Passivity-Based Teleoperation With Variable Rotational Impedance Control

In this work, we present a novel bilateral teleoperation architecture with variable impedance control for orientational contact tasks. We exploit Unit Quaternions and tools from Lie Theory to model and manipulate robot orientations, as well as Learning-from-Demonstration to learn a stiffness adaptation policy from the demonstrated task dynamics. The learnt policy then shapes the rotational stiffness of the remote robot during contact with the environment. We also present a passivity analysis where we use energy tanks to guarantee the passivity of the closed loop system, and hence the stable interaction. Our approach is validated on real robot hardware in a cutting task along a curve, and in a user study.