Looking out over a hilly landscape, parades of puffy clouds march across the California horizon. The five-year, statewide drought that had burned the surrounding countryside a golden brown has finally broken with a punishing, record rainfall. Now the hills are green again on the Central Coast, and a window of calm weather allows for a small group of scientists, researchers and academics to do something quite rare in the realm of unmanned systems and autonomy.
The site is Camp Roberts, a California National Guard base in south Monterey County that is also home to one of the most unique experimentation areas in the region, boasting 42,000 acres of diverse terrain and specialized ranges including the restricted air space that is critical for research in the skies. It was here that scientists from two universities have converged onto the site, February 9, 2017, to duke it out in a never-before-achieved swarm vs. swarm unmanned aerial vehicle dogfight.
For Naval Postgraduate School (NPS) Associate Professor Ray Buettner, it feels like a historic occasion.
“It is always difficult to know the impact of an event on the future. When the Wright brothers flew for the first time at Kitty Hawk it is unlikely that they or the folks watching could imagine that manned flight would take humanity to the moon only 66 years later. As we watch the first ‘dogfight’ between swarms animated by different forms of artificial intelligence, I am compelled to wonder where swarming technology will lead us six decades from now.”
The match is between two university teams that are well aware of the each other’s reputation in the field, the Naval Postgraduate School (NPS) and Georgia Tech Research Institute (GTRI), but this contest isn’t about bragging rights. Rather, this experimentation is to test the capability of autonomy and to learn how behaviors between each of their team’s aircraft react as they penetrate the opponent’s air space.
In short, the goal of this research, according to researchers and institutional leadership alike, is in the science, to answer questions in optimization and controls, and to envision how these capabilities advance the warfighter’s effectiveness.
“One of the only ways to find out if the ideas we have built in labs and in experimental test beds actually work, is to bring them outside into something that closely approximates how they might perform in an actual warfighting setting,” explained NPS Provost and Academic Dean Dr. Steven Lerman, on hand to witness the experimentation first-hand. “This is an outstanding example of the translation of great research into experimental practice.”
Officials say the collaboration between NPS, GTRI, and the Defense Advanced Research Projects Agency (DARPA) also serves as an incubator for ideas and a catalyst for innovation.
“This is a great opportunity to grow and see what happens outside the simulated environment,” said GTRI Senior Research Engineer Kevin DeMarco. “We sort of wall ourselves off in academia without having access to people with real field experience. The NPS faculty and research team has that real-world experience.”
DeMarco recently earned his doctorate in Electrical and Computer Engineering from Georgia Tech and has been conducting research with human-robot interaction scenarios for some time. DeMarco and other researchers in the GTRI lab have been improving sensors and swarming scenarios with its own flight of aircraft. In May of 2016, GTRI launched a successful, 30-drone swarm with a catapult system of 10 launchers that use compressed air and can load and launch in one-second intervals.
DeMarco’s team uses Base Theoretical Swarm Controls, which sort data between different altitudes, non-altitude separated flight and collision avoidance algorithms in their programming.
An unmanned aerial vehicle takes to the skies in the Naval Postgraduate School’s first-ever swarm on swarm aerial dogfight at Camp Roberts, Calif. The research, supported through the university’s Joint Interagency Field Experimentation collaborative research program, advances NPS leadership in unmanned systems behaviors and capabilities with a keen eye on supporting the future warfighter.
“With today’s experiment we have myself, Senior Research Scientist Charles Pippin, Research Engineer David Jensen and Research Associate Michael Day. And we also have three support staff back in Georgia at the GTRI lab,” said DeMarco. “We can e-mail code that we’re preparing here on the fly, have the support staff verify it using GTRI’s high performance computing and they send it back to us to upload into play.
“What’s really interesting about this experiment is that it’s the first intercollegiate competition in the swarm versus swarm arena and its really informing us how to compete in larger, more complex competitions,” he added.
NPS’ Advanced Robotic Systems Engineering Laboratory (ARSENL) has employed a commercial off the shelf platform, the Zephyr, since 2012 for its swarm research. The Zephyr is a fixed-wing design made of very light foam and is easy to hand launch or catapult. ARSENL researchers have also outfitted their aircraft with only commercial, off-the-shelf components, and the software used in the brains of this aircraft is also open-source and open architecture.
The ARSENL lab, made up of faculty, students and research associates, uses various networking technologies to achieve the types of autonomy needed for successful swarm operations. The lab utilizes its own unique launch system developed by an NPS student, a chain-driven catapult that pushes the aircraft out at 35 mph. Once aloft and clear of the launcher, rotation begins immediately with the small electric motor that revs up the propeller into a high pitch whir and zips the UAV into a steep climb.
DARPA Program Manager Tim Chung, a former NPS assistant professor who played a leadership role in establishing the ARSENL lab, describes how the drones conduct their engagement on each other.
“The battle cube is a vertical column of air space with a demarcation, separating each team’s swarm. The UAVs swarm and self-organize into their patterns, they select a leader based on an aircraft’s best position to engage their opponent. Each drone continually assesses its place in a tiered hierarchy so as to benefit the goal,” Chung explains. “Think of The Borg from Star Trek’s Next Generation, the collective or hive minded drones that are driven by a need for perfection.”
