Issue link: https://read.uberflip.com/i/1382624
WHITE PAPER Evaluating Safety-Certifiable Computing For Tomorrow 's Avionics mrcy.com 6 6 CERTIFIABLE ARTIFICIAL INTELLIGENCE New generations of avionic systems will incorporate a variety of Ar tificial Intelligence (AI), Machine Learning (ML) and Deep functions. These capabilities will significantly replace pilot workload in the cockpit, as well as improving pilot capability, decision-making and mission of Ar tificial effectiveness. These systems will face significant safety Deep Learning (DL) cer tification challenges, requiring robust processes for reduce and/ or ensuring safety-critical AI /ML /DL implementations. Mercur y used ar tificial intelligence as criteria for consideration for our processor evaluation. AI is rapidly gaining interest and becoming a requirement to suppor t future avionic applications. DEEP LEARNING (DL) Multi-layered neural networks learn from vast amount of data. Subset of Machine Learning DEEP LEARNING (DL) Image Stitching Pattern Recognition Blue Force/Red Force Identification Integrated Survivability Autonomous Target Recognition PILOT/MACHINE COLLABORATION Sense and Avoid Wire Detection Autonomous Landing Enhanced Cognitive Decision Making FULLY AUTONOMOUS, PILOTLESS AIRCRAFT Unmanned Aerial Vehicle (UAV) Teaming Autonomous Weapons Autonomous Flight Management System (FMS Real-Time Platform Maintenance (HUMS) MACHINE LEARNING (ML) Algorithm performance improves with data exposure. Subset of Artificial Intelligence ARTIFICIAL INTELLIGENCE (AI) A technology that appears to emulate human performance Includes both model - driven and data - driven approaches Current Fleet Capability Technology Development Future Vertical Lift Goals Urban Air Mobility platforms present huge challenges for safety certifiable process. URBAN AIR MOBILITY The emerging systems suppor ting Urban Air Mobility include both piloted and autonomous vehicles, with applications ranging from package deliver y UAVs to passenger-carr ying air taxis, all operating in densely populated areas. These systems will require sensor fusion for pilot-less and pilot- aided navigation, as well as AI applications that continually search for safe landing zones, power lines, and other airborne vehicles. Power ful processing in small, deployable form factors is needed for all these new airborne systems. Mercur y is working collaboratively with Intel to engage leading Urban Air Mobility platform providers to optimally address the technology and cer tification requirements.