Dynamic task partitioning for foraging robot swarms

Abstract

Dead reckoning error is a common problem in robotics that can be caused by multiple factors related to sensors or actuators. These errors potentially cause landmarks recorded by a robot to appear in a different location with respect to the actual position of the object. In a foraging scenario with a swarm of robots, this error will ultimately lead to the robots being unable to return successfully to the food source. In order to address this issue, we propose a computationally low-cost finite state machine strategy with which robots divide the total travelling distance into a variable number of segments, thus decreasing accumulated dead-reckoning error. The distance travelled by each robot changes according to the success and failure of exploration. Our approach is more flexible than using a previously used fixed size approach for the travel distance, thus allowing swarms greater flexibility and scaling to larger areas of operation.