Real-time estimation and prediction of origin-destination matrices per time slice.

This paper presents a general approach to real-time estimation and prediction of Origin-Destination (OD) matrices per time slices. The paper presents numerical examples using real data given by hourly aggregation of information available from toll stations in Italy. The results show that prediction procedures based on an indirect approach can perform almost as well as direct prediction procedures. This result is relevant to applications to networks where actual OD matrices per time slices are not directly observable and can only be estimated. Direct procedures are not based on directly observable data. The indirect approach firstly filters traffic counts for the future time slices, and then computes demand matrices.