Adaptively Dynamic RRT*-Connect: Path Planning for UAVs Against Dynamic Obstacles

Abstract

Path planning for Unmanned Aerial Vehicles (UAVs), especially in three-dimensional environments with dynamic obstacles, is an active area of research. In recent years, RRT-based algorithms have been attracting significant interest, and various improvements are introduced to RRT to make it applicable in such scenarios. Some methods leverage re-planning mechanisms to avoid collision with the dynamic threats, however, most of them fail to fully reuse historical information, hence having limitations in saving the planning costs. This paper presents Adaptively Dynamic RRT*-Connect (ADRRT*-Connect), a novel RRT-based path planning algorithm that enables UAVs to fly safely in three-dimensional environments with dynamic threats. To improve efficiency in sampling new nodes, we propose a strategy to automatically adjust the heuristic factor based on feedback from the sampling results. For avoiding collision with dynamic threats, we introduce a pruning-reconnecting mechanism to repair the path when new obstacles emerge. Our approach is economical in the consumption of tree nodes. In comparison to existing benchmarks, simulations have shown that our proposed algorithm only requires 3.5% new nodes to repair the path in re-planning.

Publication
2022 7th International Conference on Automation, Control and Robotics Engineering (CACRE)

At the conference CACRE 2022, I was honored the Best Presenter Award for an excellent oral presentation.

Best presenter award - Yicheng Chen

Yicheng Chen
Yicheng Chen
PhD student in robotics

I am interested in motion planning problems in aerial robotics.