Gesture-based Graph-theoretic Robot Formation Control

Abstract

Multi-robot systems have many diverse applications including but not limited to surveillance, search and rescue operations, hazardous material handling, agriculture etc. Formation control is a key requirement out of the many facets of robotic swarm control. While the formation control problem has been extensively explored, including vastly different and various approaches; consensus-based formation control has definitive advantages in terms of providing a robust and distributed control algorithm. In the proposed work, the notion of consensus-based control of the formations of a robotic network utilizing bare-handed gestures is explored. While similar examples can be found in recent research endeavors (including selection and command of groups of UAVs, UGVs and under-water robots), most utilize electronic or specialized devices for the purpose of Human-Swarm Interaction (HSI). This is despite the major breakthroughs in the field of image recognition and classification. The proposed gesture-recognition approach is based on an algorithm capable of discerning the finger count (number of open digits) in 2.5 seconds (on average) with 88.75% accuracy thus affording the user a vocabulary of up to five (5) individual commands / directives to assemble and control the robotic swarm. The gesture recognition algorithm was modified to increase the detection accuracy to 90.8%.