Enhanced least square based dynamic OD matrix estimation using Radio Frequency Identification data

The origin-destination (OD) demand is a crucial input for traffic planning and management. The Radio Frequency Identification (RFID) is an emerging technology to collect traffic data. In this paper, the RFID data were used to derive the preliminary OD matrix and the dynamic assignment matrix by matching the RFID data collected in origins and destinations along the road links. However, the preliminary OD matrix was inaccurate due to reasons like malfunction or limited market penetration rate of equipped vehicles. To solve this problem, an optimization model was developed based on a least squares model. The optimization model estimated dynamic OD matrix by integrating the preliminary OD matrix, dynamic assignment matrix, and link flows detected by the inductive loop detectors. The genetic algorithm was adopted to solve the optimization model. Two criteria were proposed to evaluate the optimization results. The first was the correlation coefficient between the OD matrices before and after optimization, and the second was the ratio of the estimated OD flows starting from an entrance to the traffic volumes of the same entrance detected by the inductive loop detectors. The case study using real world data collected in Nanjing, China demonstrated that the proposed method is efficient to derive accurate dynamic OD matrix.