e-SAFER - Computational models for a safe interaction between (automated) vehicles and e-scooters

Important results from the project

The goal of e-SAFER was to make interaction between e-scooterists and drivers safer. Within the project we created computational models that can assist drivers as they approach an intersection. The models were demonstrated on test-track (TRL4-5) and use kinematics glance behaviour to infer whether a vehicle is approaching an intersection safely when an e-scooterist may cross. The project also contributed several scientific publications detailing the models and the interaction scenarios.

Expected long term effects

The computational models developed within e-SAFER can be integrated into driver monitoring systems and used for context-dependent threat assessment in advanced driver assistance systems (ADAS) to improve the traffic safety of these systems in the future. The project also collected naturalistic data, which proved to be very useful for understanding unsafe behavior among e-scooter riders. However, we were not able to use these data for model validation.

Approach and implementation

E-SAFER investigated the interaction between (automated) vehicles and e-scoters. We leveraged crash databases and insurance claims to devise a test-track experiment where a human driver interacted with a robot e-scooter. The data from this experiment were used to create computational models of driver behaviour including glances. The models were demonstrated on the airfield of Vårgårda in September 2025.