Visually servoed deformation control by robot manipulators

Abstract

Despite the recent progress in physically interactive and surgical robotics, the active deformation of compliant objects remains an open problem. The main obstacle comes from the difficulty to identify/estimate the object's deformation properties. This paper presents a new visually servoed deformation controller for unknown elastic objects. The control law is designed using the passivity-based framework. The proposed method exploits visual feedback to iteratively estimate the deformation Jacobian matrix, avoiding any identification steps. We prove that even in the presence of inexact estimations, the controller ensures input-to-state stability (i.e. dissipativity) with respect to time-varying disturbances. Finally, an experimental study with several deformation tasks is presented to validate the theory.