Passivity Preservation for Variable Impedance Control of Compliant Robots

The notable performance exhibited by impedance controllers during robotic interaction, has led to widespread use of this control methodology. Improved position and interaction control might be attainable, through utilisation of Variable Impedance Control (VIC) techniques. Interactional performance could be further improved by combining structural compliance with VIC. However, utilisation of VIC tends to induce energy-injecting elements, which could impact on a robotic system's stability/passivity. Additionally, implementation of active VIC techniques on passively-compliant robots has not been investigated (although several works consider VIC using Variable Stiffness Actuators, VSAs), which renders the existing rigid-joint robot, passivity-inducing control schemes [1][2][3], inapplicable to compliant systems. To this end, the work presented here introduces a novel scheme, termed Passivity-Preservation Control (PPC), which suppresses the energy injections that could be introduced into compliant robots, as a result of VIC. Compared to tank-based VIC approaches [1][2], the PPC scheme is directly applicable to flexible-joint robots, even ones with non-linear passive stiffness elements, while its performance is independent of the tank-energy levels [3]. Moreover, the proposed scheme permits stable VIC using full-state feedback, thereby enabling impedance modulations relating to both motor, and link-side variables. Consequently, full-state feedback gains can be generated via Linear Quadratic Regulator (LQR) optimisation, thus enabling application of gain-scheduling techniques on flexible-joint robots for enhanced position control. Passivity and stability analyses are performed for joint and Cartesian-space versions of the PPC scheme, which justify their applicability to both interaction and ‘free-motion’ scenarios. The PPC scheme's efficacy, compared to constant gain impedance methods, in terms of convergence and interactional performance, is corroborated via simulation and experimental means involving the Sawyer robot, which is powered by Series Elastic Actuators (SEAs). Theoretical and experimental results mathematically and practically verify VIC stability, thus enabling flexible-joint robots to more accurately mimic biologically-inspired behaviours [4].