On reinforcement learning for robots

By Leslie Pack Kaelbling. 1996

Subfields: Reinforcement Learning, Robotics & Control

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Reinforcement Learning

1. Hado van Hasselt
2. Sergey Levine
3. J. Urgen Schmidhuber
4. Peter Stone
5. Matthew E. Taylor
6. Ronald J. Williams
7. Richard S. Sutton
8. Gavin Adrian Rummery

In Robotics & Control

1. Mahesan Niranjan
2. Ronald J. Williams
3. Gavin Adrian Rummery
4. Jing Peng

Citations in the corpus

• On-line Q-learning using connectionist systems (1994, background): “Lin (1993) examined the use of neural networks in this context, and provided algorithms for training based on the idea of backward replay, which involves storing the state-action pairs visited by the system until the end of a trial when the rewards are known, and then replaying them in reverse…”
• Incremental multi-step Q-learning (1994, background): “It should be pointed out that both the xed period learning process of Watkins (Watkins 1989) for sufciently long learning periods and the action-replay process of Lin (Lin 1992) produce the same beneecial eeects as that of Q()-learning.”