Humanoid Balancing Behavior Featured by Underactuated Foot Motion

Abstract

A novel control synthesis is proposed for humanoids to demonstrate unique foot tilting behaviors comparable to humans in balance recovery. Our study of model based behaviors explains the underlying mechanism and the significance of foot tilting well. Our main algorithms are composed of impedance control at the center of mass, virtual stoppers that prevents over-tilting of the feet, and postural control for the torso. The proof of concept focuses on the sagittal scenario and the proposed control is effective to produce human-like balancing behaviors characterized by active foot tilting. The successful replication of this behavior on a real humanoid proves the feasibility of deliberately controlled underactuation. The experimental validation was rigorously performed, and the data from the sub-modules and the entire control were presented and analyzed.

Learning Whole-Body Motor Skills for Humanoids

“…The first three strategies, controlling ankle torque, angular momentum around the Center of Mass (CoM), and the timing and position of steps, are analyzed with respect to their ability to reject disturbances in [11]. A control framework for the foot-tilting strategy has been proposed in [12] demonstrating a humanoid's ability to use foot-tilting for push recovery. Traditionally control schemes can be divided into predictive schemes which calculate reference motions, and reactive schemes which respond to sudden disturbances.…”

Emergence of human-comparable balancing behaviours by deep reinforcement learning

“…Li et al have done a thorough analysis on the dynamics of foot tilting and derived a foot tilting (ankle push-off) balancing strategy. They explained the underlying mechanism and the significance of foot tilting [3]. Concrete physical and mathematical proof in [3] suggest that foot tilting balance strategy is more robust against force perturbations than flat foot balance strategy, and have successfully designed a controller capable of underactuated foot tilting and implemented it on a real robot.…”

“…where the impulse J reject derived by capture point is the theoretical maximum of the impulse that can be rejected, where z c is the COM height, ∆ COP is the relative horizontal distance between the constant COP and the initial COM position, m is the total mass of the inverted pendulum [3]. The capture point as an indication of balance is considered to be within the support polygon, so the maximum reachability of the capture point is at the edge of the foot.…”

“…Biomechanical study of human walking has discussed about the advantage of rolling around the heel and toe during walking phase [1]. From a biomechanical point of view, tilting the foot creates better footground clearance allowing the maximum ankle torques to be exploited [2], [3].…”

“…Most ZMP based methods will fail during underactuation phases, as they need to restrict the ZMP or COP to be within the support polygon created by a flat foot and away from the foot edges. Controllers that permits the COP to lie on the narrow boundary of the foot have been developed to generate underactuated foot tilting behaviors, demonstrating the feasibility of using analytic engineering approach for designing controllers capable of dealing with underactuated phases during balance recovery [3].…”