Following, Surrounding and Hunting an Escaping Target by Stochastic Control of Swarm in Multi-Agent Systems

In this paper, we have proposed a quasi-static model of a multi-agent system as police agents with and without non-holonomic constraints for agents’ motion. We also consider an active thief agent which has stochastic dynamic equation of motion. Our control purpose is that these police agents pursue the thief agent and the geometrical center of them aims the thief agent current position. Also in the motion of the police agents for surrounding the thief, the unity of the swarm should be preserved and collision between them should be avoided. Police agents do not have all of the dynamic states of the thief agent’s motion and also those that are available and observable for the police agents are noisy, so we use an Extended Kalman Filter (EKF) to eliminate this imperfection with estimating all states. The results showed that even if measurements are incomplete and noisy and non-holonomic constraints are applied, police agents can follow thief agent in the sense of minimum variance.