A probabilistic estimation of traffic congestion using Bayesian network

For ensuring a robust traffic management system, monitoring traffic conditions promptly by estimating the congestion level is crucial. The current measures can only represent the variations of specific standard parameters and do not consider the probabilistic property. In this paper, a Bayesian Network (BN) based probabilistic congestion estimation approach is proposed. The proposed BN-based approach considers both speed and volume related measures and provides a probabilistic estimation of the probable congestion state. For recurring and nonrecurring congestion, two different BN models were developed and implemented in realtime datasets. The case study results showed that the proposed BN models could quantify the probable congestion level in terms of a probability for each state in a variable, at the presence of different combinations of prior variables' state. Further, the proposed BN based approach can be employed in the decision-making process that involves the probabilistic estimation of traffic congestion with a vision of the realtime circumstances.