This algorithm is based on the generation of compressed state space which is still a sufficient statistic for optimal control. Starting from the transition and reward matrices it produces a reformulation in terms of the compressed beliefes state which can then be solved using other approaches.
Boundend Policy Iteration (BPI)
Bounded Policy Iteration over Valuer Directed Compression (VDCBPI)
Point-based Value Iteration (PBVI)
Gradient Ascent (GA)
Stochastic Local Search (SLS)
RT-Bel
Real-time Belief Space Search, RTBSS (Paquet et al., 2005)
Online algorithm that uses branch and bound to search the state space, The lower bound is computed offline using some approximate method and used to discard sub-optimal actions.
Witness
AEMS (Ross & Chaib-draa,2007)
Perseus
Heuristic belief space search, Satia And Lave, 1973