The U.S. Air Force is building a software backbone for its autonomous drone fleet that could reshape how military and commercial aviation systems share the sky. By standardizing an open architecture for Collaborative Combat Aircraft missions, the service aims to let multiple vendors plug their autonomy software into any airframe without costly custom integration. The early results of this strategy are already taking shape, with two separate companies now advancing work under the same framework.
What A-GRA Means for Autonomous Flight
At the center of this effort is a certifiable architecture known as A-GRA, short for Autonomous Ground Reference Architecture. The concept is straightforward in principle but ambitious in practice: create a common software layer that any autonomous system can adopt (whether it flies a military strike drone or a civilian cargo aircraft). The Air Force and its innovation arm, AFWERX, are developing A-GRA through a Cooperative Research and Development Agreement with Reliable Robotics, a company known for its work on remotely piloted and autonomous fixed-wing aircraft. The CRADA structure matters because it signals a deliberate choice to co-develop the architecture with private industry rather than hand down a rigid government specification.
The design intent behind A-GRA is extensibility. Rather than building a system locked to one airframe or one mission type, the architecture is framed as adaptable across multiple platforms and operational scenarios. That includes integration with civilian airspace infrastructure, specifically detect-and-avoid technology and air traffic control interoperability. For anyone who has watched military drone programs struggle with National Airspace System access, this dual-use design philosophy represents a meaningful change. If A-GRA delivers on its certifiability promise, it could allow autonomous military aircraft to operate in mixed airspace without requiring blanket flight restrictions, a persistent friction point for current unmanned systems.
Shield AI Brings Hivemind to CCA
The open-architecture strategy gains sharper definition through a second, parallel development. Shield AI has been selected as a mission autonomy provider for the U.S. Air Force Collaborative Combat Aircraft program, specifically supporting the CCA Technology Maturation and Risk Reduction phase, known as TMRR. The company’s Hivemind software, an AI pilot system originally developed for small quadcopters and later scaled to fighter-class operations, will serve as the decision-making brain for autonomous wingmen designed to fly alongside crewed jets.
What ties these two announcements together is compliance. Hivemind is described as A-GRA compliant, meaning Shield AI has built its autonomy stack to fit within the same open framework that Reliable Robotics is helping the Air Force define. This is the practical payoff of an open architecture: two different companies, working on different aspects of autonomous flight, can develop their systems in parallel and still expect them to work within a shared standard. The alternative, proprietary autonomy stacks tied to specific airframes, would force the Air Force to manage a fragmented ecosystem where every new platform requires bespoke software integration.
Fury Integration and Flight Testing Ahead
Shield AI has already moved beyond paper compliance. Hivemind has been integrated on Anduril’s Fury drone, designated the YFQ-44A, for system-level testing. Anduril’s Fury is one of the leading CCA airframe candidates, a jet-powered autonomous aircraft designed to serve as an affordable, attritable wingman for crewed fighters like the F-35. The fact that Hivemind is already running on Fury hardware suggests the A-GRA compliance pathway is producing tangible integration results, not just theoretical compatibility.
Flight demonstrations of the Hivemind integration on Fury are expected in coming months. Those tests will be a critical proof point. Software that performs well in lab-level integration can still fail when exposed to the latency, sensor noise, and unpredictable conditions of real flight. If the demonstrations succeed, they will validate not just Hivemind as a product but A-GRA as a viable standard for mixing and matching autonomy providers with airframe builders. That validation would strengthen the Air Force’s negotiating position with future vendors, since any company wanting a piece of the CCA program would need to build to the same open spec.
Why Open Architecture Changes the Cost Equation
The strategic logic here extends well beyond engineering convenience. Legacy military aircraft programs have historically locked the government into long-term relationships with single prime contractors, partly because proprietary software and hardware made switching costs prohibitive. An open architecture like A-GRA directly attacks that dynamic. If the Air Force can credibly tell autonomy vendors that their software must be interchangeable, it creates competitive pressure that should drive down costs and accelerate development timelines. Think of it as the difference between buying a phone that only works with one carrier and buying an unlocked device that lets you shop for the best plan.
There is a reasonable counterargument, though. Standardization can sometimes slow innovation by forcing vendors to design within constraints that may not suit every mission profile. A company with a genuinely superior approach to autonomous decision-making might find A-GRA compliance limiting if the architecture cannot evolve fast enough to accommodate novel techniques. The CRADA model partially addresses this concern by keeping industry involved in the architecture’s development, but the tension between standardization and cutting-edge performance is real and will likely intensify as CCA missions grow more complex. The Air Force will need to update A-GRA regularly or risk the standard becoming a ceiling, not a floor.
Dual-Use Ambitions and Remaining Questions
Perhaps the most underappreciated dimension of A-GRA is its civilian airspace integration component. The architecture is designed to work with detect-and-avoid systems and air traffic control protocols, which means the same software framework governing a CCA drone in contested airspace could, in theory, govern an autonomous cargo aircraft flying a domestic route. For companies like Reliable Robotics, which are pursuing FAA certification for remotely operated aircraft, contributing to a military architecture that shares civilian safety standards is a strategic investment. It could accelerate their own certification timelines while giving the Air Force access to safety engineering expertise that pure defense contractors may lack.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
