An on-line algorithm for dynamic reinforcement learning and planning in reactive environments | IEEE Conference Publication | IEEE Xplore