Projects

This study examines the factors influencing cycling accident severity and how they relate to the built environment. Using accident data from Berlin combined with volunteered geographic information, we develop a modeling framework that integrates machine learning and econometric methods through a latent class discrete outcome model. The approach identifies different built environment contexts in which accidents occur and reveals how risk factors vary across these contexts. By capturing this heterogeneity, the model helps uncover complex relationships between urban environments and accident outcomes, providing insights that can support targeted safety interventions and more effective cycling policies.

This study develops a new approach to measure how people perceive cycling safety in urban environments. Instead of relying on traditional surveys, participants were shown pairs of real-world street images and asked to select which environment appeared safer for cycling. The responses were used to train a siamese convolutional neural network capable of learning safety preferences directly from images, including situations where participants perceived both environments as equally safe. The resulting model can predict human perceptions of cycling safety and enables scalable, continuous assessments of street environments, supporting more effective interventions to improve cycling conditions in cities.

CYCLANDS is a curated collection of 30 datasets on cycling crashes to lower the barrier in objective cycling research comprising nearly 1.6M cycling accidents. This data is vital for road safety analysis, enabling researchers to develop models to understand how different factors impact the frequency and severity of accidents. Observations include the severity and location of the accident, aiming to foster the worldwide study of cycling safety by providing a testbed for researchers.

• Website
• Article
• Data
• Code

This research explores how street network efficiency relates to mobility patterns across 41 European cities by comparing circuity levels for walking, cycling, and driving networks. Using data-driven clustering, cities were grouped into six distinct mobility cultures, such as Cycling Champion Cities, Sustainable Transportation Advocates, and Car Dependent Cities. The analysis reveals significant differences in network efficiency across these groups, with car-dependent cities showing the highest circuity and sustainable mobility cities the lowest. By linking urban network structure with mobility culture, the study provides a benchmark for cities seeking to improve transportation efficiency and transition toward more sustainable mobility systems.

This study examines how the structure of urban street networks influences mobility by analyzing circuity for Lisbon between 2013 and 2020. We simulated 4.8 million routes across driving, cycling, and walking networks to understand how circuity evolves over time and differs across transport modes. The results show that each mobility network has distinct structural characteristics and should therefore be analyzed separately in urban planning and modeling. In particular, Lisbon's recent expansion of cycling infrastructure significantly improved the efficiency of cycling routes. The research demonstrates that circuity is a valuable indicator for identifying where urban interventions can make active travel modes more attractive and support more sustainable mobility systems.