Homography-based visual servoing of wheeled mobile robots

Abstract

By comparing the features of an object from a desired image to features of the object in the current image, geometric relationships are exploited to enable a Euclidean reconstruction from a homography matrix that relates the image-space feedback to the position/orientation of the wheeled mobile robot (WMR) in a local coordinate system. The information obtained from the homography is used to develop a kinematic controller that yields asymptotic regulation of the position/orientation of a WMR system that is modeled as an underactuated "kinematic wheel" subject to nonholonomic constraints. The control design is facilitated by performing the stability analysis in terms of the unmeasurable camera/WMR Euclidean position. In contrast to many of the previous homography-based visual servo controllers, the kinematic control law does not depend on the numerical estimation of depth measurements. The control design is based on the nonlinear model of the vision system and the mobile robot system and is analyzed through a Lyapunov-based stability analysis.