Optimal Path Planning for Image Based Visual Servoing

Abstract

Visual servoing is a control technique that uses image data as feedback in a motion control loop. This technique is useful in tasks that require robots or other automated motion systems to automatically inspect parts or structures. One specific method of visual servoing is Image Based Visual Servoing (IBVS), a method that simply minimizes the differences between an observed image orientation and a desired one. This method works well for orientations where these differences are small, but in the case where the desired orientation is more difficult to reach, the motion system can follow non-optimal trajectories and requires more computation time. In this paper we attempt to address this large difference issue by applying Dynamic Programming, an optimal control method that can determine an optimal trajectory by partitioning possible trajectories into multiple smaller trajectories. Using a cost function to penalize undesirable subtrajectories, the optimal overall trajectory can be determined. We present the method in an application of a camera mounted to a servo positioning system for part inspection.