A Variable Impedance Scheme Based on Power-Shaping Signals and Partial Knowledge of Link-Side Dynamics for Flexible-Joint Robot Interaction and Tracking Control

This article proposes a novel scheme—based on power-shaping control (PSC)—that can endow flexible-joint robots with both tracking, and interactional, capabilities. In virtue of relying upon the PSC method, this approach entails modest modeling requirements restricted to computation of the gravitational torque vector, and motor-side dynamics terms (typically available in manufacturer datasheets). Hence, it distinguishes itself by obviating the need for calculation of computationally cumbersome link-side dynamics elements, such as the Coriolis and link inertia matrices. In contrast to analogous schemes, the highest order term required by the proposed design is the third derivative of the motor position vector; it therefore avoids the usage of link-jerk feedback that can be detrimental to interactional performance. Moreover, the propounded framework enables utilisation of noncollocated feedback for enhanced tracking accuracy, as well as variable impedance control for interactional performance augmentation. The aforesaid features are effectuated without any reliance on coordinate transformations; thus, the original dynamical model's structure remains immutable throughout. Also, the proposed design's complexity is dependent solely on the gravitational torque vector's dimension (i.e., not on the link inertia or Coriolis terms). Experimental results involving a flexible-joint robot, namely the Rethink Robotics Baxter, corroborate the theoretical analyses, in addition to demonstrating that interactional performance improvements can be achieved via the proposed methodology.