Robust partially observable Markov decision process

We seek to find the robust policy that maximizes the expected cumulative reward for the worst case when a partially observable Markov decision process (POMDP) has uncertain parameters whose values are