Optimization of transit transfer times: A system-wide approach

This dissertation presents a system-wide approach, based on genetic algorithms, for the optimization of transfer times for an entire bus transit system. Optimization of transfer times in a transit system is a complicated problem because of the large set of binary and discrete values involved. The combinatorial nature of the problem imposes a computational burden and makes it difficult to solve by classical mathematical programming methods. The genetic algorithm proposed in this research attempts to find an optimal solution for the transfer time optimization problem by searching for a combination of adjustments to the timetable for all the routes in the system. It makes use of existing scheduled timetables, ridership demand at all transfer locations, and takes into consideration the randomness of bus arrivals. Data from Broward County Transit are used to compute total transfer times. The proposed genetic algorithm-based approach proves to be capable of producing substantial time savings compared to the existing transfer times in a reasonable amount of time. The dissertation also addresses the issues related to spatial and temporal modeling, variability in bus arrival and departure times, walking time, as well as the integration of scheduling and ridership data