Hierarchical Linearly-Solvable Markov Decision Problems

Anders Jonsson, Vicenç Gómez

We present a hierarchical reinforcement learning framework that formulates each task in the hierarchy as a special type of Markov decision process for which the Bellman equation is…