Troops in Iraq are calling out for unmanned vehicles, but even more important, they are in desperate need of unmanned ground systems (UGSs) to dispose of land mines and booby traps.
Pentagon planners have responded by sending 163 UGS vehicles to Iraq, involving five different robot models. The vehicles range from 50 to 150 pounds, drive on caterpillar treads by remote control, and use a robotic arm to disable bombs. They are nearly identical to the vehicles that police bomb squads or homeland security agents use to investigate stray bags left in airport terminals.
Pentagon planners envision a future when such robotic vehicles will perform other dangerous tasks, from guarding base perimeters to conducting reconnaissance and scout missions behind hostile lines.
Despite the immediate advantages of robotic ground vehicles, systems designers say the vehicles will never advance to more-complex missions until they get over technology hurdles such as communications range and latency, and until sensor manufacturers start building robot-compatible electronics.
Robots report for duty
Pentagon officials had been planning a next-generation bomb-squad robot before the Iraq war began, yet the man-transportable robot system (MTRS) contract was still two or three years away when U.S. soldiers began asking for unmanned ground vehicles to combat deadly Iraqi mines.
So officials asked robot suppliers to send everything they could spare, and by April, they had shipped nearly 160 robots to Iraq and Afghanistan, says Cliff Hudson, coordinator of the Joint Robotics Program of the U.S. Department of Defense.
"The bad news is we're losing the units in the field. And the good news is we're losing the units in the field. Every one lost is a soldier's life saved," Hudson says. He made his comments in April at a press conference in Washington, D.C.
He plans to narrow the five robot models to one or two by August. The finalists include vehicles from: EOD Performance in Ottawa; Foster-Miller of Waltham, Mass.; iRobot of Burlington, Mass.; Mesa Associates of Madison, Ala.; and Remotec in Oak Ridge, Tenn.
Then Hudson will return to his long-term goals — building new robotic-technology standards such as common driver controls, common software architecture, and a common sensor interface.
Once robots have attained reasonable autonomy, they will shoulder more responsibility on the battlefield, he says. As parts of the Army's Future Combat System, they could do tasks like combat engineering; clearing mines; emplacing charges; breaching urban walls; screening for chemical, biological, and nuclear contaminants; direct-firing weapons; obscuring battlefields; and performing reconnaissance, surveillance, physical security, and force protection.
Each of those jobs needs vastly different tools, so Hudson's office monitors robots in seven different weight classes from eight pounds to 42 tons. For more information, see www.jointrobotics.com.
First, disable the bomb
Whether the user is local police, homeland security agents, or troops overseas, most UGS missions
today involve explosive-ordinance disposal (EOD), says Arnis Mangolds, vice president of Foster-Miller in Waltham, Mass.
Engineers at Foster-Miller make the Talon, which is one of the five models on the ground in Iraq. They have built 45 vehicles for the job. They are also finalists — along with iRobot — for the MTRS contract.
EOD is a simple mission compared to more autonomous tasks like perimeter security, scouting and reconnaissance, or direct action with firearms, Mangolds says. Designers have fit more than 80 different sensor types on the Talon, but bomb squads typically use just a dozen common types.
The most important ones are illumination, cameras (infrared, night vision, and standard), water cannon (to blow apart the trigger from an explosive charge), and sometimes two-way speakers or explosive sensors.
One reason users do not attach more sensors is the lack of a common interface, such as the ubiquitous USB port found in office PCs.
Pentagon planners want a common interface, which is one reason that the next-generation Talon will upgrade its 16-bit processor to an Intel Pentium. The more powerful chip would enable plug-and-play operation with various payloads. That makes it worthwhile to use the Pentium despite drawbacks such as its slow start-up time, high cost, sensitivity to heat, and increased power draw, he says.
"Today if they need extra illumination, soldiers will take a flashlight and duct-tape it on there. In the future, they would attach it to the processor and control it remotely. But no one makes flashlights with USB ports yet," Mangolds says.
That is not going to change any time soon, because the robot business is a small fraction of sensor companies' business, he says. Compatibility with robots is a low priority for them.
In another nod to Pentagon demands for common platforms, the new Talon will use software compliant with the Joint Architecture for Unmanned Systems (JAUS) protocol. That common language will enable robots to talk to each other, whether a UGS reports its findings to a hovering unmanned aerial vehicle (UAV), or warns another UGS about a nearby bomb. For more information, see www.jauswg.org.
Pentagon leaders also want a common operator-control unit, or OCU, on future robots. That would enable a single soldier in the field to use a single laptop-size interface to steer entire fleets of land, air, and water vehicles.
"But it's a little too early," Mangolds says. "People are talking about it, but it hasn't gained momentum like the software protocol has."
Weak communication links
The biggest technological limit on robot performance today is communications, he says. "We drive at miles per hour;
using a 750-watt-hour lithium battery on board, we could go 20 or 30 miles, depending on the terrain. But wireless communications range is 800 meters. So that's the holdback; that's the technological limit."
