Current point-based planning algorithms for solving partially observable Markov decision processes (POMDPs) have demonstrated that a good approximation of the value function can be derived by interpolation from the values ...