Zurich Researchers Teach Drone to Dodge with Dynamic Obstacle Avoidance

One of the many things preventing urban drone deliveries reaching Amazone Prime levels of mass adoption is the threat of interference, whether that’s from pets, kids, thieves, gun-toting neighbors or rogue footballs.

Researchers at the University of Zurich, Switzerland, are working on sense and avoid technology that could provide drones with a more dynamic form of self-defense.

The next level of obstacle avoidance?

A team from the University of Zurich’s Robotics and Perception Group have developed a drone that uses a camera and an onboard Visual-Inertial Odometry system to see an incoming ball and dodge out of the way.

The system takes a different approach to the kind of sense and avoid technologies we have seen from DJI and Skydio – which are focused on avoiding collisions with static obstacles.

Read more: New Developer Platform Could Make Skydio R1 The Go-To Commercial Drone

In part that’s because the University of Zurich researchers’ drone used a sensor called an event camera. Rather than recording frames each second and passing them on for software to analyse, event cameras work by only sending data when the pixels shift or spike in intensity.

Read more: 3 Technologies That Could Transform Future Drone Operations

This reduces the processing bottlenecks that restrict a conventional sense and avoid system, cuts down the latency and results in much-improved response times.

These kinds of sensor are much more expensive than their standard camera counterparts, so it could be a long time before they are used to develop sophisitcated obstacle avoidance systems in consumer and professional drones.

The work is outlined in a research paper, entitled ‘How Fast is Too Fast? The Role of Perception Latency in High-Speed Sense and Avoid’. It was written by researchers Davide Falanga, Suseong Kim, and Davide Scaramuzza and can be read in full here.

Read more: Exclusive: Parrot Explain ANAFI’s Lack of Obstacle Avoidance