Deep Q-learning using redundant outputs in visual doom | IEEE Conference Publication | IEEE Xplore