Vision-based reinforcement learning using approximate policy iteration | IEEE Conference Publication | IEEE Xplore