Path Overlapping Problem in Estimating Origin-Destination Matrix Using Stochastic User Equilibrium

Estimating the origin-destination (OD) matrix based on traffic counts in a transportation network has received substantial attention in recent decades. Several studies have attempted to incorporate a stochastic user equilibrium (SUE) constraint into the OD estimation framework to deal with uncertainties. These studies have mainly adopted a multinomial logit (MNL) route choice model which has a restrictive assumption that does not allow for overlapping paths. This paper addresses the path overlapping problem by employing corrected logit route choice models, namely C-logit and path-size logit (PS-logit), that partially capture similarity/correlation among paths by a correction term in the MNL structure. A gradient algorithm (developed by Spiess) is also utilized to solve the SUE-based OD matrix estimation problem. Numerical experiments on the well-known Winnipeg network show that considering correlated/overlapping paths in the OD estimation process using C-logit or PS-logit route choice models results in more accurate OD matrices than the MNL-based procedure.