Something about the way the brown pelicans flew in formation over the waves transfixed Vince Castelli. He was on vacation at the Outer Banks of North Carolina, but the movement of the birds took him back to discussions he and his engineering colleagues at the Naval Surface Warfare Center in West Bethesda, Maryland, had had in a brainstorming session. The team had hatched the idea of small, inexpensive, unmanned aircraft that would fly in formation, cooperatively conducting surveillance, reconnaissance, and other operations against an enemy. It was a vague notion at the time, yielding no concrete design plans.
Then Castelli saw the pelicans. Every so often one of the birds spotted a fish and dove into the water to grab it. Within seconds, the rest of the flock closed the gap left by the missing member. When the fishing pelican rejoined the group, it simply inserted itself at the end of the line.
“I realized that was the perfect model for our aircraft,” Castelli says. “If one (plane) got lost, because maybe the enemy jammed its communications, the remaining planes (would) compensate.” A flock of miniature planes, as imagined by Castelli and his team, would also be able to alter plans in mid-flight based on shared information, intelligently responding to unexpected events. “Like the pelicans,” Castelli says, “all members of the group (would have) to be able to assume any role depending on what occurred during the mission.”
Now, two years later, Castelli’s inspiration on the beach-christened SWARM, for Smart Warfighting Array of Reconfigurable Modules-has become an ambitious Navy project, a UAV (unmanned aerial vehicle) concept on the fast track toward a 2004 production date. Each “module” is a 4-foot-long, 20-pound plane with a 4-foot wingspan and a 4-pound payload-enough capacity for any type of camera, microphones for eavesdropping, mini chemical and biological detectors, or even, potentially, a small weapon. SWARM planes are expected to cost just $2,000 apiece. Compare that to the UAV that was much-celebrated during the Afghan conflict, the Predator. Armed with Hellfire missiles, the Predator is a 27-foot machine with a price tag of $3 million and a payload of more than 700 pounds. Castelli’s SWARM is an entirely different class of UAV.
But size isn’t the only difference. A remote pilot, two sensor operators, support staff, along with a ground control station carried by a C-130 transport plane, are required to operate a Predator. A group of SWARM planes will be managed by a single person using equipment as commonplace as a PDA.
A handful of SWARM prototypes are already being tested. The plan is for the Navy to start launching SWARM planes from ships in September to conduct advanced patrols or gather intelligence. Ultimately the equipment could be adopted by other government agencies or even private industry for a variety of purposes: weather research, traffic control, monitoring national borders for illegal immigrants, scouting forest fires, spotting stranded boaters or hikers. “At $2,000 apiece, the commercial applications are much bigger than the military applications,” says Tony Mulligan, CEO of Advanced Ceramics Research, which has a contract to manufacture SWARM planes based on Navy designs.
But SWARM isn’t going anywhere until artificial intelligence, the technology of making machines think, evolves a lot further. In the Castelli plan, SWARM fleets will be given nothing but a destination and a mission, then will be expected to identify the best route, the most effective formation, the objects and activities to observe and image, and the information to send back to base. That’s a daunting task for a single machine, not to mention a cluster of machines that must communicate with, react to, and continually update one another.
The most significant AI experimentation under way that could advance the SWARM project is the Autonomous Intelligent Network and Systems (AINS) initiative at the Office of Naval Research. AINS is tackling the challenges of implementing autonomous behavior using distributed computing (hooking computers together so that they work as one) over wireless networks. The goal: thousands of intelligent robotic drones-submarines, satellites, ground vehicles, ships, helicopters, and planes-under the control of a single human commander who may be thousands of miles away. “What we need to figure out is when you deploy different robots with different capabilities and different sensors, how do they come up with team behavior?” says AINS program chief Allen Moshfegh.
Earlier this year, AINS conducted a war game to test the level of intelligence a network of autonomous vehicles could reach. Two teams, each composed of ground vehicles and helicopters, were instructed to pursue each other across the California countryside. “The ultimate aim was for a team to identify the enemy and trap it,” says Moshfegh.
