University of Zurich researchers have developed technology that will enable drones to detect and avoid other fast-moving objects in as little as 3.5 milliseconds using special event cameras and newly developed algorithms. The reaction time makes the aircraft “nearly” as reliable as human pilots and is directed at beyond visual line of sight operations, according to researcher Davide Falanga.
In demonstrations, the system installed onboard a test aircraft was shown to successfully detect incoming objects “more than 90 percent of the time” in both indoor and outdoor tests. Traditional camera-based detect-and-avoid technology relies on computer analysis of all the pixels in a digital image, while event cameras use smart pixels that work independently of each other. This allows the onboard computer to dramatically speed computations by detecting changes in individual pixels such as light intensity, as opposed to having to wait for analysis of the entire image, thereby slashing vehicle reaction times by 30 to 40 milliseconds.
Researchers developed new algorithms to work with the event cameras that subtract drone movement from recorded camera events over a short burst of time. Related research was funded by the Swiss National Science Foundation through the National Center of Competence in Research (NCCR) Robotics and published in the Science Robotics journal.