Header logo is ei

Using reward-weighted imitation for robot Reinforcement Learning

2009

Conference Paper

ei


Reinforcement Learning is an essential ability for robots to learn new motor skills. Nevertheless, few methods scale into the domain of anthropomorphic robotics. In order to improve in terms of efficiency, the problem is reduced onto reward-weighted imitation. By doing so, we are able to generate a framework for policy learning which both unifies previous reinforcement learning approaches and allows the derivation of novel algorithms. We show our two most relevant applications both for motor primitive learning (e.g., a complex Ball-in-a-Cup task using a real Barrett WAM robot arm) and learning task-space control.

Author(s): Peters, J. and Kober, J.
Book Title: IEEE ADPRL 2009
Journal: Proceedings of the 2009 IEEE International Symposium on Adaptive Dynamic Programming and Reinforcement Learning (IEEE ADPRL 2009)
Pages: 226-232
Year: 2009
Month: May
Day: 0
Publisher: IEEE Service Center

Department(s): Empirical Inference
Bibtex Type: Conference Paper (inproceedings)

DOI: 10.1109/ADPRL.2009.4927549
Event Name: 2009 IEEE International Symposium on Adaptive Dynamic Programming and Reinforcement Learning
Event Place: Nashville, TN, USA

Address: Piscataway, NJ, USA
Digital: 0
Institution: Institute of Electrical and Electronics Engineers
Language: en
Organization: Max-Planck-Gesellschaft
School: Biologische Kybernetik

Links: Web

BibTex

@inproceedings{5659,
  title = {Using reward-weighted imitation for robot Reinforcement Learning},
  author = {Peters, J. and Kober, J.},
  journal = {Proceedings of the 2009 IEEE International Symposium on Adaptive Dynamic Programming and Reinforcement Learning (IEEE ADPRL 2009)},
  booktitle = {IEEE ADPRL 2009},
  pages = {226-232},
  publisher = {IEEE Service Center},
  organization = {Max-Planck-Gesellschaft},
  institution = {Institute of Electrical and Electronics Engineers},
  school = {Biologische Kybernetik},
  address = {Piscataway, NJ, USA},
  month = may,
  year = {2009},
  month_numeric = {5}
}