A Vision-Based Feedback and Supervision System for Robotic Swarms

Abstract

A robotic swarm is a large collection of simplistic robots that use decentralized control to achieve a coordinated behavior. The individual robots make decisions based on data collected from their interactions with one another as well as their interactions with their environment. In this work, we are suggesting an interventional, vision-based supervision system that when needed provides feedback to individual robots in the decentralized swarm to improve the overall swarm performance. Using a spotlight, the system provides closed feedback to the system. The suggested system does not replace the individual controllers or swarm behaviors on each robot but complement them when the need arises. The system was tested on two different applications. In the first application, the system showed a 100% success rate in preventing a robot from exhibiting a negative behavior. In the second application, the system was used to support a robot perform a positive behavior. The results show that using the system results in an average speed up of more than 4 times.