deep reinforcement learning (RL) has emerged as a promising approach for
autonomously acquiring complex behaviors from low level sensor observations.
Although a large portion of deep RL research has focused on applications in
video games and simulated control, which does not connect wi