Why will the value iteration algorithm for MDPs not work for POMDPs? The value iteration algorithm requires more evidence than POMDPs can provide. The value iteration algorithm only works for continuous sets of values. The belief states are continuous in POMDPs. A state-space (in this case the belief state space) must be searched. A utility function on belief states for POMDPs will be: Convex Concave Piecewise linear Exponential Cubic How many total plans exist given |A| actions and |E| possible observations? A|A|--1 BA(-1) B. C.|A|O(E) D.|A|n|E B () C O A What kind of function is computing the number of distinct depth-d plans given a set of actions and a set of possible observations? Gaussian Exponential Linear Doubly exponential