Image based visual servoing with kinematic singularity avoidance for mobile manipulator

Abstract

This article presents an implementation of visual servoing (VS) for a redundant mobile manipulator in an eye-in-hand configuration. We used the image based visual servoing (IBVS) scheme, which means the pose control of the robot is based on the error features in the image of a camera. Conventional eye-in-hand VS requires the inversion of a Jacobian matrix, which can become rank deficient, provoking kinematic singularities. In this work, the inversion of the Jacobian matrix is solved using damped least squares (DLS) to reduce singularities and smooth out discontinuities. In addition, a task prioritization scheme is proposed where a primary task performs the eye-in-hand IBVS task, and a secondary task maximizes a manipulability measure to avoid singularities. Finally, a gravity compensation term is also considered and defined on the basis of the image space error. The effectiveness of the proposed algorithm is demonstrated through both simulation and experimental results considering the Kuka YouBot.