Safe Reinforcement Learning via Probabilistic Logic Shields
This addresses the problem of ensuring safety in reinforcement learning for applications like robotics or autonomous systems, though it is incremental as it builds on existing shielding methods.
The paper tackles the challenge of integrating logical safety specifications with continuous, end-to-end deep reinforcement learning by introducing Probabilistic Logic Policy Gradient (PLPG), which models safety constraints as differentiable functions and learns safer and more rewarding policies compared to state-of-the-art shielding techniques.
Safe Reinforcement learning (Safe RL) aims at learning optimal policies while staying safe. A popular solution to Safe RL is shielding, which uses a logical safety specification to prevent an RL agent from taking unsafe actions. However, traditional shielding techniques are difficult to integrate with continuous, end-to-end deep RL methods. To this end, we introduce Probabilistic Logic Policy Gradient (PLPG). PLPG is a model-based Safe RL technique that uses probabilistic logic programming to model logical safety constraints as differentiable functions. Therefore, PLPG can be seamlessly applied to any policy gradient algorithm while still providing the same convergence guarantees. In our experiments, we show that PLPG learns safer and more rewarding policies compared to other state-of-the-art shielding techniques.