A Biologically-Inspired Whole-Body Motion Control (WBMC) System for Humanoid Robots

In the past 20 years humanoid robots have made impressive improvements in the performance of non-trivial individual tasks such as walking, climbing stairs, manipulating objects etc., but when the scenario involves multiple complex actions and interactions, humans still far outperform even the most sophisticated robots. This inability to combine multiple diverse tasks to solve more complex problems limits the use of robots in real world applications. Whole-Body Control longs to fill this gap, exploiting the entire body capabilities to execute individual and simultaneous tasks. The work presented in this dissertation aims to advance the fast evolving state-of-the-art in Whole-Body Control, and to achieve this goal two main research directions were investigated. Firstly, a thorough analysis of human motion was performed, in order to better understand how humans can successfully handle composite tasks. Based on the experience gained through this analysis, the lesson learned was applied to synthesize a novel attractor-based Whole-Body Motion Control (WBMC) System, that presents a further step towards having robots be capable of operating in the real world. The analysis of human motion led to the identification of the so-called kinematic Motion Primitives (kMPs) for both periodic and discrete movements. The kMPs are defined as invariant waveforms underlying human motion, and can combine to produce more complex movements that simultaneously achieve both discrete and periodic tasks. The periodic kMPs were employed to generate by reconstruction a human-like walking for COMAN, the COmpliant huMANoid robot. The stable, dynamic, kMPs-based walking gait was then scaled in frequency, and it will be shown that "walking in the resonance" can be beneficial in terms of energy efficiency (15% of the energy required to track the kMPs-based trajectories is provided by the springs). The work on the kMPs was further extended to quadrupeds, with the identification of horse kMPs that describe walking, trotting, and galloping gaits. Horse-like trajectories for each of the three gaits were generated and applied to Cheetah-Cub, a small-size quadruped robot. A gait transition strategy was also proposed and successfully tested on the robot. The attractor-based Whole-Body Motion Control (WBMC) system was developed based on two specific features arising from the experience gained with the kMPs: simplicity and modularity. The proposed, novel WBMC System combines several unique concepts. Firstly, a computationally efficient gravity compensation algorithm for floating-base systems was derived. Secondly, a novel balancing approach was proposed, which exploits a set of fundamental physical principles from rigid multi-body dynamics, such as the overall centroidal linear and angular momentum, the joint momentum, which is a quantity that had been rarely used to control humanoid robots, as well as a minimum effort formulation. Thirdly, a set of attractors was used to implement both balance and movement features such as to avoid joint limits or to create end-effector movements. Superposing several of these attractors allows the generation of complex whole-body movements in order to perform different tasks simultaneously. The modular structure of the proposed control system easily allows extensions. The presented concepts have been validated both in simulations, and on the 29-dofs compliant torque-controlled humanoid robot COMAN, and the WBMC has proven robust to the unavoidable model errors.