NASA SBIR 2020-I Solicitation

Proposal Summary


PROPOSAL NUMBER:
 20-1- H6.22-5109
SUBTOPIC TITLE:
 Deep Neural Net and Neuromorphic Processors for In-Space Autonomy and Cognition
PROPOSAL TITLE:
 DNN Radiation Hardened Co-processor companion chip to NASA's upcoming High-Performance Spaceflight Computing processor
SMALL BUSINESS CONCERN (Firm Name, Mail Address, City/State/Zip, Phone)
Numem Inc
440 N Wolfe Road
Sunnyvale, CA 94085
(408) 836-8795

Principal Investigator (Name, E-mail, Mail Address, City/State/Zip, Phone)

Name:
Mr. Nilesh Gharia
E-mail:
ngharia@numem.com
Address:
440 N Wolfe Road Sunnyvale, CA 94085 - 3869
Phone:
(408) 836-8795

Business Official (Name, E-mail, Mail Address, City/State/Zip, Phone)

Name:
Mr. Nilesh Gharia
E-mail:
ngharia@numem.com
Address:
440 N Wolfe Road Sunnyvale, CA 94085 - 3869
Phone:
(408) 836-8795
Estimated Technology Readiness Level (TRL) :
Begin: 3
End: 4
Technical Abstract (Limit 2000 characters, approximately 200 words)

New Space Directive has opened up new challenges and opportunities in AI applications. Numem aims to augment NASA’s High-Performance Spaceflight Computing (HPSC) program with design of radiation hardened & ultra-low power DNN Co-processor to enable AI applications.

There is exponential increase in the use of sensors. These sensors and connected devices generate zettabytes of data per year. Machine learning with DNN capability is needed to extract meaningful and actionable information from this data. For some applications, the goal is to take immediate action based the data such as in robotics/drones, self-driving cars, smart Internet of Things whereas in other applications, the goal is to analyze and understand the data to identify trends as in case of  surveillance, portable/wearable electronics.

This radiation hardened Co-Processor will solidify HSPC avionics ecosystem with robust AI capabilities much needed for space autonomy, small sat constellations/autonomous science, human explorations and operations Habitat and deep space missions. On-board AI processing can enable spacecraft to efficiently process large volumes of raw sensor-data into scientific knowledge or actionable data to overcome limitations in downlink communication. On-board artificial intelligence can also enable spacecraft to formulate decisions for critical operations. Commercial applications

This AI core comprises of reconfigurable DNN Engine with multiple compute units which can support multiple DNN models and sizes. Embedded STT-MRAM Memory which is 1000X high performance compared to FLASH memory and improves the silicon area by 2X to 3X compared to SRAM and reduces the standby-power by about 5X with radiation tolerant memory cell.

This robust solution is crafted for low power machine vision, autonomous vehicles, facial recognition, healthcare, real-time tracking, agriculture, manufacturing.  Radiation tolerant memory technology is ideal fit for space qualifications.

Potential NASA Applications (Limit 1500 characters, approximately 150 words)

AI Applications in Spacecraft, Lander, LEO Satellites, Lunar/Mars/Deep space missions, Planetary Real Time AI/ML Data Mining for Solar Flare, Planet Structure, Space Stations, Aerospace and Defense, Satellite Imagery Analysis, Global Surveillance Analysis, Near Earth Object Trajectory, Earth Climate prediction, Satellite Scans Analysis of Illegal Fishing in Oceans, Satellite Imagery on Deforestation, Satellite based Earth Observation Market Applications such as oil and gas industry, crop yield detection, pipeline leakage detection etc.

Potential Non-NASA Applications (Limit 1500 characters, approximately 150 words)

All low power AI applications on Edge, Autonomous cars, IOT, Wearables, Mobile, Quantum Computing, Robotics, Drug Discovery, 5G, Particle Physics, Brain-Machine Interface,etc. Any DNN applications with need of low power vision classification, vision detection, speech recognition, natural language processing, audio recognition, social network filtering and machine translation are ideal fit.

Duration: 5

Form Generated on 06/29/2020 21:10:20