The Intel AI team has been reaching for the stars…literally. We have been collaborating with NASA’s Frontier Development Lab (FDL), an AI R&D accelerator that identifies and researches challenges faced by the space program. Our partnership with NASA FDL continued earlier this month at the NewSpace Europe conference in Luxembourg. Focused on the economic opportunities of space exploration, this conference highlights the exponential growth of the space ecosystem and brings together attendees from government agencies, investment banks, private space companies, start-ups, and investors.
The NASA FDL is an applied artificial intelligence research accelerator, developed as a partnership between the NASA Ames Research Center and the SETI Institute. The program tackles knowledge gaps in space exploration by pairing machine learning experts with astronomers and planetary scientists to address tightly defined problems; the format encourages rapid iteration and prototyping to create outputs with meaningful application to the space program.
For example, the NASA FDL Lunar Resource team worked with Intel and Space Resources Luxembourg to develop an algorithm based on a convolutional neural network (CNN) that analyzes optical images and digital elevation models of the Earth’s moon using an adaptive convolution filter. This analysis assists with lunar crater detection and labeling and ultimately may assist with planning future lunar exploration missions – and even determine the location of water. At the NewSpace event, Sara Jennings from NASA FDL shared that Intel will be partnering with the program again in 2018 and that more partners are needed.
Jacek Czaja, a machine learning engineer, then shared how AI can be applied to different aspects of space science and exploration utilizing varying techniques applicable to challenges including robotics, landing detection, traverse planning, remote sensing, and unknown model approximation. Jacek discussed Intel’s different innovative platforms like the Mobileye EyeQ®5 for real-time fused camera/radar inference, path planning, and in-vehicle road reconstruction, Intel® Xeon® Scalable processors for a wide variety of AI and datacenter workloads including deep learning training and inference, and the low-power Intel® Movidius™ Myriad™ X VPU for deep neural network inference at the edge.
The NASA Frontier Development Lab and Intel are continuing to develop applications and technologies that will significantly influence the future of the space industry. If you’re attending the 2017 NIPS conference, stop by the Intel booth to learn more about our collaboration – and how you can contribute!
This post was written by Sara Jennings, with input from Intel's Shashi Jain.