As space infrastructure becomes more important, so satellite-based services are now fundamental whether in terms of global navigation, weather forecasting, or secure communications, and so consequently the space industry is seeing a boom in satellite launches.
EDGX is developing a new data processing unit concept that mixes classic AI acceleration with neuromorphic computing to decrease onboard AI power dissipation, increase adaptability and ultimately enable autonomous learning capabilities onboard next generation satellites.
“We built the Akida roadmap with a vision to deliver high-performance, ultra-low power, self-contained processing that didn’t constantly need connection to the Cloud, which is critical for constrained environments like space,” said Peter van der Made, Founder and CTO at BrainChip. “BrainChip’s focus on combining neuromorphic principles with the benefits of the deep learning ecosystem, customisation, and continuous learning, matches EDGX’s requirements for disruptive technology.”
EDGX and BrainChip are building up their partnership and have signed a Non-binding Memorandum of Understanding.
“What fascinates us the most about Akida is its capability to run in two different modes: Firstly, Akida can run conventional CNN, making it compatible with so many proven neural networks out there. But then, running in event-based mode, it cuts down power demand by orders of magnitude. And power is at premium in the space environment. Event based processing may easily open a door into a new realm of processing in space to do things which were not possible before,” said Laurent Hili, Microelectronics and Data Handling Engineer at ESA.
