Drones with cameras are already able to avoid obstacles today but are not particularly good at it. They react slowly, particularly when fast-moving objects approach them. After their cameras recognize a bird, for example, it takes 20 to 30 milliseconds for them to react. For many areas of use this is not fast enough, as the University of Zurich (UZH) explains in a press release.
Researchers from the UZH have now fitted a drone with so-called event cameras, which are cameras specializing in faster recognition of moving objects. Unlike normal cameras, they do not process all of the pixels, only those that detect a change in light intensity. As such, the focus is only on moving objects and this speeds up processing.
Event cameras are a new invention, so the researchers had to develop their own algorithms for their use in drones. The researchers tested the system by throwing various objects at the drone, which was able to evade in over 90 percent of cases, even if a ball approached from a distance of only 3 meters at a speed of 10 meters per second. Moreover, the reaction took just 3.5 milliseconds, which is about ten times faster than previously possible.
Professor Davide Scaramuzza from the UZH states: “One day drones will be used for a large variety of applications, such as delivery of goods, transportation of people, aerial filmography and, of course, search and rescue.” In these situations, they would need to reliably recognize approaching obstacles. However, this ability also plays an important role in other areas, such as the automotive industry, mining, and remote inspection.
The work was funded by the Swiss National Science Foundation through the National Center of Competence in Research (NCCR) Robotics.