This comprehensive tutorial explores the foundational principles of Multi-Core Processing (MCP) analysis and its critical evolution into Artificial Intelligence (AI)-enabled safety-critical avionics systems. This session delivers a structured, hands-on examination of interference channels, resource contention, and deterministic behavior in multi-core environments, which are key challenges for modern digital avionics.
Participants will gain insight into the U.S. Army’s Multi-Core Processors Airworthiness Playbook (MAP) framework, including its roles-based lifecycle approach, Methods of Compliance (MoC), and alignment with industry standards such as DO-178C, DO-254, AC 20-193, and AMACC. Practical analysis techniques for interference mitigation, including cache coloring, shared memory partitioning, and formal scheduling methods, will be demonstrated through case studies.
The tutorial then transitions into the frontier of AI integration, introducing Multi-Kernel Execution (MKE) as a proven methodology to extend traditional multi-core safety assurance principles to GPU-based AI hardware. Attendees will analyze real profiler data from the BOB AI Engine, a mission-critical inference system, demonstrating how MKE overlays map parallel kernel behavior onto conventional scheduling frameworks for airworthiness validation.
The session concludes with an exploration of AI as an enabler and a tool: leveraging machine learning not only to power autonomous functions like sensor fusion and predictive maintenance, but also to assist in the certification process itself. Through demonstrations and structured material, this tutorial aims to equip engineers, architects, and safety practitioners with actionable methodologies for safely modernizing avionics platforms through MCP and AI, bridging legacy rigor with next-generation performance.
