AMP-RRT*: an adaptive multi-layer path planning algorithm for robots in complex environments

Abstract

Abstract To enhance the path planning capability of robots in complex environments, this paper proposes an Adaptive Multi-layer Path planning algorithm, AMP-RRT*. Built upon a heuristic reverse RRT* framework, the proposed method integrates modules for avoiding both unknown and dynamic obstacles, forming a three-layer cooperative path optimization mechanism. The algorithm improves perception and avoidance of unknown obstacles through local resampling, adaptive neighbour node searching, and an optimized Pareto filtering strategy. Additionally, it incorporates dynamic detection and collision prediction mechanisms to enable real-time obstacle avoidance and path repair in dynamic environments. Systematic simulations are conducted on maps of three complexity levels, containing static, unknown, and dynamic obstacles, respectively. Experimental results demonstrate that AMP-RRT* consistently outperforms baseline algorithms in terms of planning time, path length, smoothness, and success rate, with particularly notable advantages in complex and highly uncertain scenarios. The proposed method exhibits strong generalizability and stability, offering an efficient and practical solution for robot path planning tasks in complex environments.