Paper accepted in IFToMM Asian conference on Mechanism and Machine Science

Link to paper

Finite-Time Formation Convergence of Vision-Based Nonholonomic Systems Without Explicit Communication

Abstract

This study considers the problem of synthesizing a decentralized control policy for leader-follower formation tracking of a group of nonholonomic agents in the absence of explicit inter-agent communication. Each follower robot is equipped with a perspective camera, and by tracking a single point feature attached to the leader agent, the leader-follower formation kinematics is developed in the image space of the follower agent. An adaptive continuous sliding mode controller is then proposed for the follower agent, and it is shown that the resulting closed-loop observer-controller system achieves global finite-time convergence of the tracking error to a small uniform ultimate bound around the origin in the absence of either explicit measurement/estimation of the leader agent’s position and velocity. Furthermore, it is shown that the size of this error bound can be arbitrarily reduced through a sufficiently large choice of the controller gain. Simulation results are used to validate and demonstrate robust performance of the proposed scheme in the presence of measurement noise.