Robust Reinforcement Learning with Dynamic Distortion Risk Measures

TL;DR

This paper introduces a novel framework for risk-aware reinforcement learning that accounts for environmental uncertainty using dynamic distortion risk measures and neural network-based estimation, demonstrated on portfolio allocation.

Contribution

It develops a new robust risk-aware RL approach with neural network estimation, policy gradient derivation, and an actor-critic algorithm for dynamic distortion risk measures.

Findings

Effective in portfolio allocation tasks

Handles environmental uncertainty robustly

Uses neural networks for risk measure estimation

Abstract

In a reinforcement learning (RL) setting, the agent's optimal strategy heavily depends on her risk preferences and the underlying model dynamics of the training environment. These two aspects influence the agent's ability to make well-informed and time-consistent decisions when facing testing environments. In this work, we devise a framework to solve robust risk-aware RL problems where we simultaneously account for environmental uncertainty and risk with a class of dynamic robust distortion risk measures. Robustness is introduced by considering all models within a Wasserstein ball around a reference model. We estimate such dynamic robust risk measures using neural networks by making use of strictly consistent scoring functions, derive policy gradient formulae using the quantile representation of distortion risk measures, and construct an actor-critic algorithm to solve this class of robust risk-aware RL problems. We demonstrate the performance of our algorithm on a portfolio allocation example.