Improving Generalization of Reinforcement Learning with Minimax Distributional Soft Actor-Critic | IEEE Conference Publication | IEEE Xplore