Date of Completion

12-17-2015

Embargo Period

12-16-2015

Keywords

Transit Networks Design Problem; Genetic Algorithm; Equity; Network Optimization

Major Advisor

Dr. Nicholas E. Lownes

Associate Advisor

Dr. Karthik Konduri

Associate Advisor

Dr. Amy Burnicki

Field of Study

Civil Engineering

Degree

Doctor of Philosophy

Open Access

Open Access

Abstract

The equitable provision of public transportation services is a major concern for transit planners and service providers around the world. However, very few tools are available to planners seeking to incorporate equity concerns into their transit network designs. This research proposes innovative models for directly incorporating equity into the stop sequencing and stop grouping components of the Transit Network Design Problem (TNDP) and develops a solution method, a genetic algorithm, for solving these models. First, a single route model is developed to test the effects of designing routes according to different definitions of equity. This research explores nine possible inequity minimizing objective function formulations, drawing from horizontal, vertical, and intermodal equity perspectives. The single route model is largely based on the classic traveling salesman problem (TSP); every stop must be visited once and only once on a single circulating route. The primary difference between the two is the single route model replaces the TSP’s cost minimizing objective function with an inequity minimizing function. The Sioux Falls and Willimantic networks were used to test the single route model and to develop a genetic algorithm capable of solving this complex problem. These experiments narrowed the list of possible equity objective functions from nine to six. Extensive testing was conducted on the genetic algorithm, both on the algorithmic structure and on the input parameters, to validate its quality and efficiency. After testing the single route model and developing a solution method, the model was expanded into a multiple route model which addresses both the stop sequencing and stop grouping components of the TNDP. This model also considers route transfers, walking paths to and from transit stops, stops serving the same demand zones, multiple paths between demand zones, and idle time. This model was applied to a subset of the University of Connecticut’s, USA, shuttle bus system and solved using an expanded and updated version of the genetic algorithm applied to the single route model.

COinS