Lightweight autonomous dragonfly avoids obstacles

December 17, 2013

The DelFly in flight with stereo vision system that allows for autonomous obstacle avoidance (credit: TU Delft)

TU Delft researchers have developed the DelFly Explorer, which they claim is the first Micro Air Vehicle with flapping wings that can avoid obstacles by itself.

Its low weight (20 grams) opens up new possible applications for MAVs, the researchers suggest.

With its two flapping, transparent wings, the DelFly resembles a robotic dragonfly. The micro aircraft has a binocular vision system that weighs four grams and consists of two cameras and a tiny computer.

By combining the images of both cameras, the distances to obstacles are determined in the same way as in humans (depth perception). The tiny computer processes the images in real time so that the DelFly knows precisely where obstacles are located.

If an obstacle is detected, the DelFly will continue to fly a small distance ahead before flying in a circular motion until the obstacle is no longer visible; it then continues its journey in that direction. In this way, the micro aircraft is able to explore unknown spaces independently, without outside assistance.

The DelFly Explorer can perform autonomous takeoff, maintain its height, and avoid obstacles for as long as its battery lasts (~9 minutes). All sensing and processing is performed on board, so no human or offboard computer needs to be in the loop.

There are other MAVs that can also avoid obstacles, such as quadcopters, but these weigh 50 times as much or more, the researchers say. They suggested the DelFly can be used, for example, to detect ripe fruit in greenhouses, to stream live video of parties or concerts from above, or as an autonomous flying fairy in a theme park. There are also obvious military and law- enforcement reconnaissance uses.

The developed stereo vision system will also be valuable for use by larger MAVs: it can contribute to their obstacle avoidance capabilities while minimally affecting their payload capability, the researcher say.