Linearly-solvable Markov decision problems
–Neural Information Processing Systems
We introduce a class of MPDs which greatly simplify Reinforcement Learning. They have discrete state spaces and continuous control spaces.
Neural Information Processing Systems
Dec-31-2007