Adversarial Behavioral Exploration for Safe Reinforcement Learning (AdvEx-RL)

This is the code repository of the paper "Adversarial Behavioral Exploration for Safe Reinforcement Learning". Paper link: IJCAI23

Instructions:

Follow the instruction written inside the READ_ME_INSTRUCTIONS.txt file

AdvEx-RL agent under 100% adversarial attack:

SAC agent under 100% adversarial attack: