Intel's Movidius Myriad 2 VPU Takes Artificial Intelligence Into Space Aboard the PhiSat-1

AI chip analyzes imagery for cloud cover then discards useless imagery prior to transmission, saving about 30 percent of its bandwidth.

The compact system performs image analysis in space, rather than waiting for transmissions. (📷: Tim Herman/Intel)

Intel's Movidius Myriad 2 vision processing unit (VPU), designed to accelerate computer vision applications in a small power envelope, has reached new heights — literally. It's now in space, on-board the experimental PhiSat-1 sun-synchronous satellite.

"The capability that sensors have to produce data increases by a factor of 100 every generation, while our capabilities to download data are increasing, but only by a factor of three, four, five per generation," explains Gianluca Furano, data systems and onboard computing lead at the European Space Agency, which led the collaborative effort behind PhiSat-1, about the problem which led to the novel solution: Putting an AI accelerator on the satellite itself to perform processing in-space, rather than transmitting raw data back to Earth.

"Space is the ultimate edge," says Aubrey Dunne, chief technology officer of Ubotica, the Irish startup which built and tested the technology. "The Myriad was absolutely designed from the ground up to have an impressive compute capability but in a very low power envelope, and that really suits space applications."

The AI system needed to be as low power as possible, to run from the satellite's solar panels. (📷: Tim Herman/Intel)

The PhiSat-1 is designed to monitor polar ice and soil moisture while simultaneously testing inter-satellite communications. Its camera takes shots of the Earth as it orbits, but any with cloud cover need to be discarded as useless — a task now performed on-satellite through the Myriad 2, saving around 30 percent of the satellite's bandwidth compared to transmitting all imagery and discarding the useless shots back on the ground.

The Myriad 2 chip needed to be tested for its reaction to radiation exposure, which it passed with flying colors in its off-the-shelf form. The system was trained on synthetic data generated from previous missions based on differing cameras, and the whole unit launched this September and confirmed now as fully-functional — and it could be updated in the future to run a different neural network for a different scientific mission.

The European Space Agency has confirmed plans to launch a second Myriad 2-equipped satellite in the near future.

ghalfacree

Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.

Latest Articles