Cascading Decomposition and State Abstractions for Reinforcement Learning | IEEE Conference Publication | IEEE Xplore