This paper compares the performance of the Howard (1960) policy iteration algorithm for infinite-horizon continuous-state Markovian decision processes (MDP's) using alternative random, quasi-random, and deterministic discretizations of the state space, or grids. Each grid corresponds to an...