A gestural language for a humanoid robot

摘要

Figure 1: (left) Cog, the humanoid research platform. (right) A kinematic schematic of the 8-DOF used in this work. The Problem: Humanoid research has long been concerned with the quality of the robot’s movement. However, obtaining the elusive tempo and grace of the human motor systemhasprovento be a very difficult problem. The complexity of controlling high degree of freedom (DOF) humanoid robots, combined with insights provide by neurophysiological findings, has lead researchers to look at motor primitives [6] as an organizing methodology. We are investigating a data-driven approach to motor primitives in building a motor language for the humanoid robot Cog. The proposed model is implemented on Cog and applied to the task of human motor mimicry. Motivation: The difficulty of the posed problem comes in part from the complexity of the human motor system. The human arm, for example, contains seven degrees of freedom and 26 muscle groups. It is a highly redundant system. A task such as pointing to an object in the world can yield an infinite number of solutions, yet the human system utilizes just a few. This in itself suggests that it is necessary to look for methods to simplify and structure the problem in such a way that it becomes tractable. The approach taken in this work can be considered a variant of what are commonly referred to as motor primitives. The crux of this approach, supported by biological findings, is that we compose complex motor acts out of a small set of simpler motor acts. This small set of motor primitives provides the building blocks of our complete