Single-level approach to estimate origin-destination matrix: exploiting turning proportions and partial OD flows

This research proposes a novel single-level approach to traffic counts-based OD estimation using observed turning proportions and partial OD flows. The key idea is to relax the need for explicit assignment-based bi-level models and exploit the observations from big traffic data. The observed turning proportions are used to estimate traffic counts; and the structural knowledge of partial OD flows is used to improve the quality of OD estimates. Numerical experiments considering different spatial coverage of partial OD from Bluetooth are conducted on two networks. The study findings indicated that the quality of estimates improve with higher percentage of partial OD information. The proposed approach is computationally faster and yielded slightly better results than the traditional bi-level. The methodology is not limited to data from Bluetooth but adaptive with other sources such as GPS, mobile phone, and travel surveys that can provide turning proportions and/or partial OD flows.