Between September 6 and October 3, 2016, the US Navy's science and technology wing sent four small boats on patrol missions in the lower Chesapeake Bay. The only unusual thing was the fact that the boats were autonomous robots.
These robots are not like drones with humans operating them from a remote trailer in Arizona. The vessels, loaded with cameras and sensors, were more like autonomous individuals within a larger robotic collective. The Office of Naval Research, which put on the demonstration, had not done this before.
Well, not exactly. In 2014, the ONR sent 13 boats into the James River in Virginia and commanded them to "swarm" around vessels in different scenarios. The boats escorted "friendly" vessels or simulated attacks on potential "enemies."
But the swarm was also kind of dumb. Where one boat went, the others went too. And at the same time, each boat had to individually determine how it would get there.
Two years later in the Chesapeake, the ONR sent fewer boats, albeit smarterones that work better as a team.
The team of Navy rigid hull inflatable boats, or RHIBs, patrolled an area while watching out for incoming vessels, and when spotting one, worked together with a shared software package called CARACaS to assign a singleboat to approach it, classify the type of vessel it was looking at, and then transmit that information to the rest of the swarm.
At one point during the demonstration, two "enemy" vessels approached the swarm, one after the other, and the swarm dispatched a single boat, and then a second, to follow each one. The remaining two boats held back.
The researchers tilted the scales slightly by programming the boats in advance with images of "enemy" vessels, but the fact the little flotilla could spot, identify, and then track ships with individual robots is noteworthy. For one, it means it can go after multiple threats instead of just one at a time.
And instead of simply escorting or attacking—the only two previous options in the 2014 demonstration—the swarm can now patrol an area on its own in addition to trailing an enemy, classifying it as good or bad, or tracking it with sensors.
In theory, the ONR could order a boat to automatically attack a vessel identified as an enemy, but these boats did not include that particular mission as an option—this time.
The only humans involved were researchers sitting at remote stations with low-bandwidth connections. They could have intervened if necessary, but the boats did their own thing. The ONR even provided a video showing the demonstration in action:
As to why this is significant, look at how automation is not just making inroads within the military. In the 20th century, the assembly line transformed the economy, mechanized transportation and created a revolution in warfare.
In the 21st century, automation is laying off assembly line workers, and turning vehicles into robots with enormous consequences for major industries. Now apply that to a battlefield—or a naval engagement.
"We envision unmanned boats doing protection missions, escort missions, we envision them doing intelligence, surveillance and reconnaissance missions," Cmdr. Luis Molina of the ONR told reporters on December 14. "Really, helping our ships and our sailors getting out and expanding their sphere of influence and being able to control larger areas of the ocean and determine what's going on."
ONR has been working on CARACaS for around 13 years, but in a span of a mere two years, went from having a robotic boat swarm that does one thing at a time, to a swarm capable of tracking and intercepting multiple targets.
That's fast. If you gave the boats some machine guns—not included on the ONR's demonstration boats—and a human weapons operator sitting at a remote terminal, you'd have a pretty effective robotic harbor defense force. Give the boats torpedoes and have them go on the attack, and how could a navy possibly defend itself? Not well.
"We have every intention of using those unmanned systems to enable our sailors to engage a threat, and destroy it if necessary," the ONR's then-chief Rear Adm. Matthew Klunder said after the James River test two years ago. "But let me anchor on one really important factor here that's critical to us. There is always a human in the loop of designation of a target, and if so be it, destruction of the target."
Now fast forward another 20 or 30 years. The same technology could be combined with larger ships such as the Sea Hunter, a 132-foot-long robotic trimaran being developed by the Defense Advanced Research Projects Agency to hunt submarines—which, by the way, has components shared with CARACaS.
This barely touches on the advances in aerial drones, which could one day communicate with unmanned vessels —all autonomously—and combine their sensors into a powerful package.
One problem is that doing so requires a lot of bandwidth for all the data packets flying back and forth. A swarm of boats at sea will need a powerful network. Molina cautioned that the swarm's network was rudimentary on purpose, and just enough to complete this particular demonstration.
Make the swarm bigger, and assign it more missions, and you'll need a lot more bandwidth.