Reducing accident risks for cyclists with AI: Recording movement data more precisely

The movement data of cyclists can be recorded with an accuracy of around one meter using commercially available smartphones. This is shown by a current research project by the Department of Communication Technology headed by Prof. Dr.-Ing. Klaus David at the University of Kassel. The department is working on the development of protection systems that warn drivers and unprotected road users of potential collisions and thus help to prevent accidents.

Lennart Busch presenting the results at PerCom in Pisa.

The researchers have developed a method with which the movement data of cyclists can be recorded with a location accuracy of around one meter or 0.64 meters. This is implemented using commercially available smartphones. The technical approach is based on a so-called "digital twin", which enables localization with the inclusion of environmental information, in particular by using the position, size and height of traffic signs. This is a new methodological approach to positioning that differs from classic GNSS/GPS-based methods, whose accuracy is typically in the range of three to ten meters.

Compared to existing approaches, the method enables significantly higher accuracy in the range of less than one meter for the first time, while at the same time using hardware that is suitable for the mass market. Alternative solutions, on the other hand, are significantly more cost-intensive and are less suitable for everyday use.

The research was published as part of the paper "SignAlign: A Smartphone-Based Visual Positioning System for Enhanced Bicycle Safety" and presented at the international conference IEEE PerCom in Pisa. Lennart Busch presented the results there on March 19.