The HMDP Planner for Planning with Probabilities

HMDPP is a probabilistic planner that computes two heuristic values for each state and chooses an action to apply through bounded heuristic search methods. One of these heuristics is the value of the well-known hadd heuristic on a relaxation of the underlying MDP in which probabilities are compiled into costs, and the other is computed from an iteratively generated pattern database based on the idea of mutex sets of literals, as computed by the h2 heuristic.