A Lightweight and Efficient Improved RRT* Algorithm for Global Path Planning in Complex Environments

In complex obstacle environments, the RRT* algorithm, an asymptotically optimal variant of the Rapidly exploring Random Tree (RRT), and its related variants often suffer from slow generation of the initial feasible solution, unstable sampling efficiency, and high computational costs associated with nearest-neighbor search and collision checking. To address these issues, this paper proposes a coordinated lightweight improved RRT* algorithm. First, a bidirectional growth mechanism combined with goal-biased sampling is introduced to enhance search directionality and improve the efficiency of initial feasible path generation. After an initial path is obtained, informed elliptical sampling is adopted, and the sampling weights are adaptively allocated among the elliptical region, the global space, and goal-biased sampling, thereby balancing local convergence and global exploration. Furthermore, a spatial-hash structure with a dynamic neighborhood radius is employed to accelerate nearest-neighbor search, while lazy collision checking and a two-stage collision-detection mechanism are incorporated into parent selection to reduce redundant expansions and unnecessary exact collision checks. Simulation results in mixed-type and single-type obstacle environments show that the proposed algorithm improves planning efficiency while maintaining competitive path quality. These results demonstrate that the proposed method has good engineering applicability for global path planning in complex environments.