Pentagon’s New Autonomy Command Should Learn from History and Today’s Advances

Pentagon leaders are pursuing a fresh push to speed up the development and deployment of autonomous and robotic systems. This effort can leverage key lessons from past attempts and recent breakthroughs.

The forthcoming Autonomous Warfare Command, expected to launch by October, aims to foster “not a particular drone, but the ability to adapt in combat” against “realistic opposition,” according to a September 30 memo from Defense Secretary Pete Hegseth. Hegseth warned, “We do not have a decade to transition our procurement and doctrine.”

Rapid adaptation is essential because modern battlefields demand changes within weeks, whereas defense contractors often need months or years to revise designs or issue software patches. The Pentagon, however, is not starting from scratch.

Start with the mission, not the robot

Effective planning begins by consulting commanders and operators to understand their needs, rather than impressing them with demonstrations that may not align with operational requirements.

A case in point is the Maven Smart System, an AI intelligence platform adopted by U.S. and NATO forces, which exemplifies how to deliver emerging technology quickly. A 2024 CNAS report noted that “a clear, urgent, but broad problem definition allowed Maven to meet user needs more directly.”

Combatant commanders are now seeking systems that enhance domain awareness. As General Frank Donovan, head of U.S. Southern Command, explained at a Defense One event in May: “Most of the systems we’re looking at are primarily our domain awareness systems, and…We can match the robots to the environment whether it swims, it flies, it has feet, whatever it does.” Donovan, who previously led the Defense Autonomous Warfare Group, observed that robots are needed not only to replace personnel but also to improve situational understanding in areas where troops are not present. He emphasized the challenges faced by partners operating in difficult terrains, jungles, and maritime zones.

Build communicative autonomy

Much of the work at the Defense Autonomous Warfare Group focused on fostering dialogue among partners about the data they require from robotic systems and the information those robots should possess. “We don’t talk about robots. We talk about the data environment,” Donovan noted. “With the different data layers we need at the forward edge, our special operations forces and conventional teammates can plug a cellphone into that network and instantly leverage any capable robot that arrives.” Autonomous systems become even more critical when adversary electronic warfare disrupts communications. A former senior defense official highlighted that, “The next step is true autonomy that can operate when GPS and communications are denied—something we already see in Ukraine and the Middle East and will certainly encounter in future conflicts.”

Buy the brains and body separately

The automotive industry offers a useful model: it separates vehicle hardware from autonomous driving software by partnering with specialized firms. In contrast, the Pentagon has typically asked manufacturers to integrate autonomy with the platform itself. For instance, buyers looking for autonomous drones have simply requested that producers modify existing remote‑controlled aircraft. This approach is flawed, according to the former official. “Drone makers excel at building platforms, but autonomy is its own discipline. Commercial self‑driving solved this years ago by having automakers build great vehicles and partner with autonomy companies for the driver. Defense needs the same model.”

Make it modular

Procurement should mandate modular autonomy packages that can be upgraded as threats evolve and technology advances. This principle extends beyond individual weapons to the entire force and allied networks. “Open standards are what make that partnership work at scale,” the official added. A recent NATO experiment in Portugal illustrated this concept. The Centre for Maritime Research and Experimentation (CMRE) concluded its annual Robotic Experimentation and Prototyping exercise with maritime unmanned systems. This year’s demonstration showcased robots receiving commands from software rather than humans. “For the first time we materialized our vision for future operations,” said João Alves, a CMRE principal scientist. “Algorithms assessed the mission and automatically assigned tasks to drones.” This success stemmed from a policy shift—partner nations agreeing to adopt common system standards.

Instill trust

The primary obstacle for any command using autonomous weapons is determining when to rely on them. A 2017 study of autonomy safety failures, particularly involving Patriot missile batteries, underscored this point. The most notable incident occurred in 2003 when a Patriot inadvertently shot down an F‑18 in Iraq. “The tactical director at the battalion command and control node gave the order, ‘Bring your launchers to ready,’” wrote Army engineering psychologist John Hawley. “That directive was tantamount to an order to engage, but that was not what the tactical director intended.” Even with a clear mission, open standards, shared data, and proper training, Hawley concluded that the greatest challenge remains “sustained high reliability in a complex and unpredictable operational setting.” The question of whether software truly understands user intent has become increasingly urgent as AI agents occasionally act contrary to expectations. For AUTOWARCOM to chart a viable path, its leaders must learn from past autonomous system successes and failures and recognize that human trust cannot be automated.

Source link