Survey of Machine Learning Methods Applied to Urban Mobility

Abstract

To increase the sustainability in urban mobility, it is necessary to optimally combine public and shared vehicles throughout a passenger's trip. In this work, we present a survey on urban mobility based on passengers' data and machine learning methods. We focus on four applications for urban mobility: public datasets, passenger localization, detection of the transport mode and pattern recognition and generation of mobility models. Public datasets lack data of multimodal trips and are in need of guidelines to facilitate the data collection and documentation processes. Passenger localization is predominantly done through fingerprinting in indoor environments; and fingerprinting relies on unsupervised learning to survey access points. The most common mean of transport detected is the bus, followed by walking and biking, while e-scooters are not included within the detected transport modes. The existing works focus on predicting the travel time of the passenger's trajectory and no machine learning method stands out to estimate the travel time. There is still a need for works that analyze how passengers make use of the urban infrastructure, which will support municipalities and transport mode operators in resource planning and service design.