Researchers at New York’s Cornell University have developed software that will help drones dodge obstacles without the assistance of humans. Up until now, the flying robots have always relied on humans at the control to operate them by remote and prevent crashing.
In experiments conducted, a miniature helicopter was equipped with a camera and captured images as it flew by. The software developed by assistant professor of computer science Ashutosh Saxena and his team, turned the image in the drone’s camera into a 3D model of its own environment – the robotic brain uses an algorithm to determine which objects are obstacles.
Out of 53 flights, which included many obstacles, the robot managed to find a path without crashing. In the final two flights however, the flights failed due to wind gusts. Eventually, the team hopes that the drone will be able to calculate wind patterns and avoid moving objects like birds.
The research is being funded by US Defense agency DARPA, along with drone-producer Lockheed Martin. In a paper entitled Low-Power Parallel Algorithms for Single Image based Obstacle Avoidance in Aerial Robots, it outlines the success of the experiments – saying:
“In outdoor robotic experiments, our algorithm was able to consistently produce clean, accurate obstacle maps which allowed our robot to avoid a wide variety of obstacles, including trees, poles and fences.”