Velocity Domain-Based Distributed Pursuit- Encirclement Control for Multi-USVs With Incomplete Information

This paper aims to investigate the distributed pursuit-encirclement control of multiple unmanned surface vehicles (multi-USVs) with incomplete information, characterized by local observation, communication constraint, and velocity absence, in complex environments. A velocity domain-based approach decomposing the pursuit-encirclement issue into velocity planning problem and velocity tracking problem is proposed to deal with this scenario, while considering obstacle avoidance and incomplete information. Firstly, the developed neural network (NN)-based unscented kalman filter (UKF), which is sufficiently real-time and adaptable to environmental features, is established to predict the velocity of the observed USVs. Subsequently, optimal reciprocal collision avoidance (ORCA) is employed to transform the pursuit-encirclement problem into the velocity domain, which enables the planning of optimal velocities that are collision-free while accomplishing the pursuit and encirclement mission. Furthermore, the velocity tracking problem for planned optimal velocity is considered as an optimization problem, where a cost function is designed and solved with adaptive dynamic programming (ADP) technique. Finally, the simulation results verify the effectiveness of the proposed distributed pursuit-encirclement control method.