Sense Limitedly, Interact Locally: the Control Strategy for Swarm Robots Search

Abstract

Swarm robotics is a novel area for search applications. Victim search & rescue operations with swarm robots in some disaster scenarios have attracted increasing attention within the field of multi-agent research community. Swarm robots possess the potential to save lives by providing the rescuers location of victim during the most critical early hours. The ongoing project of swarm robots for search victim in coal mine disaster areas focuses on the modeling method and control strategies. This paper modeled such distributed system at an abstract level in accordance with the swarm intelligence principles, which used a virtual multi-agents search to identify the target position in a closed 2-D workspace. The control over all robots for searching target was divided into global search and local search: before the target signal was detected by both certain robot itself and its neighbors belonging to the same neighborhood structure, it would move spirally rather than random walk for search cues to offer evidence for using extended particle swarm optimization algorithm for search target. Taking inherent characteristics of physical robots, finite sense ability and local interaction mode into account in control policy, we obtained the properties of the system by changing different parameters such as number of robots, communication range and sense scope. Simulation indicates the validity of the control strategy.