A leader-following approach based on probabilistic trajectory estimation and virtual train model

Abstract

The paper proposes a novel multi-robot leader-following approach that can offer accurate trajectory estimation and tracking control in scenarios where 1) temporary outage of vision detection happens due to the leader robot moving out of view, illumination variation, vision occlusion, motion blurring, etc., and 2) the leader moves at a time-varying linear velocity. The approach estimates the trajectory of the leader robot within the local reference frame of the follower using noise-corrupted odometry information and intermittent inter-robot relative observations based on detection of fiducial markers using an RGBD camera. It also introduces the virtual train model in the trajectory tracking, with which the follower is controlled to keep its linear velocity synchronous with that of the leader and a constant separation distance. Results are obtained based on evaluating the proposed leader-following approach in a test containing sharp turns and a zig-zag pattern in the trajectory, and variation in the speed of the leader robot.