Exploiting Domain Symmetries in Reinforcement Learning With Continuous State and Action Spaces

doi 10.1109/icmla.2009.41