An approach to adaptive swarm surveillance using social potential fields

Swarm robotics is a field of research inspired from how a biological system coordinates in a distributed and decentralized fashion. This distributed autonomous control mechanism makes swarm robotics a better alternative for mobile surveillance application. This paper presents a swarm intelligence algorithm to detect intrusion on a land under surveillance. The system consists of hundreds of robots governed by a set of decentralized control laws. Social Potential Field, a popular approach for distributed autonomous control, is used for the controlling motion of every robot in a swarm. Since intrusion is highly dynamic in nature, a parameter optimization approach using online gradient descent is introduced to make intrusion detection and response fast and effective. An artificial simulation environment consisting of Guard robots, Castle and Invader is created to verify the working of the proposed approach. Computer simulation results are presented for illustration and a comparison is made with the results obtained when the system lacks the ability to adapt to the changing surrounding environment.