Skip to Main Content
Interactive dynamic influence diagrams (I-DIDs) are graphical models of sequential decision-making in uncertain multi-agent setting. Algorithms for solving I-DIDs face the challenge of an exponentially growing space of candidate models over time. In this paper, we discuss a class of candidate models that are auto orphism POMDPs and present a method of solving candidate models of I-DIDs. We do this by removing the models which has permutable belief and selecting a representative set of candidate models. Next, we permute the solution of representative model to represent the solution of removed model. Further, we give approximate algorithms and discuss the error bound of the approximately technique and demonstrate its empirical performance.