. In every planning horizon, each UAV utilizes the local A* algorithm to estimate all probable paths and then the results serve as input for the task assignment system. A simple assignment algorithm is briefly introduced to validate the effectiveness of MPP. Based on the assignment, the polygonal paths are further obtained by using the global A* algorithm. Then these paths are smoothed to be flyable by using the cubic b-spline curve. In the last stage, the trajectories are modified for rendezvous of the UAVs to execute specific many-to-one tasks. Results of different stages are continuously revised and delivered to the task assignment system as feedback in the whole mission process. Numerical results demonstrate the performance of the proposed approach for stochastic scenarios.