Robotics, Learning and Autonomy at Brown (r,lab)
Manipulation Manifolds: Uncovering sensorimotor structure from teleoperation

Odest Chadwicke Jenkins
Alan Peters
Bobby Bodenheimer

Abstract

We explore the use of manifold learning techniques to uncover structure in sensorimotor time series from teleoperated humanoid manipulation tasks. Data from Robonaut, NASA's humanoid robot, was recorded while it was being teleoperated through four tool manipulation tasks. We show that one algorithm, Spatio-Temporal Isomap, is capable of uncovering behavioral structures that can be difficult to find with other dimension reduction techniques (Principal Component Analysis, Multidimensional Scaling, and Isomap). Through uncovering sensorimotor structure, we aim to provide a useful diagnostic tool for task analysis. In particular, we show that for two of the four tool manipulation tasks, projection onto to the manifold determined by Spatio-Temporal provides classification of the task into successes or failures at rates exceeding standard Support Vector Machines. One corollary of these results is that we demonstrate that both motor and sensory data are important to understanding the behavior of the robot. These techniques may be useful in providing operator feedback to improve robot performance.

Papers

R. A. Peters and O. C. Jenkins. Uncovering manifold structures in Robonaut's sensory-data state space. In IEEE International Conference on Humanoid Robotics (Humanoids 2005), Tsukuba, Japan, Dec 2005.
[ bib | .pdf ]

O. Jenkins, R. Bodenheimer, and R. Peters. Manipulation manifolds: Explorations into uncovering manifolds in sensory-motor spaces. In International Conference on Development and Learning (ICDL 2006), Bloomington, IN, USA, May-Jun 2006.
[ bib | .pdf ]

R. Peters, O. Jenkins, and R. Bodenheimer. Sensory-motor manifold structure induced by task outcome: Experiments with Robonaut. In IEEE International Conference on Humanoid Robotics (Humanoids 2006), Genoa, Italy, Dec 2006.
[ bib | .pdf ]

O. Jenkins, R. Peters, and R. Bodenheimer. Uncovering success in manipulation. In Robotics: Science and Systems Workshop on Manipulation in Human Environments, Philadelphia, PA, USA, Aug 2006.
[ bib | .pdf ]

Sponsors

DARPA MARS Program