A dynamic path planning algorithm for underwater robots integrating the improved BI-APF-RRT* and DWA

Abstract To address the problems of low global search efficiency in path planning for complex underwater environments, and insufficient obstacle-avoidance robustness, this paper proposes a hybrid path-planning method.Firstly,at the global planning level, a bidirectional rapidly exploring random tree star(BI-RRT*) algorithm integrating dynamic elliptical constrained adaptive sampling with intelligent obstacle-bypassing and an improved artificial potential field (APF) guidance mechanism is proposed. A dynamic ellipse contraction mechanism is introduced to reduce ineffective sampling.Additionally, a weighted fusion of obstacle-avoidance direction and potential field guidance is used to address the local minima and path redundancy issues inherent in traditional APF methods.Secondly,at the local planning level, an improved Dynamic Window Approach (DWA) that incorporates ocean current effects is developed. A vortex-type ocean current model is embedded into the kinematic prediction process. A continuous proximity cost function applicable to both circular and rectangular obstacles is constructed. Furthermore, a risk index based on the time-to-collision (TTC) concept is introduced by integrating minimum safety distance and velocity information. Finally,the global planning result is provided to the local planner as a guiding path to enhance global consistency during local trajectory generation.Comparative simulation results demonstrate that the proposed method outperforms conventional algorithms in path-planning efficiency, obstacle-avoidance safety, and planning robustness.