Potential of Autonomous Mobility on Demand operations across Switzerland

Abstract

Autonomous vehicles (AVs) are said to have an important role in transport in the near future. The design and control of a fleet composed of AVs that satisfies a specific demand have occupied many researchers in the topic. That is why simulations with different parameters and assumptions have been carried out to study the viability of the service. This thesis focuses on finding an alternative way of defining an optimal AVs fleet size for a certain area, without the need to carry mobility simulations on it. At the same time, it gives an image of large scale scenarios for an Autonomous Mobility on Demand(AMoD) service. For this purpose, some areas of different sizes and differently populated have been defined across Switzerland. First, metrics quantifying different aspects of their road network and demand distribution have been computed. In parallel, Multi-AgentTransport Simulations (MATSim) have been carried out based on a static demand in each area for different users’ shares. In the last step, two regression methods have been applied to estimate an optimal fleet size for each study area based on the computed attributes that show a higher impact on the fleet size. Results show a range of predictions that rely on certain highly correlated metrics, obtaining in the best case an average prediction error of 34% compared to the results of the simulations.