Theoretically, users could extend the range of wireless communications links to almost a mile by using high-gain antennas, or relaying the signal off a hovering UAV. But it is hard enough for a soldier to get the robot into the field, much less ask for an additional UAV, he says. Forests or buildings will block a wireless signal anyway.
Another trick to extend the range is to use a fiber-optic tether, which offers high bandwidth and good security without requiring line of sight. In tests, Foster-Miller engineers have driven a Talon for about four miles on a fiber-optic cable. They can extend the trip for another 500 yards by instructing the Talon to drop its empty cable spool, which contains a wireless receiver.
Latency is another communications problem, even when the vehicles stay within range, Mangolds says. The problem affects video feedback more than control signals. Latent video can make it nearly impossible for the driver to use the robot's manipulator arm to nudge an object half an inch over, or target someone running across a field.
"Say you're driving at five miles per hour, which is eight feet per second. If you have a 0.1-second delay, or more commonly, a two-to-three-second delay, that's a lot of distance, especially if there are bombs around."
Talon vehicles try to work around the problem by using digital and analog cameras, he says. While digital signals provide the best imagery, they cut out suddenly, as soon as the robot gets out of range. An analog camera signal will deteriorate gradually, giving the driver warning before his machine goes blind.
Simple robots for simple missions
In terms of navigation and control systems, unmanned vehicles are surprisingly simple. Users always ask the company to make them lighter and faster, but not necessarily smarter, Mangolds says.
"Yes, there's a desire to make them faster. But the robots do not necessarily have to be smarter to do it; that was a revelation to us," he says. "Our original systems were more intelligent than the ones we use now. Everybody's talking about more autonomy for the robot, but users typically say they want more control."
The original Talons had functions such as cruise control, global positioning systems (GPS), waypoint navigation, and obstacle avoidance. As the engineers made them simpler, reliable, and affordable, however, they gradually took those systems out.
"Maybe when missions increase in length, users will need them again, but typically EOD missions are so short that autonomy's not necessary," he says. "When they start doing direct action or recon, and sending these things out a week in advance or 100 miles away, they'll need more autonomy."
In the meantime, the typical EOD mission is a line-of-sight trip of 600 to 800 meters — just far enough for the driver to disable a bomb from a safe distance.
"There is interest in weaponization, for scout or recon missions. That's coming and it is coming fast, but it is not here yet. People tend to get excited, but the warfare of the future is not here yet," Mangolds says.
Call for better sensors
Several years ago, designers at iRobot in Burlington, Mass., created the PackBot as a platform in the Defense Advanced Research Projects Agency's (DARPA) Tactical Mobile Robotics program, now terminated.
Today, that robot is a versatile, lightweight platform that has searched tunnels under the Baghdad airport, looked for Iraqi soldiers hiding in an agricultural center building, and examined a booby-trapped airfield.
The standard PackBot model is 39 pounds, and drives eight miles per hour — about the speed of a slow jog. The robot has a modular design, so soldiers can outfit it for specific missions, says Tom Ryden, iRobot director of sales and marketing for government and industrial robotics.
To build a PackBot for reconnaissance mode, engineers strip off all the sensors except an infrared camera, replace them with extra batteries, and then drive it out to spy on an airstrip for several days.
To build a PackBot for EOD, designers add a 15-pound robotic arm. They also replace its wireless controls with cable to avoid tripping sensitive explosives with RF interference. That also boosts the bandwidth, resulting in higher-resolution pictures.
In the future, designers will need better sensors to equip their robots for more advanced missions, says Ryden.
Sensors are the key to advanced robot abilities, and sensors were the weakness in most entries for DARPA's March 13 Grand Challenge, he says. The agency offered a $1 million prize to the team that built a robot that could drive the 150-mile course through the Mojave Desert in fewer than 10 hours. They didn't have to award a dime, because the best entrant made it just 7.4 miles.
To gird PackBot for more dangerous military missions, iRobot designers are now looking for small, lightweight versions of a chemical- and biological-agent detector, radiation detector, metal detector, and ground-penetrating radar.
In the meantime, company researchers are trying to boost PacBot's speed and endurance.
To build a faster PackBot, the engineers must create a driver-assist system — software that enables the vehicle to avoid obstacles automatically, Ryden says. They also need better navigation sensors like a tool that can discern whether a puddle is too deep for PackBot to wade through. To give it longer life, they are looking for better batteries. Currently, a nickel cadmium batteries work well, but new lithium batteries might be better, he says.
The final goal is to give each PackBot the autonomy to operate without human guidance. "To really be a big help in the military, the robot must be a force multiplier," Ryden says. "It can't do that if you've got one soldier driving each robot. We're trying to give the robot autonomy to operate alone, so one soldier can operate 10 robots."
