An intelligent agency requires the capability to predict what the world looks like as a consequence of its actions. It also needs to explain present observations in order to infer previous states. This thesis proposes an approach to realize both capabilities, that is prediction and postdiction based on temporal information. In particular, there is always some uncertainty in the knowledge about the world which the autonomous agent inhabits. Therefore we handle uncertainty using probability theory. None of the previous works dealing with quantitative (or numerical) approaches addressed on the postdiction problem in designing an intelligent agent. Our thesis presents a method to resolve this postdiction problem under uncertainty.