Enhancing and productizing the DVMR rad-hard digital AI coprocessor

The project will take a TRL 6 rad-hard AI co-processor developed during the SBIR Sequential award. The co-processor, called DVMR, represents the state-of-the-art for AI inference at the Edge, both on the ground and in Space. The DVMR is capable to run large neural network inference at speed higher than real-time, while consuming orders of magnitudes less power than the alternatives (14x more efficient than the best NVidia Jetson system). With this project, we will enhance the DVMR with capabilities requested by our customers' engagements. We will design, fabricate, and test a faster chip, capable of analyzing much larger data sizes, and utilizing the latest revolutionary Transformer-based Neural Networks. Moreover, we will implement the necessary Design For Testing (DFT) features that are required for high-volume production. We will deliver, by the end of the project, 2 fully working enhanced DVMR systems to NASA that will have been tested and verified under radiation. Other fabricated chips will mounted on Evaluation Board Kits to be sent to customers for Sales development. The targeted markets are: Space, Drones, High-end security cameras and Police bodycams.
Benefits: The DVMR has been designed to survive more than 200krad of TID and can be used in Space, from LEO to Lunar environment. The DVMR is compatible with neural networks that can enhance and reconstruct low-quality data (images, radar/lidar data, comms) and analyze data to prioritize the downlink or to be used for autonomy and decision making. Also, the DVMR can be used for data fusion and anomaly detection in complex data streams. The DVMR is the most SWaP-efficient AI solution and can, therefore, be applied for Earth observation, deep space (except in the Jovian environment), Lunar observation, or on the Lunar surface (ex, autonomous rovers for ISRU). Our current customer engagements are in: - Earth Observation satellites - Police Bodycams - High-end security cameras - drones navigation In the future we will engage customers also in the smartphones, Automotive ADAS systems, Healthcare and Robotics markets.