Conventionally, selection of traffic counting/survey stations for volume studies is purely subjective, relying much on the political jurisdictions, natural boundaries or man-made barriers. Two conflict objectives are generally kept in mind when allocating counting stations: obtaining as much representative traffic information as possible and saving subsequent manpower requirement in data collection. Therefore, it is crucial to determine the optimal traffic counting locations. However, very limited attention has been devoted to developing efficient traffic counting location strategies. The thesis aims to fill this gap and create a computer- based mathematical tool for systematically selecting traffic counting locations for Origin-Destination (O-D) matrix estimation....[ Read more ]

Conventionally, selection of traffic counting/survey stations for volume studies is purely subjective, relying much on the political jurisdictions, natural boundaries or man-made barriers. Two conflict objectives are generally kept in mind when allocating counting stations: obtaining as much representative traffic information as possible and saving subsequent manpower requirement in data collection. Therefore, it is crucial to determine the optimal traffic counting locations. However, very limited attention has been devoted to developing efficient traffic counting location strategies. The thesis aims to fill this gap and create a computer- based mathematical tool for systematically selecting traffic counting locations for Origin-Destination (O-D) matrix estimation.

In this thesis, we have developed network traffic counting strategies for O-D matrix estimation according to the existing location rules. We consider two aspects of the counting location problem: how to choose the optimal locations for a given number of counting stations to intercept as many O-D pairs as possible; how to determine the minimum number and locations of counting stations required for interception of all the O-D pairs. The problems of interest are formulated as two different models. Firstly, we formulate them as binary integer programming models and the Genetic Algorithms is employed to solve them. Furthermore, the traffic counting location problem is reformulated as an integer linear programming model and an efficient algorithm based on column generation scheme is developed. Finally, we use the proposed models and algorithms to evaluate the impacts of the resultant counting location patterns on the O-D matrix estimation quality in terms of several measures. All the analyses and developments will be illustrated by numerical examples.