Safe Exploration via Policy Priors
This addresses the problem of safe online learning for RL agents in real-world applications, representing a strong specific advance rather than a foundational breakthrough.
The paper tackles safe exploration in reinforcement learning by using conservative policy priors, with their SOOPER approach guaranteeing safety throughout learning and achieving state-of-the-art performance on benchmarks and real-world hardware.
Safe exploration is a key requirement for reinforcement learning (RL) agents to learn and adapt online, beyond controlled (e.g. simulated) environments. In this work, we tackle this challenge by utilizing suboptimal yet conservative policies (e.g., obtained from offline data or simulators) as priors. Our approach, SOOPER, uses probabilistic dynamics models to optimistically explore, yet pessimistically fall back to the conservative policy prior if needed. We prove that SOOPER guarantees safety throughout learning, and establish convergence to an optimal policy by bounding its cumulative regret. Extensive experiments on key safe RL benchmarks and real-world hardware demonstrate that SOOPER is scalable, outperforms the state-of-the-art and validate our theoretical guarantees in practice.