M.Sc Thesis


M.Sc StudentLev-Ran Eldar
SubjectSemi-Supervised Travel Mode Detection of
Micro-Mobility Electric Vehicles
DepartmentDepartment of Civil and Environmental Engineering
Supervisor PROF. Sagi Dalyot


Abstract

Transportation systems allow the distribution of goods and services, providing people with the means to reach their destinations efficiently and perform their activities. Investigating the mobility patterns associated with these systems often requires a rich mobility dataset that best represents the typical traffic of the research area - not just in a spatial and temporal sense, but also regarding the context relevant to the analysis. With today’s unprecedented volume of crowdsourced data, many large scale and public travel datasets exist, but they lack the semantics and context of a user’s travel, which is mostly manually documented in prompted recall surveys and may suffer from low response rate and ambiguities. To utilize these datasets, qualitative and reliable travel mode detection algorithms are required. Detection algorithms do exist, but they are not compatible with newly collected datasets in urban areas that contain large volumes of emerging micro-mobility vehicles. In this research, the investigation of the unique mobility patterns of e-bicycles and e-scooters is made, focusing on the development of machine learning algorithms that use travel properties to correctly distinguish these emerging travel modes from other common urban travel methods. Contributions include a state-of-the-art semi-supervised neural network model that can preserve its performance between spatial domains with small amounts of data volumes. These can be used by policy makers and researchers to produce a comprehensive understanding of existing urban travel behavior, contributing to better urban transportation systems planning and design.