Finally, company designers are trying to shave weight off the robot, which is a requirement in their contract to build the small unmanned ground vehicle (SUGV) for the U.S. Army's Future Combat System. Working with Boeing and SAIC, iRobot designers will help create a reconnaissance and tactical robot that is durable and smart enough to traverse obstacles without human assistance
Payloads that fit the job
Another robot seeing duty in Iraq is the Mini-ANDROS II, a 175-pound robot that can climb stairs or crawl along at two miles per hour, while taking its commands through its tether or wireless link.
It reaches out with a six-foot robotic arm with a 12-inch grip that can spin 360 degrees to turn a car key or door knob, says Royce Hollman, customer service manager for Remotec, a Northrop Grumman subsidiary in Oak Ridge, Tenn.
With a little practice, a good robot driver can use the arm to pick up a raw egg or a plastic foam cup of coffee, he says.
The Mini-ANDROS II can do almost any job, if it has the right payload. Military users can attach weapons, a hostage negotiator can attach a two-way speaker, or a bomb technician can attach a disrupter gun to separate detonator from explosive with a burst of high-pressure water.
"End users always want faster speed and lower weight," Hollman says. "Yes, it is heavy, but so is an 85-pound bomb suit."
In the long term, robots need better communications technology, says Remotec President Mark Barber.
Research and development must focus on long-range control signals, and better ways for robots to share data, whether they are beaming it to a commander 10 miles away or to a UAV circling above.
Once they can communicate from afar, robots will be able to do far more complex tasks, Barber says.
"They could do a robotic follow. There was a very long logistics trail to Baghdad, so why not have one manned vehicle with 10 robotic ones following it, and have fewer drivers under fire?"
UAV builders design planes for fast technology refresh
By Ben Ames
Compared to unmanned ground vehicles, unmanned aerial vehicles (UAVs) carry a wider array of sensors and can operate with far greater autonomy.
The latest generation of the Northrop Grumman Global Hawk high-altitude, high-endurance UAV relies on one button-push from its controller to taxi down the runway, take off, climb to altitude, vector to its target heading, perform its mission, and return to base, says Douglas H. Patterson, director of marketing for Vista Controls, a Curtiss-Wright company, in Santa Clarita, Calif.
The integrated package of electronics aboard Global Hawk demands different technical challenges from ground-based robots. The main challenge for UAV designers is technology refresh — or the ability to swap new electronics into the platform without taking apart the entire plane.
The current version of Global Hawk relies on a dual-redundant mission-management computer to fly the aircraft, manage other onboard flight systems, and integrate data from the sensor suite. Vista engineers build the two boards that power that AIR-based system in each vehicle.
But the next version of Global Hawk — the Spiral Four edition — will split the sensor computer from the flight computer, Patterson says. That will enable designers to change the sensor platform on a vehicle without requalifying its flight-control software, which is a slow and expensive process.
Such flexibility is particularly important for this version because Northrop Grumman engineers will build two versions of the plane — one for the U.S. Air Force (the current customer), and one for the U.S. Navy and Coast Guard.
Although the two will use different radars, cameras of different resolution, and even different printed circuit boards they will share the new split architecture — a redundant, improved mission-management computer (IMMC) alongside a single sensor-management unit (SMU) on each plane.
The new SMU will use a CompactPCI format for the Air Force version, and a split backplane CompactPCI and VME format for the Navy, Patterson says. The rest of the design is parallel; both used 6U-sized boards, conduction-cooled units that rely on Gigabit Ethernet, PICMG 2.16 for data flow, and lightweight aluminum and carbon fiber chassis.
These design choices contribute to the Global Hawk's mission as a long-term loiterer, circling at high altitude with jet engines and enormous wings. The vehicle has a short, stubby fuselage but has the same wingspan as a Boeing 737 commercial jet, Patterson says.
As opposed to General Atomics' Hellfire missile-armed, propeller-driven, low-altitude Predator, the Global Hawk acts more like a 21st century U-2 spy plane.
To extend the comparison, UAV electronics must meet many of the standards as avionics gear. In fact, from a board designer's point of view, there is not much difference between a UAV and a cruise missile, says Ian Stalker, manager of marketing communications for Dy 4, also a Curtiss-Wright company, based in Kanata, Ontario.
Engineers at Dy 4 supply air- and conduction-cooled boards for the Global Hawk and Predator. For either combination, the demands are the same: high computing density, ruggedization, and small size.
The only difference between the ruggedization necessary for UAVs and missiles revolves around staggering amount of vibration that missiles must withstand from air buffeting while attached to combat aircraft, Stalker says. Dy 4 is building a VME card for the Taurus, a European cruise missile just starting production.
The greatest design difference is a UAV's need for constant technology refresh, he says. Dy 4 designers first used the Motorola 68040 class of single-board computers for the early Predator. They have since advanced to the PowerPC 603, the PowerPC 740, and beyond.
Aside from a bottomless appetite for processing power, UAV builders have not changed their original electronics specifications much, Stalker says. The Predator is now flying with a conduction-cooled VME card using the PowerPC 740, and the Global Hawk is using CHAMP-AV quad processor cards for general computing, not signal processing.
