Dynamic Multi-Task Planning for Unmanned Surface Vehicles: Spatiotemporal Clustering and Game-Enhanced Intelligent Coordination

Unmanned surface vehicles (USVs) are increasingly utilized in dynamic maritime missions requiring high levels of autonomy, responsiveness, and coordination. However, real-time multi-task planning in open-sea environments remains constrained by limited computational efficiency, frequent task conflicts, and inefficient resource utilization. This paper proposes an intelligent cooperative task planning framework—Spatiotemporal Clustering and Game-enhanced Intelligent Coordination (SCGIC)—to enable robust and adaptive decision-making in dynamic USV operations. The proposed approach first introduces a joint spatiotemporal distance metric and an incremental update mechanism to construct a spatiotemporal informed incremental K-Means (SIIK) algorithm. This algorithm maintains spatiotemporal coherence within evolving task clusters while supporting online task insertion and adaptive cluster structure adjustment. Furthermore, a game-theoretic task allocation mechanism based on Self-Organizing Map (SOM) neural networks is incorporated, wherein neurons engage in a bidding process to handle task priority conflicts, temporal constraints, and resource optimization. To handle emergent tasks and resource contention, a lightweight negotiation protocol is introduced, combining a Nash bargaining model with a budget-aware auction strategy to enable rapid local task reconfiguration. Simulation results demonstrate that SCGIC consistently outperforms baseline methods across key metrics—including path efficiency, conflict mitigation, and response latency—showcasing superior robustness and real-time performance. This study advances scalable and adaptive mission planning for autonomous USVs, with broad applicability to scenarios such as ocean surveillance and search-and-rescue operations.