Virtual Masses Description Forming Conservative System for Periodic Trajectory Modification of Robust Intermittent Controller

Abstract

An intermittent controller, which alternates between phases of no control and feedback control, provides a robust, computationally and energetically efficient method by leveraging the robot's passive dynamics. However, this approach, especially during rapid motions and when facing disturbances, becomes unstable due to model errors during the single support phase; this is the no-control phase. This letter proposes a novel multi-mass model that enables intermittent controllers to reduce model errors by periodically modifying the trajectory for the no-control phase. To this end, we introduce a virtual masses description that enables trajectory modification, forming a conservative system. Its trajectory is adjusted directly before the no-control phase, ensuring stability against disturbances. As a result, robots using our model demonstrated robust walking in both simulations and hardware experiments, even when faced with varying external disturbances and obstacles, while also exhibiting improved efficiency.