Institute of Transportation, MOTC

Chih-Chao Fan, 2006.07

Feng Chia University - Graduate Institute of Traffic and Transportation Engineering and Management

 　　The origin-destination (O-D) matrix is important information for proposing traffic management policy and planning transportation facilities. However, for a freeway system without ramp-closed tolling or vehicle tracking system, a large-scale survey is the only way to obtain such information exactly. Thus, numerous models have been proposed to estimate the O-D matrix from link traffic counts. However, the accuracies of these estimation models depend upon the traffic patterns and geometric layout of network. Their performance can not be guaranteed if no further information can be acquired except for limited information of link traffic. License plate recognition system (LRS) can automatically recognize and compare the license plate number of passing vehicle, and then determine the track of the vehicle. Therefore, with the extra information provided by LRS, the accuracy of estimated O-D matrix can be remarkably enhanced. Of course, if LRS are installed at all ramps of freeway, the O-D matrix can be exactly collected. Nevertheless, wide installation of LRS may invade the privacy of road users and consume a great deal of government budget. Therefore, it is an essential topic to determine the optimal number of LRS and their locations for the purpose of enhancing the estimation of O-D matrix. Based on this, the present paper proposes a mathematical programming model to determine the optimal locations of LRS. Three objectives are considered: maximization of the accuracy of estimated O-D matrix, minimization of LRS installation cost and minimization of the privacy invasion. The O-D matrix estimation method proposed by Robillard (1975) is adopted, because the method is easy to incorporate the extra traffic equations provided by LRS and simple to manipulate. These extra traffic equations are derived from pairwise comparison of the recognized plate numbers between two arbitrary LRSs installed in different locations. Due to the combinatorial characteristics of problem, genetic algorithms (GAs) are employed to solve the optimal locations of LRS.
　　For investigating the validity of proposed model, The exemplified examples is a freeway with 8 interchanges and 42 interchanges randomly generated O-D matrix. The result shows, the error rate estimated of OD matrix decrease by the numbers of LRS increase. The decrease range about 5%-10%. The location of LRS discovered that we can set one or two LRS first at the middle and both sides of high flow section, then we increase the LRS in the both sides of high flow section. The best number of LRS is six to nine. For verifying the applicability of model, this research takes the example application of No.1 national freeway of Taiwan of OD matrix. The result shows, best number of LRS is seven, and we need to set them at Shijr Interchange、Yuanshan Interchange、Linkou Interchange、Hukou Interchange、Taichung Interchange、Shinying Interchange、and Kaohsiung Interchange. It can make error rate drop by 19.66%, this result shows the validity and applicability of proposed model.