English

Stochastic Approximation for Risk-aware Markov Decision Processes

Optimization and Control 2019-12-05 v4 Artificial Intelligence

Abstract

We develop a stochastic approximation-type algorithm to solve finite state/action, infinite-horizon, risk-aware Markov decision processes. Our algorithm has two loops. The inner loop computes the risk by solving a stochastic saddle-point problem. The outer loop performs QQ-learning to compute an optimal risk-aware policy. Several widely investigated risk measures (e.g. conditional value-at-risk, optimized certainty equivalent, and absolute semi-deviation) are covered by our algorithm. Almost sure convergence and the convergence rate of the algorithm are established. For an error tolerance ϵ>0\epsilon>0 for the optimal QQ-value estimation gap and learning rate k(1/2,1]k\in(1/2,\,1], the overall convergence rate of our algorithm is Ω((ln(1/δϵ)/ϵ2)1/k+(ln(1/ϵ))1/(1k))\Omega((\ln(1/\delta\epsilon)/\epsilon^{2})^{1/k}+(\ln(1/\epsilon))^{1/(1-k)}) with probability at least 1δ1-\delta.

Keywords

Cite

@article{arxiv.1805.04238,
  title  = {Stochastic Approximation for Risk-aware Markov Decision Processes},
  author = {Wenjie Huang and William B. Haskell},
  journal= {arXiv preprint arXiv:1805.04238},
  year   = {2019}
}

Comments

34 pages, 4 figures, 2 tables

R2 v1 2026-06-23T01:51:39.610Z