English

An Efficient Solution to s-Rectangular Robust Markov Decision Processes

Machine Learning 2023-02-01 v1 Optimization and Control

Abstract

We present an efficient robust value iteration for \texttt{s}-rectangular robust Markov Decision Processes (MDPs) with a time complexity comparable to standard (non-robust) MDPs which is significantly faster than any existing method. We do so by deriving the optimal robust Bellman operator in concrete forms using our LpL_p water filling lemma. We unveil the exact form of the optimal policies, which turn out to be novel threshold policies with the probability of playing an action proportional to its advantage.

Keywords

Cite

@article{arxiv.2301.13642,
  title  = {An Efficient Solution to s-Rectangular Robust Markov Decision Processes},
  author = {Navdeep Kumar and Kfir Levy and Kaixin Wang and Shie Mannor},
  journal= {arXiv preprint arXiv:2301.13642},
  year   = {2023}
}

Comments

arXiv admin note: substantial text overlap with arXiv:2205.14327