Vision-Based Object Tracking Using an Optimally Positioned Cluster of Mobile Tracking Stations

Abstract

This paper presents a novel, highly capable strategy for utilizing a multirobot network to track a moving target. This method optimizes the configuration of mobile tracking stations in order to produce the position estimate for a target object that yields the smallest estimation error, even when the sensor performance varies. This is verified in both simulation and physical experiments using groups of two and three quadrotor aerial robots as mobile tracking stations controlled using a cluster control approach. These quadrotors track and follow an autonomous robot using only the data provided by the quadrotors’ onboard cameras. This results in a simple, robust system that can accurately follow a moving object.