An HRMCTS-Based Optimization Method for Efficient Multi-Objective Path Planning

Path planning for unmanned systems in complex environments must simultaneously satisfy safety, kinematic feasibility, and real-time performance requirements. Monte Carlo Tree Search (MCTS) offers advantages such as model-free operation, strong interpretability, and anytime planning capability, but it suffers from large branching factors, excessive search depths, and poor convergence under sparse reward conditions in high-dimensional state spaces. To address these challenges, this paper proposes a Heuristic Rolling Monte Carlo Tree Search (HRMCTS) framework. First, the path planning problem is formulated as a constrained Markov decision process, where the state consists of position and heading, and actions are discretized heading changes. Second, a heuristic selection strategy incorporating goal-directed guidance and obstacle safety margins is introduced to improve search directionality, while a limited-depth forward simulation with branch pruning is employed during the rollout phase. A multi-objective reward function is designed to integrate distance, goal progress, tangent-based obstacle avoidance, smoothness, and efficiency, thereby jointly optimizing path quality and computational performance. Experiments are conducted in three scenarios: static polygonal environments, dynamic circular obstacle environments, and dynamic polygonal obstacle environments. Simulation results demonstrate that the proposed method offers significant advantages in terms of planning efficiency, environmental adaptability, generalization capability, and interpretability.