Abstract

The objective of any transit system is to provide better level of service to its passengers. Better level of service can be provided by proper scheduling. The scheduling problem was formulated as an optimization problem to minimize the overall transfer time (TT) of transferring passengers and initial waiting time (IWT) of the passengers waiting to board a bus/train at their point of origin. In this thesis, a mathematical programming (MP) formulation of the scheduling problem at one transfer station is presented. Large number of decision variables and the non-linear nature of some constraints and the objective function in the traditional MP formulation resulted in high computational complexity Genetic algorithms (GAs) - search and optimization methods based on natural genetics and selection - were used to solve the scheduling problem. Function based declaration and coding of variables are the main advantages of using GAs for the scheduling problem. These advantages allowed an efficient reformulation of the problem. The computational complexity of the reformulated problem was considerably less than the MP formulation of the problem. Results obtained from a number of test problems indicate that GAs are able to find optimal schedules with minimal computational requirements, very unlike the traditional methods. A number of extensions to the present study are presented. The success of GAs in finding a solution to this problem suggests that GAs are efficient as a solution tool for similar optimization problems arising in transportation systems.