The University of Zurich and ETH Zurich have managed to take obstacle avoidance to a new level, allowing drones to dynamically avoid static and non-static objects with a latency of only 3.5 milliseconds. The drone is equipped with event cameras that detect light changes in the individual pixels instead of the whole frame.
Dynamic obstacle avoidance allows drones to dodge moving objects
The University of Zurich and ETH Zurich engineers have been able to develop a new obstacle avoidance system. The system allows drones to dynamically track and avoid static and non-static objects they encounter.
The new system uses event cameras that pick up changes in light on a per-pixel basis, allowing unchanged pixels to not produce any data, lowering the amount of data being analyzed and therefore improving reaction times. The team has also developed a new algorithm to support the event cameras with an overall latency of only 3.5 milliseconds. Which is “sufficient for reliable detection and avoidance of fast-moving obstacles.”
The engineers threw various types of balls at the drone both indoors and out with the drone managing to dodge the ball every time without any issues. This is because the algorithm monitors the last 10 milliseconds of data from the event cameras to detect if the object is static or non-static and then move the drone out of its path accordingly. The autonomous drone was able to avoid obstacles at speeds of around 10 meters per second, only using on-board sensing and computation. Be sure to take a look at the video below.
Would you want this dynamic obstacle avoidance implemented to consumer drones? Let us know in the comments below.
Stay in touch!
If you’d like to stay up to date with all the latest drone news, scoops, rumors, and reviews, then follow us on Twitter, Facebook, YouTube, and Instagram, or sign up for our daily email newsletter that goes out every weekday at 6 p.m. ET.
Buy your next drone directly from manufacturers such as DJI, Parrot, or Yuneec, or retailers like Adorama, Amazon, B&H, BestBuy, DroneNerds, or eBay. By using our links, we will make a small commission at no additional cost to you. Thank you for helping DroneDJ grow!
Photo: Science Robotics