RCF-RRT*: A risk-clearance cost-field-guided RRT*-based planner for global USV path planning in static risk-constrained maritime environments

Abstract

Path planning for unmanned surface vehicles (USVs) in risk-constrained maritime environments requires the simultaneous consideration of obstacle avoidance, environmental risk, safety clearance, and trajectory executability. Existing sampling-based planners can efficiently explore feasible regions; however, their expansion process is often dominated by geometric metrics and therefore does not fully capture the coupled influence of environmental risk and obstacle clearance. In addition, unconstrained post-smoothing may reduce clearance, violate risk constraints, or increase local heading variation. To address these issues, this paper proposes RCF-RRT*, a risk-clearance cost-field-guided RRT*-based planning framework with feasibility-certified spline refinement. First, a traversal cost field is constructed by integrating an obstacle-distance field with an environmental risk field, and the corresponding cost-to-go information is used to guide tree expansion. Second, an adaptive step-size strategy and hard-constrained point/edge feasibility checks are introduced to reject unsafe candidate nodes and connections under collision, risk, and clearance constraints. Third, a Pareto-dominance parent-selection strategy is designed to improve the tree structure by jointly considering path length, length-normalized risk exposure, normalized turning cost, and clearance margin. Finally, a feasibility-certified and quality-gated spline-refinement mechanism is developed to prevent unsafe or degraded smoothing results from being accepted. Simulation results show that the proposed Pareto parent-selection strategy reduces the average path length by approximately 5.43% and the normalized turning cost by approximately 68.69% while maintaining a success rate of 1.00. In the complex risk-constrained scenario, RCF-RRT* reduces the average final path length by approximately 3.18% and the normalized turning cost by approximately 32.51% compared with Bi-APF-RRT*, while preserving zero sampled feasibility violations under the adopted collision, risk, and clearance criteria. These results indicate that the proposed framework improves path compactness, turning smoothness, and feasibility preservation, providing an effective global planning method for static risk-constrained USV navigation.