Global–local unmanned surface vehicle path planning with bidirectional GOOSE–A* Fusion and predictive Dynamic Window Approach

For port, inland waterway, and near-sea operations, unmanned surface vehicle (USV) autonomous navigation faces increasingly stringent demands for path quality and real-time performance. Although the Dynamic Window Approach (DWA) retains computational efficiency, it converges to local optima and systematic tuning of the prediction horizon remains difficult; conventional global planners in complex environments also exhibit reduced search efficiency, insufficient path quality, and performance bottlenecks tied to single-algorithm reliance. To address these issues, this paper proposes an integrated global–local framework. At the global planning level, a Bidirectional Information Exchange Fusion Algorithm (BIEFA) fuses GOOSE and A* through bidirectional information exchange, raising path quality and supplying higher-quality global solutions for subsequent fusion with local planners; the same idea extends to other heterogeneous algorithm combinations for synergistic complementarity. At the local planning level, DWA is cast under finite-horizon constrained optimal control as a simplified model predictive control (MPC) solver over discrete velocity space, with online prediction-horizon optimization: lengthening the horizon within constraints enlarges lookahead, alleviates local optimality, and improves trajectory smoothness; shortening it before detour risk increases suppresses unnecessary detours. Numerical simulation results show that the proposed approach attains a superior balance among trajectory smoothness, global consistency, and real-time capability while honoring safety constraints.