Publications

Sort by: [year] [type] [author]

Speeding up Reinforcement Learning using Manifold Representations: Preliminary Results

Robert Glaubius, Motoi Namihira, and William D. Smart.
In "Proceedings of the IJCAI 2005 Workshop on Reasoning with Uncertainty in Robotics (RUR 05)", Edinburgh, Scotland, July 2005.

Reinforcement Learning (RL) has proven to be a useful set of techniques for planning under uncertainty in robot systems. Effective RL algorithms for this domain must be able to deal with large, continuous state spaces, and must make efficient use of experience. In this paper, we present methods to better leverage observed experience by reusing these experiences across parts of the problem state space that are known to be similar. We present experimental results in a simple goal-based navigation domain. We also present an preliminary approach to identifying portions of the world that appear similar based on observed transition samples.

Paper: [PDF]

@inproceedings{ijcai05,
  author = {Glaubius, Robert and Namihira, Motoi and Smart, William D.},
  title = {Speeding up Reinforcement Learning using Manifold Representations: Preliminary Results},
  booktitle = {Proceedings of the {IJCAI} 2005 Workshop on Reasoning with Uncertainty in Robotics ({RUR 05})},
  address = {Edinburgh, Scotland},
  month = {July},
  year = {2005}
}