Development of origin-destination trip matrices using mobile phone call data

Abstract

Reliable Origin-Destination (OD) matrices are important for transportation planning and analyses. However, detailed field data collections by means of extensive questionnaire surveys or license plate matching are very expensive. On the other hand, mobile phone users leave footprints of their approximate locations whenever they make a call or send an SMS. As mobile phones have reached a high penetration rate in both developed and developing countries, the mobile phone call records can be used to extract the OD patterns of travelers. This can be scaled up to determine the actual OD matrix of a transport network. In this thesis, two kinds of data are used for constructing the OD matrices: mobile phone Call Detail Records (CDR) and traffic counts extracted from video recordings. CDR is collected from Grameenphone (GP), the largest mobile operator of Bangladesh. The Dhaka city major road network has been divided into 29 nodes. Each BTS (Base Transceiver Station) is assigned to a specific node. So if someone makes a phone call, his/her location is represented in terms of a definite node. If he moves to another node and makes another call then its position is updated to that node and an entry is recorded in the corresponding OD. The database management software MySQL and MATLAB are used to process data from 1,397,118 users (one day) to determine the OD pattern (termed as seed OD in this thesis). Video data is also collected for three selected days of the same month (eight hour each day) in thirteen important locations of Dhaka city road network. The traffic of Dhaka city is then simulated using a modified version of the microscopic traffic simulator MITSIMLab. The mobile phone CDR data are used as seed OD and an optimization based approach is deployed to determine the scaling factors that minimize the differences between the observed and simulated traffic counts. The main outcome of this research is a methodology to calculate OD matrix for microscopic traffic simulator using mobile phone CDR data and limited traffic video data. This approach is not only economic but also suitable for easy periodic update of the OD matrix and also extendable in the context of dynamic traffic simulation. This could be a useful tool for local transportation agencies for generating their local OD matrices and subsequently using those in the microscopic simulation software for evaluation and impact assessment of alternative transport policies. It may be noted that although this research has taken Dhaka network as a test scenario, this methodology can be used for other city networks also.