When Russia invaded Ukraine in February 2022, many military pundits expected a truncated campaign that would be over in just a matter of days. Even the former U.S. Chairman of the Joint Chiefs of Staff, Gen. Mark Milley, predicted that Kyiv would fall within 72 to 96 hours.
What actually happened was quite different. Ukraine’s willingness to fight and its ability to rapidly innovate and deploy new weapon systems have enabled it to hold off an overwhelmingly large and incredibly sophisticated military threat for more than four years.
One of the keys to the country’s success has been its use of unmanned aerial vehicles (UAVs). Ukraine’s military has managed to destroy some of Russia’s exquisite bomber aircraft and naval vessels, together costing billions of dollars, leveraging inexpensive UAVs costing thousands of dollars each. This incredible feat has the world’s largest militaries reevaluating the solutions they’re developing and how they’re spending defense dollars, with a new emphasis on low-cost, attritable autonomous drones.
In fact, recently, Secretary of War, Pete Hegseth, released a memo illustrating the importance of autonomous UAVs to the future of America’s military.
The directive outlined in the memo effectively pulls all drone and autonomous programs away from the disparate services and consolidates them under a Direct Reporting Portfolio Manager for Unmanned Offensive and Defensive Systems (DRPM-UxS). Calling drones and autonomous systems “the most consequential battlefield innovation of this generation,” the memo claims that America has fallen behind other militaries in producing unmanned systems and must make changes to “move at the speed this moment demands.”
However, as this strategic shift in focus and investment into autonomy continues, and America explores autonomy more earnestly, it’s important to remember that not all approaches to autonomy are the same.
Remotely Piloted vs Preprogrammed vs True Autonomy
I recently had the opportunity to observe the Ukrainian military’s drone operations in person and was incredibly impressed with what they’ve managed to accomplish. However, it appears many of their drone systems continue to depend largely on remote piloting, or on a simpler form of autonomy involving preprogrammed missions executed by advanced UAVs.
Remotely piloted UAVs have been part of the U.S. military’s arsenal for approximately 25 years. They played key roles in the Global War on Terror and will invariably remain a useful and effective tool for the military for years to come. However, remotely piloting UAVs has its challenges – particularly when it comes to leveraging UAVs as a force multiplier.
The U.S. Department of War (DoW) is preparing for conflicts against peer and near-peer adversaries with increasingly sophisticated intelligence, surveillance, and reconnaissance (ISR) capabilities and next-generation effectors. These adversaries can identify and hold traditional weapons systems at risk well before they’re within range to attack. To overcome this, each service has introduced a new operational concept that would leverage an overwhelming force of distributed weapons systems capable of attacking from numerous directions at once – effectively making it impossible for the adversary to defend itself.
For these operational concepts to work – including the Navy’s Distributed Maritime Operations (DMO) concept, and the Air Force’s Agile Combat Employment (ACE) concept – a large number of attritable unmanned vehicles are needed. Scaling these operations through remotely piloted, unmanned aircraft alone raises significant personnel demands, since each platform typically requires dedicated operators.
For operational concepts such as DMO and ACE to succeed, autonomous solutions are required. This is why there is such an increased focus on autonomy.
Unfortunately, many of the autonomous UAVs and other unmanned vehicles being championed by the military’s industry partners today fall short of true autonomy. These autonomous UAVs don’t truly think for themselves. Rather, they execute a series of actions preprogrammed before the mission.
In this system, a UAV or a swarm of UAVs will fly to established way points that are predetermined prior to the mission. When reaching those waypoints, they then execute a task, which could involve generating ISR video or imagery, or deploying effectors. For the mission to be changed, a human in the loop must manually update the waypoints and directions during the mission.
There are two main problems with this type of autonomy. The first is the need for humans in the loop. As we’ve already explained, there simply aren’t enough warfighters to pilot the unmanned systems needed for the more distributed military missions of the future.
But the second issue is even more important to mission success – these UAVs aren’t truly thinking for themselves, just following orders. True autonomy involves unmanned systems that can sense their environment and situation and then make the necessary changes to improve likelihood of mission success.
For example, let’s say a UAV or a swarm of UAVs encounters anti-aircraft defenses between its point of origin and its destination. Preprogrammed UAVs will continue to follow their assigned path, even as UAVs in the swarm are shot down. UAVs that are truly autonomous will detect that they’re being targeted by an adversary’s anti-aircraft defenses and take the necessary evasive maneuvers to accomplish the mission, even without a human in the loop to change their programming.
In that scenario, true autonomy is the difference between mission success and mission failure, including the loss of an entire swarm of UAVs. However, it’s not enough for UAVs to be autonomous; they must also work together in ways that make sense for the mission.
