Sense and Avoid for Drones: New Algorithm Allows Drones to React in 3.5 Milliseconds, Avoiding Fast Moving Obstacles

Drones still can’t dodge a bullet  – but new research will now help drones avoid relatively fast moving objects.

A team of three researchers at the Department of Informatics, University of Zurich in Switzerland have created a method of sense and avoid for drones using event cameras so that the drones can safely avoid obstacles at very close range range.

Today’s autonomous drones have reaction times in tens of milliseconds.  The new method reduces the overall latency to 3.5 milliseconds,  so that drones can more reliably detect and avoid relatively fast-moving obstacles.   This progress represents a critical step in drone technology: given expectations of an increasingly crowded airspace, drones need to be able to avoid other aircraft in flight.

As described in Science Robotics, the researchers are using a different type of sensor to acheive results:  “To safely avoid fast moving objects, drones need low-latency sensors and algorithms,” says the paper abstract. “We departed from state-of-the-art approaches by using event cameras, which are bioinspired sensors with reaction times of microseconds. Our approach exploits the temporal information contained in the event stream to distinguish between static and dynamic objects and leverages a fast strategy to generate the motor commands necessary to avoid the approaching obstacles. ”

The video below demonstrates and explains the new research.

[embedded content]


CEO,, and CMO of Principle at Spalding Barker Strategies. Has enjoyed working with and around the commercial drone industry for the last 10 years.   Attendance and speaker at Industry Events such as Commercial UAV, InterGeo, Interdrone and others.   Proud father of two. Enjoys karate, Sherlock Holmes, and interesting things.  Subscribe to all things drone at DroneLife here. Email is [email protected]   Make Sure that you WhiteList us in your email to make sure you get our Newsletter.  [email protected]