AI's Emerging Role as 'Copilot'
Edge AI is increasingly being framed as a cognitive augmentation tool for pilots, not a replacement. A recent technical presentation highlighted how AI can reduce pilot workload by filtering and prioritizing data for better situational awareness. The industry push is toward explainable AI (XAI) that makes recommendations transparent and trustworthy in the cockpit.
DARPA's Aircrew Labor In-Cockpit Automation System (ALIAS) program aims to develop a removable kit to add high levels of automation to existing military aircraft, reducing the need for onboard crew. Sikorsky, a Lockheed Martin company, is a key partner in this initiative, leveraging its MATRIX Technology to enable complex missions with minimal human oversight. The system is designed to manage an entire flight from takeoff to landing, including handling in-flight emergencies. The U.S. Air Force is aggressively pursuing AI-driven autonomous systems through its Collaborative Combat Aircraft (CCA) program, developing "loyal wingmen" to fly alongside human pilots. Companies like Anduril Industries, General Atomics, Lockheed Martin, and Northrop Grumman are creating prototypes, such as the YFQ-44A and YFQ-42A, designed for reconnaissance, electronic warfare, and strike missions with minimal human intervention. This initiative is supported by AFWERX, the innovation arm of the Air Force, which accelerates the development and integration of advanced technologies from startups and small businesses. For edge AI deployment in resource-constrained aerospace environments, Field-Programmable Gate Arrays (FPGAs) are often favored due to their reprogrammability and efficiency in parallel processing, which is crucial for real-time applications like image processing. Application-Specific Integrated Circuits (ASICs) offer high performance and energy efficiency for a single, predefined neural network but lack the reconfigurability of FPGAs. AMD Xilinx's XQR Versal is an example of a dedicated AI accelerator designed for the high-radiation environment of space. Certifying AI-based systems under the DO-178C standard presents a significant challenge for commercial and military aviation. Since AI models can have "hallucinations" and produce erroneous outputs, they are best used as advanced assistants rather than autonomous creators of certified code. Current approaches treat AI-generated code like human-written code, requiring that every line be traced back to a requirement, be independently reviewed, and pass deterministic testing. Explainable AI (XAI) is critical for building trust in these autonomous systems. Techniques like LIME (Local Interpretable Model-agnostic Explanations) and SHAP (SHapley Additive exPlanations) help to make the "black box" of AI decision-making more transparent to pilots and regulators. This transparency is essential for validating AI-driven strategies and ensuring they comply with stringent safety standards. Several startups are emerging in the AI copilot space. Beacon AI, founded by a former US Navy F-18 pilot, has secured $20 million in funding to develop an AI assistant for both commercial and military cockpits. Meanwhile, MIT's "Air-Guardian" system uses eye-tracking technology to monitor a pilot's focus and can take control of the aircraft in an emergency. These systems aim to enhance pilot decision-making and improve flight safety.
