Worst Cases Policy Gradients

TL;DR

This paper introduces a risk-sensitive actor-critic reinforcement learning framework that models future uncertainty to optimize policies based on conditional Value-at-Risk, improving safety and generalization in driving simulations.

Contribution

It presents a novel actor-critic method that incorporates future return uncertainty to learn risk-aware policies with dynamic risk levels.

Findings

Risk-averse policies outperform others in safety-critical tasks.

The approach enables dynamic risk-based decision making.

Policies generalize better across different simulation parameters.

Abstract

Recent advances in deep reinforcement learning have demonstrated the capability of learning complex control policies from many types of environments. When learning policies for safety-critical applications, it is essential to be sensitive to risks and avoid catastrophic events. Towards this goal, we propose an actor-critic framework that models the uncertainty of the future and simultaneously learns a policy based on that uncertainty model. Specifically, given a distribution of the future return for any state and action, we optimize policies for varying levels of conditional Value-at-Risk. The learned policy can map the same state to different actions depending on the propensity for risk. We demonstrate the effectiveness of our approach in the domain of driving simulations, where we learn maneuvers in two scenarios. Our learned controller can dynamically select actions along a continuous axis, where safe and conservative behaviors are found at one end while riskier behaviors are found at the other. Finally, when testing with very different simulation parameters, our risk-averse policies generalize significantly better compared to other reinforcement learning approaches.