Humanoid Robot Acrobatics Utilizing Complete Articulated Rigid Body Dynamics

TL;DR

This paper introduces a control architecture for humanoid robots that enables highly dynamic acrobatic motions by directly utilizing complete articulated rigid body dynamics, overcoming previous approximation limitations.

Contribution

It presents a novel control framework combining trajectory optimization and whole-body control with a matching model abstraction based on full equations of motion.

Findings

Successful simulation of acrobatic maneuvers

Effective handling of constraints and posture behaviors

Demonstrates the feasibility of physics-based control for dynamic humanoid motions

Abstract

Endowing humanoid robots with the ability to perform highly dynamic motions akin to human-level acrobatics has been a long-standing challenge. Successfully performing these maneuvers requires close consideration of the underlying physics in both trajectory optimization for planning and control during execution. This is particularly challenging due to humanoids' high degree-of-freedom count and associated exponentially scaling complexities, which makes planning on the explicit equations of motion intractable. Typical workarounds include linearization methods and model approximations. However, neither are sufficient because they produce degraded performance on the true robotic system. This paper presents a control architecture comprising trajectory optimization and whole-body control, intermediated by a matching model abstraction, that enables the execution of acrobatic maneuvers, including constraint and posture behaviors, conditioned on the unabbreviated equations of motion of the articulated rigid body model. A review of underlying modeling and control methods is given, followed by implementation details including model abstraction, trajectory optimization and whole-body controller. The system's effectiveness is analyzed in simulation.