Using Homomorphisms to transfer options across continuous reinforcement learning domains

http://dl.acm.org/citation.cfm?id=1597618