Multi-agent modeling and analysis of pedestrian and vehicular traffic

Abstract

Evaluating urban infrastructures through computer models is a useful way of gath­ering feedback and improving a proposed design before finally implementing it in concrete. Many current models focus on vehicular traffic, to the detriment of pedes­trians. To address this shortcoming, this thesis contributes a multi-agent model of vehicular and pedestrian traffic, where agents can plan, adapt and execute, produc­ing statistical feedback, and an intuitive visualization. This includes a description of how the environment is represented and how the agents plan and execute. Through the use of a software prototype, analysis of traffic flow is explored in several simple scenarios, highlighting the effect of agent variations. Through a model of the Uni­versity campus, a more realistic and holistic scenario is analysed, demonstrating the applicability and limitations of the model.