Development of robot swarm algorithms on an extensible framework

Abstract

Swarm intelligence for robots is inspired by observation of how homogenous collections of animals behave in nature to succeed in finding food and avoiding predators. Swarm robots usually lack centralized control to determine each robots individual behavior, however global behaviors can emerge through many local interactions, which are simple in nature. Studies show that simple rules executed on the individual robot can explain complex group behaviors and it is sufficient to support only local sensing and communication. The advantages of swarm intelligence are robustness at the level of the group where individual failure is not a significant problem; individual behaviors are easy to implement, and the approaches are scalable since the control mechanisms do not depend on the number of individuals in the swarm. We present an ongoing project that leverages hardware and extends software for swarm robotics. The hardware consists of robots called swarmies. A swarmie is a small robotic vehicle with a webcam, a GPS system, sensors like IMU, ultrasonic obstacle detector; a Wi-Fi antenna for wireless communication and an on-board computer. The objective of the project is to develop algorithms for the Swarmies so they can communicate and thus execute cooperative april-tag collection autonomously. The software is a ROS (Robot Operating System) controller framework for the Swarmie robots. We were able to improve the swarmie's behavior to search a space more effectively and utilize computer vision methods.