Is the risk worth the ride? Crash causation analyses of naturalistic e-scooter data

Licentiate thesis, 2025

The rapid increase in e-scooter popularity has brought an increase in crashes resulting in injuries and fatalities, creating growing concern about e-scooter safety. The aim of this thesis is to investigate which limitations in human behaviour, vehicle design, riding environment, and infrastructure contribute to e-scooter crashes. The two included studies address the limitations of conventional crash databases, using naturalistic riding data from instrumented rental e-scooters in an urban environment. The high-frequency kinematic and video data elucidated behaviours and factors contributing to safety-critical events (SCEs: crashes and near-crashes). The studies focused on two topics: identifying the key risk factors for e-scooters and evaluating the impact of methodological choices on the risk assessment. This unprecedented research used kinematic triggers to identify trips with at least one SCE, which were then verified through video footage. The identified events were labelled and relevant variables related to the rider, infrastructure, environment, and trip characteristics were extracted.

The results highlight the need to adapt definitions of crashes and near-crashes to reflect the unique characteristics of e-scooters (and perhaps other forms of micromobility). The results also show the need to prioritise safety interventions based on both crash risk and crash prevalence to optimise their impact on safety. In fact, the factors such as riders using the e-scooter for leisure trips, the presence of intersections, trips taken on Fridays and Saturdays, pack riding, and inexperienced riding—listed in decreasing order of prevalence—were nonetheless all significant contributors to risk. The results challenge assumption derived from conventional crash databases; for example, if nighttime riding is not as risky as previously believed, nighttime bans might not be necessary. Identifying risk factors from SCEs requires a baseline for comparison, which captures typical riding scenarios with no SCEs. In this thesis, two different approaches to baseline selection (random and matched) were compared. The results indicate that both random and matched baselines are necessary to get the full picture of crash causation.

In conclusion, this thesis contributes to the field of micromobility safety by identifying several factors influencing e-scooter crashes and evaluating the impact of baseline selection. Additionally, the need for tailored definitions of e-scooter SCEs was identified. These insights can guide the development of suitable interventions, such as rider training programs, targeted campaigns, risky-riding detection systems, and intelligent communication systems, to enhance e-scooter safety.