快速平滑收敛策略下基于QS-RRT 的UAV运动规划
摘要
Rapidly exploring random tree (RRT) based motion planning algorithm constructs collision-free paths by biasing the exploration toward the unexplored task space with a random sampling scheme. This algorithm is probabilistically complete and computationally efficient. However, the length, safety and smoothness of the generated path must be improved in motion planning applications for unmanned aerial vehicles (UAVs). This paper models the threat environment, the UAV’s maneuverability and sensory ability, and then designs several optimal strategies with respect to sampling, obstacle avoidance, path navigability and path smoothing globally and locally. Consequently, a quick and smooth convergence RRT (QS-RRT) is obtained, and two improved optimal motion planning algorithms are presented for both known and unknown task spaces. Simulation results show that the algorithms can not only guarantee the convergence and path safety, but more importantly shorten the flight distance and remarkably improve the path navigability and smoothness. Furthermore, the algorithms can trade-off between the optimal degree and computational efficiency, which will optimize the adaptability to different practical mission requirements.