Adaptive bidirectional heuristic rapidly exploring random tree* for efficient path planning

Abstract

Sampling-based path planning algorithms such as rapidly exploring random tree* (RRT*) are widely used for autonomous navigation in complex environments. However, many RRT variants suffer from slow initial exploration, suboptimal convergence, and search inefficiency in dense spaces. Based on this, adaptive bidirectional heuristic-RRT* (ABH-RRT*) is proposed. It is a novel method introduced as a unified path planner. ABHRRT* integrates bidirectional tree growth, heuristic-based parent selection, fast-informed hybrid sampling, and adaptive reordering to improve exploration efficiency and path optimality. The algorithm speeds up the initial path recovery caused by the presence of dual tree expansion and fast sampling. In addition, the algorithm also refines the solution using informed sampling and adaptive reordering to improve convergence toward near-optimal paths. The performance of ABH-RRT* is evaluated in four environments with different complexity levels and compared with RRT, RRT*, Fast-RRT*, Smart-RRT*, and Informed-RRT*. Experimental results show that ABH-RRT* consistently produces shorter paths and faster convergence, reduces path cost by 2–24% and increases convergence speed by 40–58% in dense and constrained environments. These results show that ABH-RRT* is a better and adaptive solution for path planning in complex scenarios.