Analysis of urban air mobility’s transport performance in São Paulo Metropolitan Region using MATSim Simulation metamodel based on active learning to predict the simulations outputs

Sérgio Fonseca dos Santos

Key information

Authors:

Sérgio Fonseca dos Santos (Sérgio Fonseca dos Santos)

Supervisors:

João Pedro Faria Mendonça Barreto (João Pedro Faria Mendonça Barreto); Zafeiris Kokkinogenis

Published in

12/02/2021

Abstract

Due to the rapid population growth, it is estimated that the level of congestion increases in cities, making it necessary to integrate new transport modes, in order to reduce travel times. Recent advances in aeronautical technology have allowed the development of a new transport mode, Urban Air Mobility (UAM). When implemented, it will make possible to move people from one point to another using vehicles with vertical take off and landing capabilities. While these vehicles have been extensively studied, there is little research regarding the impact the introduction of these vehicles will have on cities. It is, therefore, essential to predict the behavior of these when introducing this new transport mode, in order to adopt the best measures. That said, this work has two objectives: to predict the impact that the introduction of this new transport mode will have in the city of São Paulo, Brazil, through an agent-based Mobility Simulator (MATSim) and to create a simulation metamodel based on Gaussian Processes (GP) and active learning, which allows to predict the result of simulations, instead of them being carried out. Regarding the simulation results obtained, it is verified that in the simulated base scenario there is great air congestion. However, the introduction of this new transport mode can serve to alleviate congestion on the roads. Regarding the metamodel, it can predict, with some accuracy, the results of simulations carried out. However, it is necessary to improve the strategy in order to bring the metamodel closer to the simulator.