Racing Drones Help Space Agency Develop Next-Gen Navigation Systems

The European Space Agency (ESA) is turning to the high-speed world of drone racing to test AI-powered control systems, aiming to prepare future spacecraft for navigating space.

Unlike on Earth, space is unpredictable, with gravitational fluctuations and atmospheric turbulence likely to force spacecraft to deviate off course.

The researchers are attempting to develop a low-cost solution that would enable a spacecraft to maintain its course without the need for brute-force corrections. 

Their approach, called Guidance & Control Networks (G&C Nets), allows spacecraft to continuously replan optimal trajectories rather than sticking to a set course.

Dario Izzo, the scientific coordinator for the ESA’s advanced concepts team, explained that drone racing provides a real-world testing environment for its neural networks.

“We’ve been looking into the use of trainable neural networks for the autonomous oversight of all kinds of demanding spacecraft maneuvers, such as interplanetary transfers, surface landings and dockings,” Izzo said. “Such a neural network approach could enable optimal onboard operations, boosting mission autonomy and robustness.”

The tests involved drones equipped with the ESA’s neural-network-based AI control systems, challenging them to race against time in Delft University of Technology’s research and test laboratory, known as the “Cyber Zoo.”

Related:NASA Charts AI, Robotics, 3D Printing as Path for Mars Sustainability

In a 33-foot by 33-foot test area, the drones navigated a course, beating set times to ensure fast operations while adapting to changes in their environment. 

The drones flew separately but were powered by the same neural network-powered control system. They were able to continuously replan optimal trajectories, rather than sticking to a set course to handle environmental changes.

Human-steered Micro Air Vehicle quadcopters were alternated with the autonomous drones to compare their performance.

“The main challenge that we tackled for bringing G&CNets to drones is the reality gap between the actuators in simulation and in reality,” said Christophe De Wagter, TU Delft’s principal investigator. “We deal with this by identifying the reality gap while flying and teaching the neural network to deal with it. For example, if the propellers give less thrust than expected, the drone can notice this via its accelerometers. The neural network will then regenerate the commands to follow the new optimal path.”

The ESA’s upcoming Hera mission will incorporate some of the learnings from the drone tests conducted in the Netherlands. 

Related:NASA Appoints Agency’s First AI Officer

The probe will attempt to assess the impact of NASA’s DART mission. 

In September 2022, NASA fired a projectile at an asteroid in the Didymos binary asteroid system to evaluate potential planetary defenses against objects approaching the planet.

Hera will attempt to evaluate the DART mission’s impact site while navigating nearby asteroids as it gathers the data.

The ESA wants future probes and spacecraft to be able to maneuver around objects like asteroids autonomously and effectively.

“The traditional way that spacecraft maneuvers work is that they are planned in detail on the ground [and] then uploaded to the spacecraft to be carried out,” said Sebastien Origer, an ESA graduate trainee. “Essentially, when it comes to mission Guidance and Control, the Guidance part occurs on the ground, while the Control part is undertaken by the spacecraft.”

"Our alternative end-to-end Guidance & Control Networks approach involves all the work taking place on the spacecraft. Instead of sticking to a single set course, the spacecraft continuously replans its optimal trajectory, starting from the current position it finds itself at, which proves to be much more efficient."

The graduate said the drone-based efforts “represent a way to build trust, develop a solid theoretical framework and establish safety bounds, ahead of planning an actual space mission demonstrator."