Adaptive patrol for a group of robots

Abstract

Patrolling is a basic task for a group of robots. All regions of an area are checked at regular interval and robots may look for intruders or garbage to collect. This work proposes a reactive and adaptive approach of this problem. In a virtual environment shared by robots, task data are propagated from place to place with modifications in order to represent locally distant-tasks value. A robot bas access to task data situated in the region where it stands and follows a gradient that guides it to valuable regions. First experiments on a simulator are presented. In addition of being efficient in achieving the work, the proposed architecture shows interesting properties of adaptability concerning the group size, the environment size and type.