
This paper proposes considering robust reinforcement learning (RRL) problems within the framework of zero-sum positional differential games, which allows for the development of centralized Q-learning approaches with theoretically justified intuition.


coremsg

robust-reinforcement-learning-via-zero-sum-positional-differential-games-a-deep-q-learning-approach


Robust Reinforcement Learning via Zero-Sum Positional Differential Games: A Deep Q-Learning Approach
