Decentralised Coordination in Swarm Robots Through XGBoost-Enhanced Colour Light Communication

Abstract

Inspired by natural swarm systems, robotic swarms aim to solve complicated problems through the emergent behaviour of coordinating robots (agents). Communication among the robots is of paramount importance for their effective coordination, cooperation, and overall performance. This research presents a colour light-based communication system for miniature mobile swarm robots, on which a pre-trained supervised machine learning model runs and is responsible for effective colour recognition, enhancing inter-robot local communication. The performance of various supervised machine learning techniques was examined, and XGBoost performed best overall, with a classification accuracy of 96.66%, an execution time of 0.403 ms, an average sensing distance of 87.38 cm, and an acceptable size of 402.1 kilobytes while running on a 32-bit embedded microcontroller. The current work also demonstrates various swarming behaviours, utilising the developed communication as proof of concept.