On his bike like no other, Holger Caesar sets off into the middle of traffic: equipped with laser sensors and scanners, the Dutch researcher is in search of data which, according to him, could one day save lives. Its blue electric mount darts into the chaos of the afternoon, while thousands of students cycle through the campus of the University of Delft (South Holland). A maze of cycle paths like everyday life in a country where the number of bicycles exceeds that of the population. “We record the behavior of cyclists and hope that this data will have many applications in the future,” Mr. Caesar explains to theAFP.
The ultimate goal: to help cyclists avoid obstacles, facilitate the construction of self-stabilizing bicycles or even teach autonomous vehicles to avoid hitting cyclists. “Concerning cars, it’s relatively simple (…): they go left, they go right, they go straight, but for cyclists, it is very difficult to predict how they will behave,” continues Mr. Caesar. “We could, for example, use the data to develop an application that alerts motorists when a cyclist makes an unexpected movement,” he adds.
Equipped at the front and rear with three LiDAR sensors, “Light Detection And Ranging” or light detection and ranging, the bike called “Delft SenseBike” could just as easily appear in a science fiction film. This technology is often used for autonomous vehicles, where laser sensing helps generate a three-dimensional image of their surroundings. The sensors emit a large amount of infrared rays which then reflect off surfaces and are sent back to the LiDAR, allowing it to “map” the area the SenseBike passes through, including moving objects and people such as cyclists.
The collected data is then processed using a technique called “labeling”, which involves associating everything visible in the images with a description such as “tree” or “cyclist”. This technique should allow a car to recognize a cyclist and avoid a collision. “The first step will be to make this data publicly available, so that academics and entrepreneurs can benefit from it,” says Mr. Caesar. From this data, artificial intelligence algorithms can be developed to detect, track and predict the behavior of cyclists and “plan our own route to get around them,” adds the researcher.
The Netherlands has no less than 37,000 kilometers of cycle paths and 22 million bicycles. But data is lacking in this area. It is “difficult” to quantify the number of accidents because “not all are recorded”, recognizes the Dutch Cyclists’ Federation on its site. “About 270 people died in bicycle accidents in 2023,” according to the country’s Central Bureau of Statistics. Nearly half of deaths are caused by collisions between cyclists and cars, trucks or buses.
-“Cars are becoming safer for passengers, but not for other road users,” notes Esther van Garderen, director of the cycling federation. After self-driving cars, could data science enable bicycles to drive themselves? “I think it would take the fun out of cycling,” Mr. Caesar laughs. “But we believe it is still possible to make cycling safer,” he concludes.
Are you already following us on WhatsApp?
Subscribe to our channel, activate the little ???? and you will receive a news recap every day in early evening.
