A decentralized localization scheme for swarm robotics based on coordinate geometry and distributed gradient descent

Abstract

In this paper, a decentralized localization scheme using coordinate geometry and distributed gradient descent (DGD) algorithm is presented. Coordinate geometry is proposed to provide a rough estimation of robots' location instead of the traditional trigonometry approach, which suffers from flip and discontinuous flex ambiguity. Then, these estimations will be used as initial values for DGD algorithm to determine robots' real position. Evaluated results on real mobile robots show an average mean error of 2.56 cm, which is closed to the minimum achievable accuracy of the testing platform (2 cm). For a team of eight robots, the total average run time of the proposed scheme is 66.7 seconds. Finally, its application in swarm robotics is verified by experimenting with a self-assembly algorithm named DASH.