When people think about autonomy in aviation, they often focus on what it can do — flying complex missions, reacting to dynamic environments, making intelligent decisions in real time. But the real challenge isn’t getting autonomy to work, rather it’s proving that it can be trusted. 

For autonomous flight to scale across aircraft and mission sets, it must meet the same airworthiness expectations we apply to human pilots. This isn’t just paperwork; it’s the difference between a demo and a system that can safely operate in the real world. 

Laying the Groundwork for Assured Autonomy 

This principle drives Resilient Shield, an internal research and development initiative focused on creating an ASTM F3269-21–compliant, modular, and adaptable Multi-Monitor Runtime Assurance (MM-RTA) capability for Hivemind, Shield AI’s AI-enabled mission autonomy software. The program is building the technical and certifiable foundation that allows Hivemind to operate on airworthiness-critical platforms with confidence and safety assurance.  

Resilient Shield is a collaboration between Level 5 Autonomy Labs (L5A), led by Mark Skoog, a pioneer in safety-critical autonomy. Last year, Shield AI and L5A teamed up to define and build the foundational MM-RTA architecture — a certifiable framework for bounding AI-driven behaviors in flight. Skoog, who spent decades at NASA’s Armstrong Flight Research Center leading the development of automatic collision avoidance systems and multi-monitor autonomy architectures, provides a strong foundation for how Shield AI approaches airworthiness and certification readiness. 

Since its kickoff in the fall of 2024, Resilient Shield has advanced rapidly through design, simulation, and flight testing. The development of the MM-RTA provides the safety backbone that enables Hivemind mission autonomy while maintaining confident, safe operations. 

How Runtime Assurance Works 

MM-RTA functions as a continuous safety layer for autonomous systems. It sits between Hivemind and the aircraft’s flight control software, constantly monitoring commands and system behavior. If an unsafe condition arises such as a potential ground or air collision, MM-RTA intervenes to maintain safe flight and enable recovery. Its responsibility is simple: ensure that autonomy commands do not pose a flight safety risk. 

By building a modular and platform-agnostic MM-RTA architecture, the Resilient Shield team is ensuring that Hivemind can be integrated safely across a range of aircraft from small Group 1 systems to large Group 5 Collaborative Combat Aircraft (CCA). 

Milestones and Momentum 

This year marked major progress for the Resilient Shield team. In June, Shield AI completed the first flight of an MM-RTA-enabled Ground Collision Avoidance System (GCAS) on a Cessna 170. The test aircraft carried an IMX8 processor running GCAS that successfully issued “Pull Up” alerts when it detected a simulated terrain conflict. The test validated the core safety logic and responsiveness of the system. 

In August, the team integrated GCAS with Hivemind aboard a Believer remote control aircraft, testing the system with Hivemind mission autonomy in the loop for the first time. Hivemind, running on an Orin Nx, communicated with the Believer’s Ardupilot system on a CubeOrange. During testing, Hivemind autonomously planned waypoints and Heading Speed Altitude (HSA) into simulated terrain hazard. GCAS successfully interved to prevent impact proving a major step toward real-time safety assurance.  

In November, we reached another milestone by combining GCAS and Well Clear capabilities aboard a Bearhawk 5 aircraft. Well clear provided baseline separation standards for air collision avoidance using an ADS-B transceiver as the traffic sensor in four different collision scenarios: Overtake, Head On, Shallow Crossing, and Climbing Into. A combination Well Clear and GCAS scenarios was also flown where the aircraft was pinched between air traffic to the right and terrain, making a left turn and climb to avoid a collision. 

Looking Ahead 

Next, the Shield AI and L5A teams are developing an Air Collision Avoidance System (ACAS) capability. Phase 1 will focus on handling non-cooperative aircraft, addressing late detections, unexpected maneuvers, and other hazards. Phase 2 will integrate with existing ACAS/Traffic Alert and Collision Avoidance System (TCAS) equipment to meet civil and commercial airspace certification requirements. 

By mid-2026, Shield AI intends to make MM-RTA available as part of the Hivemind Autonomy SDK, enabling customers to integrate certified safety assurance into their own autonomous systems. MM-RTA is being developed to meet rigorous software assurance standards, supporting customers in building certifiable, safety-critical autonomy solutions. 

 Each Resilient Shield flight test reinforces our confidence that Hivemind can meet the safety and reliability standards required for certified flight. By combining GCAS, Well Clear, and soon ACAS under a unified MM-RTA framework, Shield AI is laying the groundwork for certifiable autonomy, advancing toward the day when Hivemind can safely operate on any aircraft, in any airspace.  

About the Author 

Willy Logan is a Senior Director of Engineering at Shield AI, where he leads a team of Chief Engineers integrating the company’s Hivemind software across unmanned aircraft. Before joining Shield AI, Willy spent more than 12 years at General Atomics Aeronautical Systems (GA-ASI), where he held roles including Deputy Systems Engineering Director for the MQ-9B UK Protector and Mission Systems Chief Engineer. He holds both a B.S. and M.S. in Mechanical and Aeronautical Engineering from UC Davis, and lives in San Diego, where he enjoys surfing and finding the city’s best tacos with his family.