Unmanned Doesn’t Mean Autonomous — And the Industry Keeps Getting This Wrong
One of the most overused words in defense technology right now is “autonomy.” The Department of War and lawmakers are focused on the future of drone warfare, and defense tech is the new VC-fueled start-up flavor of the year, with the word “autonomy” showing up everywhere, ranging from PowerPoints to budget documents, briefings, and headlines. The problem is that companies and people use the word to mean different things, depending on their agenda. But this is not simply a matter of differing opinions, or a situation where reasonable minds can disagree.
ASTM International, the primary global standards organization that develops technical requirements for aviation and aerospace systems, defines an autonomous system as one that has the authority to “independently determine a new course of action in the absence of a predefined plan to accomplish goals based on its knowledge and understanding of its operational environment and situation.” This compares to an automated system that ASTM defines as “hardware and software that automate a predefined process without the need for human intervention…“
I spend a lot of time in conversations where the word “autonomy” is used to describe what is in fact nothing more than automation, not dissimilar to how people often describe “if-then” software programming as “AI.” Both misuses create real confusion. The industry has taken one word and stretched it to make it sound sexier than it really is.
Waypoint flight, collision avoidance, formation flying, and object tracking all get called autonomy. Those are all useful capabilities. But they are not all the same thing as autonomy. They are automation. And referring to them as autonomy obscures the fact that automated machines cannot change the script after launch to react to real-time conditions without human intervention.
The distinction matters. Automation means a machine follows instructions that a human defined in advance. A factory robot welding the same seam over and over again is automated. A drone flying to a pre-programmed coordinate is automated. A system that locks onto a selected vehicle and follows it is performing an automated task.
Autonomy is different because it begins when the machine encounters something unexpected and has to decide what to do next within the mission parameters a human has given it.
The automotive industry ran into this same problem years ago. Cruise control, lane keeping, and highway assist are not the same as a vehicle that can reason through an unexpected situation and safely determine its own next action.
That is why the auto industry eventually needed a levels framework, because the terminology became too loose. Defense technology needs the same discipline.
Consider a simple drone mission. A drone follows a vehicle, and when the vehicle stops, a person exits and runs away. What should the drone do? If the drone was instructed in advance to “follow that vehicle no matter what,” then the drone sticks with the vehicle. That is automation. In a true autonomous architecture, the drone understands the mission priority, evaluates what changed, determines that the person is now the priority object to follow, and adjusts its behavior accordingly; that is autonomy.
Unmanned Is Not the Same as Autonomous
There is a second mix-up happening alongside the automation-autonomy confusion, and it may be the more common one: people are calling anything without a person on board “autonomous” simply because it is unmanned.
Unmanned only describes the platform. It means no human is physically riding inside the vehicle. It says nothing about who, or what, is making the decisions. A drone flown by a soldier holding two joysticks is unmanned. It is also completely manual. There is no autonomy in that system, automation or otherwise, because a human is making every decision in real time from a remote location.
This distinction matters because so much of the current defense narrative treats “unmanned” as a proxy for “intelligent.” Headlines describe “autonomous drone swarms” when what is actually flying is a set of remotely piloted aircraft with no autonomy and no real collaboration. Investors hear “unmanned” and assume the human has been removed from the loop entirely, when in many cases the human has only been removed from the cockpit.
The right question is never whether anyone is sitting inside the vehicle. The right question is who, or what, is making the decisions, and under what conditions. A system can be unmanned and fully manual. It can be unmanned and automated. It can be unmanned and autonomous. Those are three very different capabilities that happen to share the same empty seat.
A soldier with a joystick in each hand trying to manually fly a drone is managing altitude, camera angle, speed, direction, battery life, and is not thinking about the whole battlefield. That is a lot of cognitive load for an unmanned aircraft with no autonomy, which keeps the operator preoccupied and without the bandwidth to keep the bigger picture in focus.
Automation can reduce that load, until something goes off script. Autonomy not only reduces the load, but it also keeps the cognitive load low when the unforeseen happens, which it frequently does. Collaborative autonomy takes it a step further.
A single autonomous drone can make decisions about its own mission. A collaborative swarm allows multiple drones to share information, adapt roles, and work together as a team. We believe that represents a fundamentally different capability from a choreographed formation or a group of drones waiting for instruction from a central controller. Coordination can still imply that the behavior was written in advance. Collaboration means the machines exchange useful information and adjust their behavior in real time.
Nature solved this problem a long time ago.
Ants do not wait for their queen to tell them what to do. Nor do bees. Wolves hunting as a pack do not operate from a fixed script. They sense, communicate, adapt, and change roles based on what the pack is trying to accomplish. That is the model we have built around.
At Palladyne AI, we call this philosophy Decentralized Embodied Collaborative Autonomy, or DECA. The intelligence is embodied on each machine. It operates at the edge. It is decentralized, so it does not require centralized connectivity. It is domain-specific, which means the machine has the intelligence it needs for the mission it is assigned, rather than a generalized brain trying to do everything.
That matters in defense because the battlefield is a contested environment. Communications can be degraded. GPS can be denied. Bandwidth can disappear. Latency, or more importantly, the lack thereof, can become the difference between success and failure. A machine that has to wait for instructions from a distant cloud is not truly autonomous in the way the battlefield requires.
The next phase of defense technology will not be defined only by how many drones we can build, though that matters and we need to build far more of them. The harder question is how those drones will be used. One operator per drone does not scale. One human manually managing every camera, every flight path, every sensor, and every tactical adjustment does not scale. Pre-programmed routes do not scale when the environment changes faster than the plan.
That is why words matter. When every feature gets called “autonomy”, operators, policymakers, investors, and procurement officials lose the ability to distinguish between a useful function and a true mission-level capability. The better question is not simply whether a system is “autonomous.” It is what decisions can the machine make, under what conditions, with what human oversight, and at what level of collaboration with other machines?
That is where the industry needs more clarity because in defense technology, the difference between automation and autonomy is operational.