Ubotica Technologies and NASA JPL Present on Benchmarking Deep Learning Inference Of Remote Sensing Imagery Onboard The International Space Station
Kuala Lumpur, Malaysia, 20 July 2022,
Ubotica Technologies and NASA JPL are testing advanced computing that could pave the way to more advanced space missions using software based artificial intelligence running on advanced processors in space.
Future space missions will process imagery onboard to detect science events, create insights and respond autonomously – enabling new science. Spacecraft will respond to image dust devils or fresh impact craters on Mars, or automatically image around clouds on Earth!
One of the challenges to these future mission concepts is that current space computing has limited capabilities and is derived from Computers of the 1990’s to ensure reliable performance in the extreme environment of space – particularly radiation. Modern processors provide significant improvements in size, weight and power through innovative designs and manufacturing techniques.
Leveraging HPE’s Spaceborne Computer-2 on the International Space Station (ISS), JPL has partnered with Ubotica to test the CogniSat platform with the Intel Myriad X processor (used, for example, in terrestrial drones and low earth orbit satellites).
Among the models being tested are neural network models to interpret Mars satellite imagery. “Someday such a neural net could enable a satellite to detect new impact ejecta, evidence of a meteorite impact, and tell other spacecraft or take follow-up images”, says JPL Data Scientist, Emily Dunkel. “Rovers could also use these processors with neural networks to determine where it is safe to drive”, Dr. Dunkel adds.
“Testing CogniSat in space with algorithms currently in use on the ground, helps prove that this advanced hardware and software system is ready now for space missions,” says Léonie Buckley, Senior Engineer at Ubotica. “CogniSat provides significant efficiency improvements over alternatives to generate actionable insights in real time.”
Of significant importance is the deployment of a memory health checker, which determines the health status of the onboard DDR memory of the Myriad X. This memory checker provides vital information on possible radiation-induced memory corruptions which can lead to unreliable inference results. Thus far, no memory corruptions have been detected.
Software to plan rover activities is also being tested. By making the spacecraft smarter, the spacecraft will operate more autonomously and provide real insights using on-board smart processing which will accelerate scientific discovery in space. This heralds an new era of autonomous (or self-directed) exploration based on the latest compute paradigms.
Results from the testing will feature in a paper to be presented jointly by NASA JPL and Ubotica Technologies at IGARSS – the International Geoscience and Remote Sensing Symposium – the flagship conference of the IEEE Geoscience and Remote Sensing Society (GRSS), which takes place in Kuala Lumpur, Malaysia and online 17 – 22 July, 2022.
Founded in 2017, Ubotica Technologies provides smarts for smart satellites. Our products and services are used by global space industry partners to deliver real-time insights directly to users.
Ubotica is headquartered in Dublin, Ireland with a team of AI Engineers based in DCU Alpha, and has a team of Computer Vision Engineers in Spain based in UCLM, Ciudad Real, and a team of space systems experts in the Netherlands based in the Aerospace Innovation Hub at the TU Delft Campus.
The founders of the company have deep expertise in the advanced technology sector globally and have developed innovative technology, successfully brought complex products to market and delivered high value exits including Parthus Technologies, GloNav and most recently Movidius.