Comparative Analysis of A* and RRT Variants for Mobile Robot Path Planning in Multi-Scale Grid Environments

Abstract

To address the trade-off between planning efficiency, path optimality, and trajectory tracking accuracy of mobile robot path planners under different spatial environments, this paper constructs square $(\mathbf{5 0 0} \times \mathbf{5 0 0})$ and elongated $(\mathbf{1 0 2 4} \times \mathbf{3 5 0})$ grid simulation maps covering short-range local, full-map large-span, and narrow ultra-long horizontal navigation tasks. The performance of three heuristic functions, heuristic weights for A*, expansion step sizes of RRT, as well as unidirectional and bidirectional search frameworks, including bidirectional A* and RRT-Connect, are comprehensively compared. Three quantitative evaluation indicators, namely planning time, total path length, and DTW-based average tracking error, are adopted to measure algorithm performance. Experimental results indicate that Manhattan distance achieves balanced overall performance among the tested heuristics; small heuristic weights and small RRT step sizes preserve high-quality paths in confined scenes, while larger weights and step sizes greatly accelerate planning for wide-range environments. Bidirectional search remarkably cuts computational overhead for both graph-based and sampling-based planners. Bidirectional A* generates near-optimal paths, whereas RRT-Connect shows overwhelming real-time advantages with minor path redundancy, especially in narrow large-scale maps. This work offers clear parameter selection strategies and algorithm application references for differential-drive robots navigating various complex static grid environments.