Power-Shaping Model-Based Control With Feedback Deactivation for Flexible-Joint Robot Interaction

Abstract

This letter presents a novel control methodology that is based on an amalgamation of model-based, power-shaping control (PSC), and feedback deactivation. It has been hypothesised that proportional position control of flexible-joint robots can impinge on their interactional performance, since position feedback causes a ‘stiffening’ of the robot's joints. In order to reduce reliance upon feedback terms, this work proposes usage of feedforward, model-based control terms, whilst accounting for both the actuator and link coordinates. The introduced PSC signal enables stable tracking control, even when the proportional controller's position error term employs solely non-collocated state feedback. It is also demonstrated that the user can stably deactivate position feedback pertaining to specific joints, in a real-time manner, so as to further enhance interactional performance. However, since feedback deactivation can lead to abrupt generation of unfeasible control signals, a variable impedance control (VIC) technique is proposed to overcome this limitation. Despite VIC's potential to inject undesirable amounts of energy into the closed-loop system, an amended PSC term is introduced to guarantee stability preservation. Moreover, to achieve saturation prevention in an energy efficient manner, a novel Lyapunov function is proposed that enables unconstrained modulation of the system's active impedance gains. Experimental results involving the Rethink Robotics Baxter robot corroborate the theoretical stability analyses, in addition to demonstrating that interactional and tracking performance improvements can be achieved via the proposed methodology.