Institute Homepage DE Sign In

Back

Autonomous Motion Conference Paper 2001

Real-time statistical learning for robotics and human augmentation

Thumb ticker sm ss
Autonomous Motion
Stefan Schaal
  • Director

Real-time modeling of complex nonlinear dynamic processes has become increasingly important in various areas of robotics and human augmentation. To address such problems, we have been developing special statistical learning methods that meet the demands of on-line learning, in particular the need for low computational complexity, rapid learning, and scalability to high-dimensional spaces. In this paper, we introduce a novel algorithm that possesses all the necessary properties by combining methods from probabilistic and nonparametric learning. We demonstrate the applicability of our methods for three different applications in humanoid robotics, i.e., the on-line learning of a full-body inverse dynamics model, an inverse kinematics model, and imitation learning. The latter application will also introduce a novel method to shape attractor landscapes of dynamical system by means of statis-tical learning.

Author(s): Schaal, S. and Vijayakumar, S. and D’Souza, A. and Ijspeert, A. and Nakanishi, J.
Book Title: International Symposium on Robotics Research
Year: 2001
Editors: Jarvis, R. A.;Zelinsky, A.
Bibtex Type: Conference Paper (inproceedings)
Address: Lorne, Victoria, Austrialia Nov.9-12
URL: http://www-clmc.usc.edu/publications/S/schaal-ISRR2001.pdf
Cross Ref: p1490
Electronic Archiving: grant_archive
Note: clmc

BibTex 

@inproceedings{Schaal_ISRR_2001,
  title = {Real-time statistical learning for robotics and human augmentation},
  booktitle = {International Symposium on Robotics Research},
  abstract = {Real-time modeling of complex nonlinear dynamic processes has become increasingly important in various areas of robotics and human augmentation. To address such problems, we have been developing special statistical learning methods that meet the demands of on-line learning, in particular the need for low computational complexity, rapid learning, and scalability to high-dimensional spaces. In this paper, we introduce a novel algorithm that possesses all the necessary properties by combining methods from probabilistic and nonparametric learning. We demonstrate the applicability of our methods for three different applications in humanoid robotics, i.e., the on-line learning of a full-body inverse dynamics model, an inverse kinematics model, and imitation learning. The latter application will also introduce a novel method to shape attractor landscapes of dynamical system by means of statis-tical learning.},
  editors = {Jarvis, R. A.;Zelinsky, A.},
  address = {Lorne, Victoria, Austrialia Nov.9-12},
  year = {2001},
  note = {clmc},
  slug = {schaal_isrr_2001},
  author = {Schaal, S. and Vijayakumar, S. and D'Souza, A. and Ijspeert, A. and Nakanishi, J.},
  crossref = {p1490},
  url = {http://www-clmc.usc.edu/publications/S/schaal-ISRR2001.pdf}
}