Dynamic and Versatile Humanoid Walking via Embedding 3D Actuated SLIP Model With Hybrid LIP Based Stepping

Abstract

In this paper, we propose an efficient approach to generate dynamic and versatile humanoid walking with nonconstant center of mass (COM) height. We exploit the benefits of using reduced order models (ROMs) and stepping control to generate dynamic and versatile walking motion. Specifically, we apply the stepping controller based on the Hybrid Linear Inverted Pendulum Model (H-LIP) to perturb a periodic walking motion of a 3D actuated Spring Loaded Inverted Pendulum (3D-aSLIP), which yields versatile walking behaviors of the 3D-aSLIP, including various 3D periodic walking, fixed location tracking, and global trajectory tracking. The 3D-aSLIP walking is then embedded on the fully-actuated humanoid via the task space control on the COM dynamics and ground reaction forces. The proposed approach is realized on the robot model of Atlas in simulation, wherein versatile dynamic motions are generated.

Risk-Averse Control via CVaR Barrier Functions: Application to Bipedal Robot Locomotion

“…However, the S2S dynamics P h (18) cannot be obtained in analytical form due to the nonlinear and hybrid dynamics of the robot walking. Our previous work [33] suggests that the S2S dynamics of the walking of the Hybrid-Linear Inverted Pendulum (H-LIP) Fig. 3.…”

“…Planning and controlling of bipedal walking are challenging problems, and there has been various related approaches [32] in the literature. In this paper, we apply the approach in [33] to approximate the bipedal walking dynamics via a discrete linear system subject to stochastic uncertainty.…”

“…Let x h = [c, p, v] T denote the horizontal state, where c is the horizontal position of the center of mass (COM) of the robot relative to the inertia frame, p is the horizontal position of the COM relative to its stance foot, and v is the horizontal velocities of the COM. Then, the horizontal step-to-step (S2S) dynamics [33] of the horizontal COM state is represented as…”

“…T is the discrete preimpact state of the H-LIP, and u t H-LIP is the step size. The expressions of A, B can be found in [33]. By approximation, (18) can be rewritten as…”

Sequential Motion Planning for Bipedal Somersault via Flywheel SLIP and Momentum Transmission with Task Space Control

“…In our previous work [28], [34], we applied the control Lyapunov function based quadratic programs (CLF-QPs) [35] for output stabilization with momentum outputs. A different optimization-based control, i.e., task space control (TSC) [36], [37], [38], [39], can serve the same purpose of stabilization. Here we derive TSC with minimum variable formulation for illustration.…”