Vision-assisted adaptive target tracking of unmanned ground vehicles

In this paper, we develop a vision-based adaptive control algorithm for target tracking from a collaborative UAV-UGV platform. By estimating global camera motion and registering video frames into a common coordinate system, the UAV detects moving targets, as well as the UGV on the ground. It estimates their locations and sends this information to the UGV to guide its tracking of the targets. Based on this vision-based location estimation, we develop an adaptive fault-tolerant control scheme for the UGV to accurately track the target. We have designed an adaptive controller to adjust the control currents to the motors which are mounted on its left and right wheels so that the vehicle is able to follow the target trajectory. We investigate actuator fading of the vehicle, unknown parameters in the vehicle system model, together with the impact of the uncertainty and noise in computer vision processing on the overall tracking performance. Our simulation results demonstrate that the proposed algorithm is very efficient.