Real-Time Sensor Data Processing and Compression Performed On-board Unmanned Aerial Systems (UAS)

ABSTRACT: The K-128 co-processor leverages insights from neuroscience to achieve superior performance. Unlike previous efforts along those lines, our approach is to create a software programmable co-processor that runs existing applications. One of the most salient characteristics of such a co-processor is its power efficiency: not just a few percent more efficient than a traditional processor, but many times more efficient. Dragonfly Pictures, Inc. (prime) and Cognitive Electronics (CE) propose to test the K-128 as an onboard visual sensor processing capability to support automated real-time identification, tracking, and selection of regions of interest (ROIs) based on environment, target sets, and mission priorities. BENEFIT: High Performance Computing hardware is a $10 billion dollar market that is growing at 8% per year according to market research firm IDC. In addition, the market for High Performance Cloud Computing hours is currently $177 million, and growing at 71% per year, again according to IDC. Among the end markets for HPC, the Federal Government is the largest, buying over $2 billion of High Performance Computing hardware per year. The K-128 processor will be tested using video processing software and ISR data to serve automate real-time identification, tracking, and selection of regions of interest. The K-128 brings a 100x performance per watt advantage.