Automatic Guided Vehicles: Robust Controller Design in Image Space

We have been interested in Automatic Guided Vehicles (AGV) for several years. In this paper, we synthesize controllers for AGV applications using monocular vision. In particular, we are interested in road following and direction change tasks, and in analyzing the influence of extrinsic camera parameter perturbations on vehicle behavior. We use the bicycle as the kinematic vehicle model, and we choose the position of the white band on the road as the sensor signal. We define an interaction between the camera, which is mounted inside the vehicle, and the white band detected in the image space. Using this kind of interaction, we present how to use a pole assignment technique to solve the servoing task. We show the simulation and experimental results (1/10 scale demonstrator) with and without perturbations. We then investigate the use of a robust controller to slow down the effect of perturbations on the behavior of the vehicle.