A. Harutyunyan, S. Devlin, P. Vrancx, and A. Nowe, “Expressing Arbitrary Reward Functions as Potential-Based Advice”, AAAI, vol. 29, no. 1, Feb. 2015.