Individual motorcycling safety: creating a safety profile from riding data

Abstract

Motorcycle riders are vulnerable road users, who suffer fatal accident outcomes at significantly higher rates than car drivers. Lifesaving assistance systems are considerably harder to integrate in the operation of a motorcycle, compared to the operation of a more predictably moving car. Here we present a methodology of estimating an individual rider's classifier of “risky” dynamics from their riding behaviour on several popular motorcycling routes in an experimental set up. Using clustering of common motions and data obtained at known accident sites, as well as an updating regime for the model fit to the individual rider, we are able to identify potential riding risks in an online methodology and determine the driving factors (i.e., the most relevant dynamics for a motion to be classified as risky) in the risk estimate, to potentially base interventions on.