A temporal reasoning method based on maximum a posteriori estimation in situation assessment

Abstract

Stochastic temporal reasoning in situation assessment (SA) is very important in many applications. The approach based on Maximum Likelihood Estimation (MLE) treats the unknown temporal variable as a constant which doesn't use a priori information and generates a larger estimate variance. In this paper, the relation model of known temporal information and unknown temporal variable has been established which can also be used to MLE-based method. In the model, the forward and backward reasoning algorithm about time instants has been derived by treating the unknown temporal variable as random variable and introducing MAP estimation into temporal reasoning. The performance analysis between MAP-based method and MLE-based method shows that under some conditions, the estimate variance of MAP-based method is lower than MLE-based method, and we have given these conditions by experiments.