Reinforcement Learning with Parameterized Actions

Authors

  • Warwick Masson University of the Witwatersrand
  • Pravesh Ranchod University of the Witwatersrand
  • George Konidaris Duke University

DOI:

https://doi.org/10.1609/aaai.v30i1.10226

Keywords:

Reinforcement Learning, Parameterized Actions, Parameterized Policies

Abstract

We introduce a model-free algorithm for learning in Markov decision processes with parameterized actions—discrete actions with continuous parameters. At each step the agent must select both which action to use and which parameters to use with that action. We introduce the Q-PAMDP algorithm for learning in these domains, show that it converges to a local optimum, and compare it to direct policy search in the goal-scoring and Platform domains.

AAAI 2016 Proceedings Cover

Downloads

Published

2016-02-21

How to Cite

Masson, W., Ranchod, P., & Konidaris, G. (2016). Reinforcement Learning with Parameterized Actions. Proceedings of the AAAI Conference on Artificial Intelligence, 30(1). https://doi.org/10.1609/aaai.v30i1.10226

Issue

Vol. 30 No. 1 (2016): Thirtieth AAAI Conference on Artificial Intelligence

Section

Technical Papers: Machine Learning Methods