Stability-Guaranteed Control via Divergent-Component-of-Motion Feedback for Force-Based Balancing in Articulated-Soft Floating-Base Robots

Abstract

This paper aims at unifying simplified models of balancing – based on linear, centre-of-mass (CoM) expressions – and articulated-soft floating-base (ASFB) robot dynamics models, for control design. Two distinct controller variations are introduced: the first operates by mapping gross applied force references directly to the joint space, while the second relies upon converting CoM impedances to joint torques. Moreover, it is revealed that variable, Cartesian CoM impedance gains can be stably utilised via these schemes. In both cases, the user need only specify a desired Divergent Component of Motion (DCM) trajectory, and the remaining controller terms are computed automatically via the proposed approach. Since the ASFB model accounts for the passive elasticity present in the robot's joints, and, therefore, the motor-side dynamics, the proposed schemes incorporate full-state feedback signals to construct the various controller terms. In spite of this added level of complexity (due to the underactuation), the power shaping-signal control (PSC) method enables the development of controllers that can demonstrably ensure convergence to referential DCM positions, while considering the full ASFB dynamics; this effectively proves stability and – for the first time – balance in the sense of Lyapunov. The proposed methods are corroborated through practical push recovery experiments involving the Bipedal Robot Unit with Compliance Enhanced (BRUCE), which testify to stability/balance maintenance during the execution of DCM regulation/tracking, and disturbance rejection, tasks.