Risk-Averse Total-Reward Reinforcement Learning

Kishan Panaganti (Tencent AI Lab) · Xihong Su (University of New Hampshire) · Jia Lin Hau (University of New Hampshire) · Gersi Doko (University of New Hampshire, Durham) · Marek Petrik (University of New Hampshire)
convergence guaranteesdynamic consistencyelicitabilityentropic risk measureentropic value-at-riskinfinite-horizon objectivesmarkov decision processesmodel-based algorithmsnumerical resultsoptimal stationary policyperformance guaranteesq-learningrisk-averserisk-averse value functiontabular domainstotal-reward

Risk-averse total-reward Markov Decision Processes (MDPs) offer a promising framework for modeling and solving undiscounted infinite-horizon objectives. Existing model-based algorithms for risk measures like the entropic risk measure (ERM) and entropic value-at-risk (EVaR) are effective in small problems, but require full access to transition probabilities. We propose a Q-learning algorithm to compute the optimal stationary policy for total-reward ERM and EVaR objectives with strong convergence and performance guarantees. The algorithm and its optimality are made possible by ERM's dynamic consistency and elicitability. Our numerical results on tabular domains demonstrate quick and reliable convergence of the proposed Q-learning algorithm to the optimal risk-averse value function.