Practical and Theoretical Research into
Humanoid Motion and Interaction
by Joshua G. Hale

Figure 1: `CB' 31 DOF Humanoid robot (produced by SARCOS), and simulated jumping motion (vertical lines correspond to contact forces)

Introduction

There is a symbiotic relationship between the development of humanoid robotics and the study of human behavior. Principles and hypotheses regarding human skills such as visual processing and motor control inform and inspire research into replicating such processes with robot technology. Developments in humanoid robotics make a corresponding return to these fields in terms of embodied computational implementations that facilitate new experimental paradigms, concretely reveal that computational complexity and plausibility of neural processing strategies, and inspire new hypotheses regarding human capabilities.

A central theme of our research has been the study of human behavior for and through humanoid robotics and vice versa. The development of humanoid robotic technology has been aided by the consideration of biologically plausible methods, and likewise, the computational feasibility and effectiveness of such methods for humans themselves has been explored through practical implementation. In particular we have explored the learning (14), planning and performance of motion, trajectory formation (16), computational tools facilitating motion production and analysis (13), human perception of motion (18, 28, 17, 2, 29), and concepts of human-robot interaction (15).One fundamental component of our research has been synthetic motion generation. In the following exposition we demonstrate its relevance to robotics, animation, human motion production and perceptual psychology.

Among the tools for generating and analyzing motion that have been central to our work, dynamic simulation is the most broadly applicable and is particularly important for robotic and simulated humanoids. Our most recent research has been focused on improving simulation fidelity and computational efficiency in dynamic simulation. In particular we discuss below an effect boosting analytical technique for resolving contact forces with static and dynamic friction, and a kinematic configuration manipulation technique intended to simplify the treatment of the same problem for articulated bodies such as humanoids. We have developed motion controllers for tasks such as walking and jumping by means of these techniques (see Figures 1 & 2).

Figure 2: Walking controller for simulated humanoid robot `CB'

[ read full article ]