Bus Network Scheduling with Genetic Algorithms and Simulation

Abstract

This thesis investigates the costs associated with a bus scheduling problem in an urban transit network for both deterministic and stochastic arrival processes and proposes computerized models for each. A simple genetic algorithm (SGA) with some problem-specific genetic operators is developed for the deterministic arrival process and a simulation-based genetic algorithm (SBGA) is developed for the stochastic arrival process.

The new models are applied to an artificial bus network to test their efficiency. Several sensitivity analyses and a goodness test are conducted for each arrival process. The results show that the SGA model can find the optimized solution very quickly when it uses problem-specific operators such as the coordinated headway generator, coordinated headway crossover and coordinated headway mutation. They also show that the SBGA model can find a good solution even though it uses general genetic operators.