Institute Homepage
Institute Homepage DE Sign In

Back

Autonomous Motion Empirical Inference Conference Paper 2007

Using reward-weighted regression for reinforcement learning of task space control

Thumb ticker sm 12009745 10103538825457245 7502907506146263960 n
Empirical Inference
Jan Peters
Research Group Leader
Thumb ticker sm ss
Autonomous Motion
Stefan Schaal
  • Director

In this paper, we evaluate different versions from the three main kinds of model-free policy gradient methods, i.e., finite difference gradients, `vanilla' policy gradients and natural policy gradients. Each of these methods is first presented in its simple form and subsequently refined and optimized. By carrying out numerous experiments on the cart pole regulator benchmark we aim to provide a useful baseline for future research on parameterized policy search algorithms. Portable C++ code is provided for both plant and algorithms; thus, the results in this paper can be reevaluated, reused and new algorithms can be inserted with ease.

Author(s): Peters, J. and Schaal, S.
Book Title: Proceedings of the 2007 IEEE International Symposium on Approximate Dynamic Programming and Reinforcement Learning
Pages: 262-267
Year: 2007
Bibtex Type: Conference Paper (inproceedings)
Address: Honolulu, Hawaii, April 1-5, 2007
DOI: 10.1109/ADPRL.2007.368197
URL: http://www-clmc.usc.edu/publications/P/peters-ADPRL2007.pdf
Cross Ref: p2672
Electronic Archiving: grant_archive
Note: clmc

BibTex 

@inproceedings{Peters_PIISADPRL_2007,
  title = {Using reward-weighted regression for reinforcement learning of task space control},
  booktitle = {Proceedings of the 2007 IEEE International Symposium on Approximate Dynamic Programming and Reinforcement Learning},
  abstract = {In this paper, we evaluate different versions from the three main kinds of model-free policy gradient methods, i.e., finite difference gradients, `vanilla' policy gradients and natural policy gradients. Each of these methods is first presented in its simple form and subsequently refined and optimized. By carrying out numerous experiments on the cart pole regulator benchmark we aim to provide a useful baseline for future research on parameterized policy search algorithms. Portable C++ code is provided for both plant and algorithms; thus, the results in this paper can be reevaluated, reused and new algorithms can be inserted with ease.},
  pages = {262-267},
  address = {Honolulu, Hawaii, April 1-5, 2007},
  year = {2007},
  note = {clmc},
  slug = {peters_piisadprl_2007},
  author = {Peters, J. and Schaal, S.},
  crossref = {p2672},
  url = {http://www-clmc.usc.edu/publications/P/peters-ADPRL2007.pdf}
}