Cheng Yang

We propose Fuz-RL, a fuzzy logic-based framework that enhances robustness of safe reinforcement learning by measuring and balancing uncertainties in a non-additive way, achieving better safety and performance under uncertainty.