Novel fixed point clustering with RETOPSOM for robust multi-USV task allocation

Task allocation for unmanned surface vehicles (USVs) in a multi-base departure mode substantially enhances mission execution efficiency and response speed, while reducing the risk of communication interruptions. However, traditional task allocation algorithms often encounter high computational complexity and fail to address the challenges posed by complex environments in multi-base departure scenarios. To address these issues, this paper proposes a two-layer task allocation framework. The upper layer performs task clustering among USVs, while the lower layer determines the task execution sequence for each USV formation. Firstly, an innovative fixed-point guided clustering algorithm is introduced, incorporating communication distance constraints and execution capability factors to address the task clustering problem for heterogeneous USVs in communication-constrained maritime environments. Subsequently, by introducing repulsion forces and modifying the neuron update process within the repulsive topology-preserving self-organizing map (RETOPSOM), the RETOPSOM algorithm is proposed to achieve task allocation under obstacle constraints. Finally, simulations in various scenarios demonstrate that the proposed algorithm effectively generates reasonable and efficient task allocation plans for heterogeneous USV formations departing from multiple bases in communication-constrained environments.