SD-Planner: A hierarchical A*-DWA trajectory generation framework for USV coastal inspections

The deployment of Unmanned Surface Vehicles (USVs) for autonomous coastal inspection has emerged as a critical imperative, driven by the inherent inefficiencies and safety hazards of manual methodologies. To address the conflict between global path quality and local dynamic reactivity, this paper proposes SD-Planner (Smooth and Dynamic Planner), a novel hierarchical trajectory generation framework. Within this architecture, the global routing layer utilizes a directionally-guided five-neighbor A* algorithm acting as a macro-level geometric filter, integrated with cubic B-spline smoothing. This effectively prunes kinematically unreasonable maneuvers and provides a collision-free reference corridor with improved obstacle clearance and kinematic consistency in narrow waterways. At the local reactive level, the standard Dynamic Window Approach (DWA) is improved by incorporating a reference-path deviation penalty, kinematic smoothness costs, and a hybrid normalization strategy. This enables the local planner to reduce unnecessary deviation from the global reference path while suppressing high-frequency control oscillations, thereby improving waypoint convergence and supporting real-time geometric obstacle avoidance in the tested simulation scenarios. MATLAB simulations and chart-based case studies demonstrate that the SD-Planner generates smooth and collision-free trajectories in the tested simulation scenarios, improving the stability and execution efficiency of USV operations in complex marine environments.