This fall, SWARM prototypes are expected to demonstrate their first bit of intelligent behavior: autonomous navigation. When instructed to travel from Point A to Point B, the planes are supposed to figure out how to group themselves into a pack in order to reach their destination. Two possible strategies are being studied to achieve this capability, says Castelli. In one, a plane declares itself the leader and dictates through messages where each of the other planes should go; in another, this leadership responsibility would be distributed among the various SWARM vehicles, with each plane adjusting its position based on the flying pattern of its neighbors.
But this limited set of capabilities is far from what’s needed for the complex missions SWARM might accomplish. For example, imagine that a group of SWARM planes is patrolling the no-fly zone in Iraq. One is shot down by a hidden missile launcher. A second plane, recognizing the absence, searches the area and finds a suspicious building. A third spots the hot afterglow of the missile launch with its infrared camera, confirming the sighting. Next, the two planes transmit their data to a fourth SWARM plane equipped with a transceiver, which broadcasts all images to central command for analysis. If officers agree that the planes have identified an enemy surface-to-air missile launcher, they can instruct a fifth SWARM plane to “paint” the target with laser light that will guide the smart bombs sent to attack it.
To enable SWARM vehicles to communicate in such complex ways, much secure data must be transmitted. Castelli’s team plans to equip the planes with transceivers that can send data at high bit rates via the Iridium satellite network. Another option: a line-of-sight communications link being developed by the Office of Naval Research, in which a radio transceiver sends encrypted data over a 10- to 20-mile range.
A SWARM plane is designed to be launched from a catapult or, possibly, a helicopter and to operate at 60 knots, meaning it can be employed in most weather. The vehicles don’t need to return home-from the military’s perspective, $2,000 equals disposable-so their wingspan is short. “If I had to land it, the wingspan would be three times longer,” says Dave Lacey, an aeronautical engineer on Castelli’s team. The major inflight challenge is stability, since small planes with simple avionics are susceptible to rough weather. To compensate, the plane’s dihedral wings tilt up at the tips to keep the aircraft from rolling.
The plane’s engine will be made of ceramic and plastic, lightweight materials that are preferable to metal because they are cheaper, they can’t be easily detected by radar, and they can withstand higher temperatures, enabling the engine to burn fuel more efficiently. That’s important because the miniature planes have a target range of 1,500 miles but gas tanks that hold just 1.5 gallons of diesel. But the SWARM engine design has some drawbacks as well. For one, diesel engines are loud, so the SWARM prototypes are anything but covert (Castelli hopes to fix this flaw with noise reducers). More important, the two-stroke engine stalls in the thin atmosphere of higher altitudes, so the SWARM devices will only be able to fly at heights of between 500 and 8,000 feet.
Castelli is convinced that for SWARM devices to be widely used in military missions, large and small, the machines must be disposable so that losing a few will not be a concern. He chose the target price of $2,000 to match the cost of sonar buoys that the Navy uses and frequently discards. He’s earmarked $400 for avionics, $200 for the engine, and $900 for communications, with the remaining amount allocated to the generator, airframe, and the command module. So far, though, at prototype stage, it’s been impossible to get production costs lower than $16,000.
Starting in 2004, 10,000 SWARM planes will be produced annually, Castelli says. The first tests of the SWARM system are under way in the Pacific, where the Navy is using prototype planes to search for whales and other marine mammals before testing munitions, high-powered low-frequency sonar, or missile defense interceptors-systems that can disrupt the animals’ ability to communicate, feed, and navigate.
Next up for the SWARM team, after formation flying is perfected, is to bring the cost of the vehicles closer to the $2,000 target by redesigning the avionics so that the entire system can be shrunk and integrated into a chip. After that, Castelli says he wants to focus on creating a network of sensors for the planes that would combine digital audio and video files from the entire SWARM fleet and send them back to base. Such a system would provide more detailed remote information than anything the military has today. All in all, not a bad legacy for a flock of pelicans.