Online Motion Planning for UAV under Uncertain Environment
Xiaoting Ji, Jie Li · 2015
This paper proposes a Partially Observable Markov Decision Process (POMDP) based planning framework for the fixed-wing unmanned aerial vehicle (UAV) under sensing and motion uncertainty. The objective is to navigate a UAV with a partially known map and noisy sensors to reach the goal position while avoiding obstacles. Taking the non-holonomic constraints into consideration, the off-line point-based planning algorithm is first proposed to compute the approximate value function and find the locally optimal solution, which must be collision-free and feasible. In order to be suitable to the changeable environment with unknown obstacles, a real-time on-line planning algorithm is designed. Simulation results demonstrate the efficiency and utility of our approach for UAV guidance in the uncertain environment.