Model-Free Robust $\phi$-Divergence Reinforcement Learning Using Both Offline and Online Data
Abstract
The robust -regularized Markov Decision Process (RRMDP) framework focuses on designing control policies that are robust against parameter uncertainties due to mismatches between the simulator (nominal) model and real-world settings. This work makes two important contributions. First, we propose a model-free algorithm called Robust -regularized fitted Q-iteration (RPQ) for learning an -optimal robust policy that uses only the historical data collected by rolling out a behavior policy (with robust exploratory requirement) on the nominal model. To the best of our knowledge, we provide the first unified analysis for a class of -divergences achieving robust optimal policies in high-dimensional systems with general function approximation. Second, we introduce the hybrid robust -regularized reinforcement learning framework to learn an optimal robust policy using both historical data and online sampling. Towards this framework, we propose a model-free algorithm called Hybrid robust Total-variation-regularized Q-iteration (HyTQ: pronounced height-Q). To the best of our knowledge, we provide the first improved out-of-data-distribution assumption in large-scale problems with general function approximation under the hybrid robust -regularized reinforcement learning framework. Finally, we provide theoretical guarantees on the performance of the learned policies of our algorithms on systems with arbitrary large state space.
Cite
@article{arxiv.2405.05468,
title = {Model-Free Robust $\phi$-Divergence Reinforcement Learning Using Both Offline and Online Data},
author = {Kishan Panaganti and Adam Wierman and Eric Mazumdar},
journal= {arXiv preprint arXiv:2405.05468},
year = {2024}
}
Comments
To appear in the proceedings of the International Conference on Machine Learning (ICML) 2024