Cyclists are frequently confronted with the dangers posed by turning trucks, overtaking buses or cycle paths blocked by parked vehicles. However, near misses are not recorded in accident statistics and bicycles are not equipped with sensors capable of providing relevant data. The first findings on risks faced by cyclists are however now available as part of the “SimRa-Sicherheit im Radverkehr” project led by Professor Dr. David Bermbachof the Mobile Cloud Computing research group at TU Berlin and the Einstein Center Digital Future (ECDF).
David Bermbach and his colleagues have developed an app to collect data on “near- accidents” involving cyclists and other road users. Members of the public can download the SimRa app for free from the app store on their smartphone. The app uses GPS data to display routes and acceleration sensors to detect risk situations. To date, some 20,000 routes have been compiled and analyzed for Berlin. Data make it clear that streets used by cyclists with cars traveling in both directions and with parking spaces but without cycle paths are particularly dangerous. In such streets, cyclists are often overtaken with close passes by motorists traveling at high speed or else have to contend with tailgating.
When analyzing their data and comparing it with bicycle accident statistics for 2019, it came as little surprise to the researchers at TU Berlin and the ECDF to learn that major shopping streets, such as Kurfürstendamm and Friedrichstrasse, are unsuitable locations for leisurely bicycle rides.
More surprising perhaps was the fact that streets such as Weichselstraße in the borough of Friedrichshain-Kreuzberg or Elsenstraße in the borough of Treptow-Köpenick were also classified as medium-high and high risk. The section of Kronprinzessinnenweg between Havelchausee and Königstraße in the borough of Steglitz-Zehlendorf was similarly classified as medium-high risk. A frequent danger on this stretch documented by cyclists using the app was represented by vehicles overtaking too close to the cyclist. The stretch between Hüttenweg and Havelchaussee is currently blocked for motorized vehicles. The research team recommend that this measure be extended to the entire Kronprinzessinenweg as it is a very popular route for people making an outing by bike.
“Danger is widespread in Berlin and not just restricted to three or four corners of the city. But similar solutions can be used for similar problems,” explains Bermbach, who, together with his team, has developed potential solutions for particular road sections and intersections. Until traffic planning experts conduct the corresponding on-site analyses, however, these solutions are intended to serve as a basis for discussion rather than direct recommendations for implementation.
Research findings are visualized using an interactive street map showing all road sections and intersections with either a) at least 50 journeys or b) at least 10 journeys and a hazard score of 0.25 or more. “Users can click on street sections and intersections to view recorded incidents,” explains Bermbach. The incidents analyzed include overtaking with close passes or tailgating, vehicles entering or leaving a parking space, near-accidents occurring as a result of turning, oncoming road traffic, near-incidents of dooring, and obstructions or obstacles.
A further result map shows all road sections and intersections in Berlin with at least 10 journeys and above. Users can also click on road sections and intersections here to view recorded incidents. However, there are as yet insufficient data for many sections of road.
“We have conducted analyses using both result maps to create danger zone reports. We are publishing our research findings on the GitHub platform to make our data freely accessible and to enable us to regularly update the datasets,” says David Bermbach. “We are currently working on migrating data in half-yearly packets to an open data repository.” Open data repositories are large, freely accessible databases for long-term archiving of research data and are generally operated by libraries.
In the coming months, the researchers will be releasing the new machine learning-based incident detector. This will use AI processes to detect recurring patterns, such as incidents of full braking, recorded in the acceleration data in the app. The first results for other regions will also be compiled and presented soon. The project currently has 12 partner regions, including Munich, the Ruhr, and Bern in Switzerland.