|Ph.D Student||Haas Inbal|
|Subject||Developing Models for the Optimal Selection of|
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Shlomo Bekhor|
|Dr. Amir Ziv Av|
|Full Thesis text|
Transportation infrastructure projects constitute a key element in the development of transportation networks. The selection and scheduling of the most beneficial set of projects is, however, a very complicated task. This is due to many contributing factors including many, at times conflicting, objectives; interdependencies between the benefits delivered by different projects; and uncertainty with respect to many elements such as the cost estimation and construction duration.
The transportation-related project selection problem is known in the literature as the Network Design Problem (NDP). This problem can wear different forms, according to the preferences set by the decision makers. This problem can focus on a single or multiple objectives and on a single or multiple periods.
One of the major problems frequently associated with the NDP has to do with its solution space. Depending on the examined network and the number of candidate projects, the solution space of this problem can increase very rapidly. This fact poses a challenge with respect to the size of the instances that can be solved, even when meta-heuristic methods are applied.
In this thesis, we formulate 5 different models that address the NDP, representing different objective functions and constraints. The objectives examined in this study are system time, road safety and sustainability. In addition, the models are also different from one another in the treatment of the time perspective. Several models focus on a single period (i.e. selection of projects only), while others on multiple periods (i.e. selection and scheduling of projects).
The developed models aim to fill the gap with respect to several issues concerning the NDP. In particular, the ability to solve large instances of the problem, the evaluation methods of the used objectives, and the treatment of uncertainty properties of the problem. The models were applied on a real-size network, using a large set of candidate projects and their results were analyzed and discussed.
The developed models can serve as valuable tools in the hands of decision makers, aiming at future improvements of existing networks and the optimal scheduling of these improvements.