Using Accident Location and interpretable Risk to fine tune Advanced Rider Assistance Systems for Motorcyclists

Motorcycle riding is a popular activity among riders of all ages and the number of motorcyclists is still increasing, despite safety issues being tricky to resolve for this mode of transport. Motorcycle rides constitute a type of vulnerable road user (VRU) since accidents tend to have more severe consequences for them due to the lack of physical protection for riders compared to passengers in passenger cars. Since this is a consequence of the very nature of the vehicle (being less heavy and more dynamic to move) potential safety interventions for
motorcyclists need to be based on predictive indicators for unsafe situations and aim to avoid crashes altogether.
This paper presents the results of ongoing work to improve motorcycle safety by finding causally interpretable risk characteristics based on accident data and motorcycle riding dynamics collected from test rides by individual riders. Dynamics data at known accident spots and representative data for individual rider-typical motions is associated to the type of historical accident in order to produce an estimate not only of risky areas and maneuvers, but also to associate types of riding dynamics that put the driver at risk. The relation to potential causes is essential for the inclusion of the resulting risk warnings in the activation of an Advanced Rider-Assistance System (ARAS), in order to produce a tailor-made response to the individual.