A small drone has taught itself to judge distances using only one eye during trials aboard the International Space Station (ISS), scientists have reported. Although humans can effortlessly estimate distances with a single eye, robots still lack this capability.

“It is a mathematical impossibility to extract distances to objects from one single image if the object has not been encountered before,” says Guido de Croon, assistant professor at the Micro Air Vehicle Laboratory of Delft University of Technology and one of the investigators. “But if we recognize something to be a car, then we know its physical characteristics and we can use that information to estimate its distance from us. A similar logic is what we wanted the drone to learn during our experiment.”

Stereo camera on one of the SPHERES drones. Image credit: NASA. One of the ISS' Synchronized Position Hold Engage and Reorient Experimental Satellites (SPHERES) drones was used for the test. With 12 carbon dioxide gas thrusters enabling rotation and movement in all direction, the bowling ball-sized SPHERES are essentially free-floating mini-spacecraft within the space station and are used for testing a wide variety of technologies.

In the experiment, a drone began navigating inside Japan’s ISS module while recording stereo vision information from its two camera "eyes." It then began to learn about the distances to walls and nearby obstacles so that when its stereo camera was switched off, it could begin autonomous exploration using only a single camera.

Operating in weightlessness, with no favored up or down direction, added to the challenge. However, the experiment demonstrated that machine learning would indeed allow the normally stereo-viewing drone to recover from the loss of one camera.

“It was very exciting to see a drone in space using cutting-edge artificial intelligence methods for the very first time,” says Dario Izzo, who coordinated the research contribution from the European Space Agency's Advanced Concepts Team. “Our approach, based on self-supervised learning, has a high degree of reliability and helps drone autonomy. A similar learning approach was successfully applied to self-driving cars, a task where reliability is also of paramount importance.”

According to Leopold Summerer, head of ESA's Advanced Concepts Team, the experiment represents a further step in the quest for truly autonomous space systems, which are increasingly in demand for complex operations in deep-space exploration.