Goal-Biased Rapidly-Exploring Random Trees for Efficient Marine Path Planning

This paper presents a goal-biased rapidly-exploring random tree (RRT) approach for rapid path planning in marine environments. The key innovation integrates a goal-biased sampling strategy to enable more efficient exploration towards the goal region compared to the traditional RRT algorithm. The implementation also incorporates an artificial potential field-inspired steering method for smoother paths and a path optimization technique adapted from prior work to reduce redundant nodes. The proposed goal-biased RRT planner was validated on a real-world marine map, demonstrating significantly improved path planning performance over state-of-the-art RRT variants, like Informed-RRT^∗ and RRT^∗-Smart algorithm, including optimized path length, faster computation, and better scalability. The efficiency gains address challenges of robotic marine navigation to some extent by providing a rapid, smooth, and optimized planning algorithm well-suited for applications like autonomous surface vehicles (ASV) in complex seascapes. Results highlight the method's characteristics and advantages for efficient path generation across real-world environments.