A new drone swarm piloting program from L3Harris, released today, allows a single operator to control multiple drones across several vehicle types in different domains during government-managed tests—and company officials told reporters they expect to eventually be able to scale that up to thousands. The development comes as the Pentagon struggles with how it could manage swarms of drones in a conflict rife with electromagnetic attacks on communications.
To achieve the desired scale, the company uses an approach that minimizes the data exchange between the operator and the swarm, putting more burden on the intelligence capabilities onboard the drone.
Toby Magsig, L3Harris’s vice president and general manager for enterprise autonomous solutions, said the approach used in Amorphous, where highly autonomous drones figure out what they’re doing together and collaborate with little input from the operator, is based on feedback the company received from the Army, Navy, and Defense Secretary’s office.
That feedback emphasized two things: the military need for one operator to control large numbers of drones simultaneously without the drones needing to analyze and interpret lots of data. The end result is a software architecture where the data exchange between the drones and the operator is scaled down to roughly the size of an SMS text message. That means the swarm isn’t dependent on a large enterprise cloud somewhere far away, and can operate closer to the battlefield without long communication threads.
“We’re looking for very thin messages that don’t require intense bandwidth and don’t consume an intense amount of processing power for those small, low-cost assets to be able to consume and inter operate,” Magsig said.
The drones must figure out how to carry out different aspects of the mission with limited instructions and minimal talk. In effect, he said, they “bid” on different operations or behaviors based on where they are, what they can do, and what they know about the rest of the swarm. That all happens in a way the operator can see, but without him or her having to give lots of specific instructions.
“For example, you designate a search area and you say, ‘Hey, search this area.’ You draw a polygon with basic points. They receive that. They say, ‘Hey, we just received the command to search. I’m over here. I’ll search this area.’ The other one receives a message and says, ‘Okay, I’ll search. I’m right here. I’ll do this.’ And then you iterate through this process.’”
That approach cuts down on the number of tasks the drones have to figure out how to do, he said, saving them processing power for other tasks.
“You’re able to integrate this software onto any vendor’s hardware. And we’ve done that integration improvement time and time again. Some take longer than others, but it’s absolutely an integration,” Magsig said.
Even that very low bandwidth level of communication still provides data back to the operator, he said: “So things like swarm health that allows you to monitor the quality and become the feedback loop to that smart swarm technology, things like predictive analysis. So when you task a swarm, you can see the behaviors that they’re about to do, so you feel comfortable as a user [in terms of] what you’ve asked that swarm to do, and then things like real-time ability to shift and adjust the tasks based on the emerging mission requirements.”
L3Harris isn’t the only company racing to build drones and deliver an open-architecture software platform to steer hundreds of them. Last November, the Pentagon’s Defense Innovation Unit announced it had awarded prototype contracts to three companies: Anduril, L3Harris, and Swarm Aero. The contracts are part of a program called Autonomous Collaborative Teaming, or ACT, and it’s much more important than any specific drone purchase because individual drone makers would need to make sure their platforms worked with that swarm piloting software, DIU’s deputy director for strategy, policy, and national security partnerships Aditi Kumar said at the time.
“We are buying this capability independently of the hardware systems, and so we need to be able to have open architectures, government-owned architectures, to ensure that the software that we’re bringing in is one being upgraded and then integrated into all manner of hardware systems,” she said.
The approach Amorphous uses is a change from the level of operator control that characterized drone operation during U.S. operations in the Middle East—when analysts would sit and patiently watch hundreds of hours of live drone feed coverage.
While there may still be certain instances where a drone operating in what’s called a “disconnected, degraded, intermittent, and low-bandwidth environment” is able to send back video footage, that won’t be typical. It also won’t be practical in a situation where one person is tasked with operating thousands of drones across land, sea and air.
What the means for autonomy and future human control is, “You really have to move from ‘in the loop’ to ‘on the loop,’” Magsig said, meaning less human control. “That’s the only way you’re going to get to that level of scale.”
But the most important aspect of control, the ability to stop automated processes that don’t follow commanders’ intent, is still there, designed into the system as an emergency stop function. “So if ever a user is not comfortable with what the behaviors that the swarm has exhibited, the human is always in control, simply hit that button,” he said.
