Publication

Cycling safety is a worldwide safety concern that jeopardizes cycling sustainability. In 2011, cyclist fatalities totaled 658 in Japan, 677 in US, and 2092 in EU. In Europe the proportion of bicyclists’ death is also increasing as a consequence of drivers’ deaths decreasing over time. The tremendous increase in driver safety in the last few decades has been driven by research, technology, and legislation that have no counterpart for bicyclists. The most credited method for further improving driver safety is analysis of field data such as naturalistic driving data. These field data have already provided insight into driver behavior and accident causation that were not measurable with previous data. Naturalistic data is currently used to drive the development and evaluation of intelligent safety systems, such as driving assistance systems and active safety systems in general. In contrast, naturalistic data from bicycles is extremely limited and thus not widely used to tackle cycling safety issues. In this paper naturalistic cycling data was used to explore the relation between ranges of bicycle kinematics and cycling safety. The results presented here show that such a relation exists and that longitudinal speed and vertical acceleration are the best predictors for safetycritical events. Furthermore, while vertical acceleration is the best predictor for crashes, rapid changes in longitudinal speed relate to the cyclist’s perceived safety. Future research should determine the extent to which cyclists’ safety margins can be assessed in real time and whether this assessment can occur early enough for effective intervention.