Using dimensionality reduction to exploit constraints in reinforcement learning | IEEE Conference Publication | IEEE Xplore