Iterative Reachability Estimation for Safe Reinforcement Learning

TL;DR

This paper introduces RESPO, a novel framework for safe reinforcement learning that guarantees safety while optimizing rewards in stochastic environments, with proven convergence and improved performance over existing methods.

Contribution

The paper presents RESPO, a new reachability-based framework for safe RL, along with algorithms that ensure safety guarantees and convergence in stochastic settings.

Findings

Algorithms converge to locally optimal policies.

Improved reward and safety performance over baselines.

Effective in diverse safe RL environments.

Abstract

Ensuring safety is important for the practical deployment of reinforcement learning (RL). Various challenges must be addressed, such as handling stochasticity in the environments, providing rigorous guarantees of persistent state-wise safety satisfaction, and avoiding overly conservative behaviors that sacrifice performance. We propose a new framework, Reachability Estimation for Safe Policy Optimization (RESPO), for safety-constrained RL in general stochastic settings. In the feasible set where there exist violation-free policies, we optimize for rewards while maintaining persistent safety. Outside this feasible set, our optimization produces the safest behavior by guaranteeing entrance into the feasible set whenever possible with the least cumulative discounted violations. We introduce a class of algorithms using our novel reachability estimation function to optimize in our proposed framework and in similar frameworks such as those concurrently handling multiple hard and soft constraints. We theoretically establish that our algorithms almost surely converge to locally optimal policies of our safe optimization framework. We evaluate the proposed methods on a diverse suite of safe RL environments from Safety Gym, PyBullet, and MuJoCo, and show the benefits in improving both reward performance and safety compared with state-of-the-art baselines.