Flock of Geese vs Swarm of Bees
As we’ve explained, the future of warfighting involves the use of proliferated, distributed forces. But for proliferated unmanned systems to accomplish the mission effectively, they must be able to collaborate while remaining autonomous individuals.
Many of the autonomous UAVs and drone swarm solutions being championed by the DoW’s industry partners are touted as collaborative. However, their concept of collaboration leaves something to be desired. In many instances, these collaborative drone swarms fly in formation, like a flock of ducks or geese. There is a leader at the front of the pack guiding their mission, and the pack is in constant communication, taking direction from that leader.
However, there are a number of problems with that approach. The largest and most glaring of which is the question of what happens should something happen to the leader. In many drone swarms, the entire pack can be lost if the leader is compromised, resulting in mission failure and the loss of the UAVs. This also creates a challenge when the military wants to fly heterogeneous drone swarms – a mix of different UAVs with different tasks and payloads.
For mission success, it’s more effective for drone swarms and packs of UAVs to operate more like a swarm of bees than a flock of geese. Instead of flying in formation, constantly communicating, and moving as a predictable unit, they should all operate selfishly and independently – each attempting to accomplish the mission individually and collaborating only when necessary to improve the chances of mission success.
For example, imagine a scenario in which a swarm of autonomous UAVs is tasked with surveilling a convoy of vehicles. With true autonomy, that swarm of UAVs can detect and track the vehicles, dynamically changing their path to follow the convoy by identifying and pursuing objects on the ground. However, should the convoy stop and individuals inside scatter, a collaborative autonomous swarm of UAVs could process the situation, communicate amongst themselves, and split up in a self-organizing way, enabling them to continue the mission.
In this scenario, a swarm of UAVs was deployed to an area where it was expected to operate autonomously as a group. However, when the situation changed and threatened to affect each UAV’s ability to accomplish its mission, they communicated and collaborated to better achieve mission success. Also, since each UAV in the swarm operated both individually and collaboratively as part of a team, there was no designated leader. This provides mission assurance and resiliency by eliminating a single source of failure.
This approach to autonomy is often referred to as decentralized collaborative edge-resident autonomy fully functioning on the native drone edge computer HW without the need for any external compute resources, e.g., cloud connectivity and persistent communication. And, while it may seem like a futuristic concept, recent military exercises show that it’s possible today.
4ID Exercises Show What’s Possible
The Fourth Infantry Division – the Army division often at the forefront of testing and evaluating new technologies – recently held its “Ivy Mass” and “PCC6” multi-domain exercises intended to demonstrate and evaluate different approaches to UAV autonomy and collaboration in a battlefield environment. These exercises were conducted in a “Red Team / Blue Team” scenario, with heterogeneous, autonomous drone swarms comprised of munitions and ISR drones used in “hunter-killer” exercises.
Each drone swarm comprised multiple ISR and loitering munitions drones. The ISR drones were tasked with patrolling an area, identifying targets of interest, and surveilling them. These drones sent information back to the soldiers operating each swarm, who used this information to designate an object as a “high value target.” The loitering munitions drones then autonomously located and tracked the target, but final kinetic decision authority remained with the human operator. The ISR drones then evaluated the damage to determine if the target was neutralized.
Palladyne AI’s SwarmOS decentralized collaborative embodied autonomy software fully autonomously operating on the drone edge computer hardware and Gremlin-X reusable mini-bomber were among those evaluated at Ivy Mass and PCC6. These solutions demonstrated the autonomous tasking, resilient operations, and autonomous tactical strike capabilities required for the exercise. But most importantly, we confirmed in operationally realistic exercises that SwarmOS-enabled “hunter” drones and Gremlin-X mini-bomber drones were incredibly effective and easy to use, even with minimal operator training.
If the ongoing war in Ukraine has taught the DoW anything, it’s that collaborative, autonomous drone swarms can be a force multiplier and difference maker – even against larger, more sophisticated adversaries. However, for the U.S. to continue taking steps toward the adoption and use of autonomous UAVs on the battlefield, the solutions it evaluates and deploys need to work within the operational approaches and constructs of the future.
More distributed and decentralized operations require the proliferation of UAVs and other autonomous systems. The threat from peer and near-peer adversaries requires that these systems be able to collaborate and operate in the most austere and contested environments. This is why decentralized, collaborative autonomy is the future – enabling heterogeneous swarms of UAVs to operate independently or together to accomplish the most important military missions.
The author, Dr. Denis Garagic, is Chief Technology Officer at Palladyne AI.