Generalized Policy Improvement Algorithms With Theoretically Supported Sample Reuse
We develop a new class of model-free deep reinforcement learning algorithms for data-driven, learning-based control. Our Generalized Policy Improvement algorithms combine the policy improvement guarantees of on-policy methods with the efficiency of sample reuse, addressing a trade-off between two important deployment requirements for real-world control: (i) practical performance guarantees and (ii) data efficiency. We demonstrate the benefits of this new class of algorithms through extensive experimental analysis on a broad range of simulated control tasks.
Authors
- James Queeney (ORCID: https://orcid.org/0000-0003-0655-3637)
- Christos G. Cassandras (ORCID: https://orcid.org/0000-0002-1625-7658)
- Ioannis Ch. Paschalidis (ORCID: https://orcid.org/0000-0002-3343-2913)
Institutions
- Boston University (US)
- University of Massachusetts Boston (US)
- Mitsubishi Electric (United States) (US)
Publication Details
- Journal
- IEEE Transactions on Automatic Control
- Published
- 2024-09-03
- DOI
- https://doi.org/10.1109/tac.2024.3454011
- Citations
- 2
- Primary Topic
- Reinforcement Learning in Robotics
- Type
- article
- Field-Weighted Citation Impact
- 0.69
Funders
- National Science Foundation
- U.S. Department of Energy
- Boston University
- Mitsubishi Electric Research Laboratories
- MathWorks
- National Institutes of Health
- Office of Naval Research
- Division of Electrical, Communications and Cyber Systems
- Air Force Office of Scientific Research