The European Space Agency (ESA) is setting its sights on trapping a piece of space junk 660 km above the Earth.

The European Space Agency is embarking on its first mission to remove space debris from orbit. At a rough estimate, about 34,000 pieces of junk are presently orbiting the Earth, and their presence is a danger to space missions carrying cargo and humans. At risk also are expensive satellites that perform essential services for mankind such as communications, weather prediction, and the Internet. The ESA’s first target is to recover an obsolete payload adapter – the Vespa Upper Part – that is currently orbiting at 660 km above the Earth. (EPFL)

Recovering the Vespa payload adapter

The Vespa (Vega Secondary Payload Adapter) Upper Part (see in the picture above) was once a part of the European space agency’s Vega Rocket.

The mission objective is to trap the Vespa within the robotic arms of a capture rocket. It will then pull it back into the Earth’s atmosphere and allow it to burn up in a controlled and safe manner.

Three key problems arise in this endeavor.

• First, the capture rocket’s cameras (its ‘eyes’) must reliably recognize the Vespa. Note that nobody has seen that object for over seven years as it’s been spinning in a vacuum in space.
• Second, the robotic arms must reach precisely the exact location of Vespa from the correct angle
• Third, to perform and manage the recovery operation out in space in real-time and with limited onboard computing power.

Clearspace and AI

This groundbreaking project is being tackled by Clearspace, an offshoot spun off from the EPFL Space Centre. Its mandate is to develop technologies to capture and deorbit space debris.

The mission to recover the Vespa upper part is set for 2025. Artificial intelligence and machine learning algorithms will play a central role in the project.

Deep learning algorithms are training to recognize the Vespa using synthetic images in great detail, even including the Earth as a backdrop. To help the algorithm recognize Vespa, the scientists are generating synthetic images of the target debris. To do this they are using samples of aluminum and carbon fiber panels – the material of its construction.

They are generating synthetic images using a goniophotometer.

“This is a large robotic device that spins around a test swatch to simultaneously illuminate and observe it from many different directions, providing us with a wealth of information about the material’s appearance,” says Assistant Professor Wenzel Jakob, head of EPFL’s Realistic Graphics Lab.

To solve the second problem regarding the angle and approach for the capture of the Vespa, again it is AI to the rescue.

The scientists are developing deep learning algorithms to “reliably estimate the 6D pose (3 rotations and 3 translations) of the target from video-sequences even though images taken in space are difficult.”

The team is now transferring the deep learning algorithms to a dedicated hardware platform. Note that these algorithms need to be sophisticated enough to work autonomously in space. Their complexity puts a massive strain on computational resources.

“Their implementation requires squeezing out all the performance from the platform resources,” says Prof David, Atienza, head of EPFL’s Embedded Systems Lab.

Related Story:     Intel First To Deploy AI “On Edge” In Space