Distributed sliding model predictive control-based multi-target enclosing for autonomous underwater vehicle swarm

To address the challenges of limited local perception, model uncertainty, and unknown external disturbances in multi-target enclosing tasks for autonomous underwater vehicle (AUV) swarms, this paper proposes a group cooperative enclosing strategy that integrates a distributed observer with model predictive control (MPC). First, a dynamic model of the AUV swarm is constructed under a multi-target enclosing scenario. A distributed observer with local information fusion capability is designed to enable parallel estimation of multiple target states and achieve global consensus-based reconstruction. Subsequently, based on task relevance and spatial topology, the swarm is divided into several functional subgroups, forming a dynamic “subgroup–target” coupling structure. For each subgroup, a constrained MPC controller is designed to dynamically enclose its assigned target by solving an optimal control sequence over a finite prediction horizon in a receding horizon fashion. The proposed approach explicitly considers input saturation and inter-agent cooperation constraints, thereby achieving a balance between control accuracy and system robustness. Simulation results demonstrate the effectiveness and stability of the proposed method in dynamic multi-target environments, highlighting its strong engineering applicability and potential for practical deployment.