New solution approaches for the traffic counting location problem

Traffic counts are used to monitor traffic flow and measure the vehicle population crossing a station during a certain period of time. This count information then allows us to make origin-destination trip estimation and thus have a better long-term planning in a transportation network. However, this process may demand great deal of time, investment and labor force. Many researchers focus on the optimal usage of available resources (e.g. labor, time, money) for the purpose of the more realistic origin-destination trip estimation. Hence, a manager, or decision maker, selects the location of traffic count stations such that they gather the information how the trips between origin-destination pairs took place. In this regard, the goal of this study is to effectively estimate all origin-destination trips while locating the minimum number of counting stations. This problem is known as traffic counting location problem in the literature, which aims at finding the optimal location of counting stations while minimizing the cost of location effort (i.e., locating the minimum number of counting station). Two main problems arise in traffic counting location problem literature: (1) the existing exact solution methods have poor computational performance and can thus be applied only to only a set of small or middle-sized networks. Moreover, (2) the limited set of networks that can be solved exactly restricts the set of benchmarks that can be used to assess the efficiency of new heuristic methods in this arena. The motivation of this paper is that making improvement to the solution of both problems is an important step to allow these problems to be applied to large, real-life situations addressing such important managerial concerns as managing the resources of a government, or agency, effectively. In this regard, our study presents novel and effective Integer Programming models to solve traffic counting location problem, aiming disconnect all origin-destination trip pairs. We observed that the proposed models in this thesis, tracking origin-destination pairs without needing enumeration of all potential paths, greatly reduced the needed time to solve the problem according to our extensive computational experiments tested on well-known transportation networks. Therefore, due to the tractability as well as returning the optimal solution in a short time, the proposed models in this thesis fills the important gap in the traffic counting location problem literature.