Abstract: Goal representation heuristic dynamic programming (GrHDP) is proposed in this paper to demonstrate online learning in the Markov decision process. In addition to the (external) reinforcement ...