Chung continues to describe attributes drones will take on as they learn to defeat opponents through evasion and counter-measures in the decentralized, ad hoc doctrine.
“The aircraft, either singly or in groups, choreograph their next move into a defensive or offensive pattern. What’s not being seen are the huge amounts of data being sent over a Wi-Fi network that provides each of the team’s aircraft a three-dimensional battle space picture. This simultaneous sharing of information between the drones induces the concept of mass, coordination, intelligence and speed to the battlefield,” described Chung.
One observer who cranes his neck upward as the swarms whiz by overhead in formation is retired Marine Corps Brig. Gen. Frank Kelley, who now serves as the first-ever deputy assistant secretary of the Navy for Unmanned Systems. The position puts Kelley in charge of creating and implementing a strategy for development of unmanned warfare systems and the technology that links all domains. Kelley has only been in the job a little over a year, but is very optimistic about autonomy’s future.
“I’ll tell you, my expectations have been fulfilled with coming out to Naval Postgraduate School and to Camp Roberts, to see the hardware, to see the experiment being conducted and to meet each of the researchers from NPS, DARPA and GTRI,” said Kelley.
“Swarming technology is incredibly important. Remember, we don’t accomplish any mission with just one vehicle, one aircraft, or one Sailor or one Marine. There are teams of Marines, Sailors and platforms working together in a combined arms approach. We’re going to see that combined arms in unmanned air vehicle formations. But there’s no way that we shouldn’t be able to look into the future and see how that’s going to apply across all platforms,” added Kelley.
“The technology that has been developed in our universities, our warfare centers, and institutions like NPS and DARPA now makes those concepts available. And the folks that are thinking how we’re going to operate in the future, realize that there’s a huge part to be played by autonomous systems,” said Kelley. “I can’t think of a domain where that isn’t going to work in the future – underwater, on the surface, in the air and quite possibly in space as well.”
Seeing the successful experiment has Kelley imagining the possibilities of the future of unmanned systems. “I’m absolutely confident that this will be able to scale … I’m also excited with the fact that we’ll be able to go back and talk to our leadership and say to them we’re on a positive trajectory for getting unmanned systems to work in a swarming concept of operations,” he said.
A swarm of unmanned aerial vehicles get into a self-determined formation during swarm vs. swarm field experimentation at Camp Roberts, Calif. The match between researchers at NPS and Georgia Tech Research Institute is considered the first-ever swarm on swarm autonomous drone dogfight.
“I honestly believe there are no technical hurdles that we can’t overcome. We need to think about the cultural and social aspect of the man-machine interface and tap into the potential of both,” exclaimed Kelley.
GTRI Research Scientist in robotics Michael Day, also a former research associate at NPS, continues his experimentation with swarm-based sensors and logistics in his new lab. His team has been collaborating with NPS to resolve issues with eliminating radio noise over their network communications.
“The benefit of our collaboration with NPS is that we come from different worlds. NPS is directly connected to the military. They have access to Soldiers who have been in theater recently, so they know what the existing threats are and what research needs to be done to counter them,” Day explained.
“GTRI gets a lot of insight into what is needed in the DOD and we get another point of view. We both benefit with the exchange of points of view, we both look at algorithms in slightly different ways, and we can swap ideas,” Day continued. “We also have little friendly competitions. Usually when you put game or competition on something, you add motivation for your team to try to do a little better than the other side and it gives synergy to the whole thing.”
And speaking of friendly competition, researchers take a politically-correct approach to declaring win or lose. Buettner says that NPS won one match, tied one match and lost one match. “What I would say is that both NPS and GTRI won in terms of the science,” he added.
“The simple fact is that swarms flown by each team executed their behaviors perfectly in that they generated the data that we needed to collect. The hard work of data analysis is just beginning,” continued Buettner. “The real winners of the competition are the citizens of the United States, as the U.S. Navy remains at the forefront of exploring these rapidly developing technologies.
Dr. Kevin Jones, who coordinated the launch and recovery of the NPS team in addition to being the lead aeronautical engineer on the ARSENL team, said one of most significant lessons learned from the event was in the ability to plug and play behaviors in a live environment.
“NPS, and our research partners, demonstrated the ability to allow any third party to develop swarm behaviors and to implement them in a (compatible) software stack, permitting them to be loaded into a swarm that can engage in combat behaviors,” Jones explained. “Going back to the original Swarm vs. Swarm Aerial Combat Challenge objective, all of the pieces exist to support the execution of a national grand challenge event that would eventually enable us to rapidly identify and transition swarm tactics and maneuvers from the NPS environment to more operational environments.”
As the window of optimal, rain-free weather begins to close, teams are already thinking about their next steps in this rapidly-developing field. With researchers at NPS already accomplishing what was then a record, single swarm of 50 individual aircraft in August 2015, Jones says there is no significant technical difference between what the team accomplished today with swarms of 10, and what would be needed to fly a 50 vs. 50 challenge.
They’ll save that battle, then, for another day.