A fast point-based algorithm for POMDPs
Nikos Vlassis, Matthijs T. J. Spaan · 2004
We describe a point-based approximate value iteration algorithm for partially observable Markov decision processes. The algorithm per-forms value function updates ensuring that in each iteration the new value function is an up-per bound to the previous value function, as estimated on a sampled set of belief points. A randomized belief-point selection scheme allows for fast update steps. Results indicate that the proposed algorithm achieves competitive perfor-mance, both in terms of solution quality as well as speed. 1