Toward Learning Human-Like, Safe and Comfortable Car-Following Policies With a Novel Deep Reinforcement Learning Approach | IEEE Journals & Magazine | IEEE Xplore