Friday, November 1, 2002
Minidrones: Near Term
Designed for small-unit operations, unmanned air vehicles, such as Dragon Eye, will be fielded in the near term.
Tested in Kosovo and Afghanistan, unmanned air vehicles (UAVs) are here to stay. But behemoths like Global Hawk and Predator dominate, despite their cost, complex logistics and vulnerability. Tactical UAVs exist, but the small-unit niche remains unfilled.
This situation, however, is about to change. The U.S. Marine Corps is close to fielding a 5.5-pound (2.5-kg) reconnaissance and surveillance drone for company-level operations. The Defense Advanced Research Projects Agency (DARPA) will test prototype, squad-level Micro Air Vehicles (MAVs), aimed at urban warfare, with the a U.S. Army division next year, leading to their use in a major war game. Meanwhile, the Army is looking at a downsized version of the Pointer UAV.
U.S. Navy researchers, in parallel, are developing a fault-tolerant sensor and communications network for wider-area reconnaissance and surveillance–composed of unmanned aircraft and other robotic vehicles. This "swarm" would scale from tactical to strategic dimensions (see story on opposite page).
The U.S. military’s attention, clearly, is on small UAVs, but they need to prove themselves. "The objective of seeing over the next hill, rather than into the next country, promises something of a revolution for individual units on the ground," says Richard Aboulafia, vice president of the Teal Group, an aerospace research firm. But the temptation is to wait for the next technology leap before finalizing sensor and data link requirements. Another challenge is networking. "Network-centric [warfare] isn’t here yet. Global Hawk and Predator are effectively bipolar, bilateral. You have to construct an elaborate network and then plug in the small systems."
Dragon Eye, a "backpackable" drone with of wingspan of less than 4 feet (1.2 meters), is the closest mini-UAV to production. Pending a "milestone C" decision next month, the Marine Corps plans to buy 311 systems–each comprised of three airplanes and a ground control station–over the next few years. A production contract, expected by January 2003, will go to BAI Aerosystems or AeroVironment. The first fully supported, production vehicles could be fielded in July 2003. The USMC already has employed the Dragon Eye at the company level in an urban warfare exercise at the former George AFB in California. And the Army has bought two prototype Dragon Eyes and a control station for use in Kosovo this year with the 1st Infantry Division.
Control of these UAVs will be transferable from one ground station to another. "Doing that part well will make or break the small UAV business in the military," says Maj. John Cane, UAV project officer with the Marine Corps Warfighting Lab. The vehicle might start out under battalion control, for example, but be passed down the chain to a company commander if something breaks out at that level.
The Marines wanted a "hands-off," largely autonomous system that can be operated with minimal training, says Cane. Right now the USMC has the Pioneer tactical UAV, a 15-year-old, 450-pound (204-kg) vehicle–a Marine Expeditionary Force asset. But because Pioneer has a tremendous logistics footprint, it is far from ideal for small units on the move.
Dragon Eye will be hand-launched or bungee-launched, using a 6-foot (1.83-meter) nylon string stretched out to 28 paces. The service wants a 5- to 10-kilometer (3.1- to 6.2-mile) data transmission range, 45-minute endurance, and cruising speed of 35 knots. Typical operation would be at 300 to 500 feet (91.4 to 152.4 meters).
The flight path is programmed from a 4-pound (1.8-kg) Panasonic Toughbook computer, and imagery is viewed through video goggles. Marines use drop-down menus to plot GPS waypoints on a digital map and determine flight profiles at each point. Key to the project is the 4-ounce (113-gram) autopilot developed by Lockheed Martin Skunk Works. Consuming only milliamps of power, the device permits automatic takeoff, landing and in-air operation. The autopilot also allows the use of a side-looking camera that can keep a target in view for extended dwell times.
The Dragon Eye prototypes today employ off-the-shelf daylight color video and low-light black-and-white cameras. Raytheon, however, is developing a half-pound, 640-by-480-pixel, uncooled infrared (IR) camera to be flight-tested this month. The relatively quiet, electrically powered vehicle could also carry an acoustic sensor to identify ground vehicles.
The concept of operation is evolving and will continue to develop as the vehicles reach the fleet. The plan now is to give individual infantry battalions six systems to deploy at their discretion. A battalion-level surveillance/target acquisition platoon could be attached to a company commander, or the platoon could relay the information through battalion headquarters. Both techniques have been demonstrated.
The USMC also is developing Dragon Warrior, a larger, vertical-takeoff UAV with 50-nautical mile range, 15,000-foot service ceiling and 3- to 5-hour endurance. It would be employed at altitudes of 2,000 to 3,000 feet. The UAV would belong to an infantry regiment but could move with a battalion, Cane says. The service plans an electro-optical (EO)/IR sensor and a wideband communications relay payload–based on the AN/VRC-99A tactical radio–that could be the "reach-back to the ship from the guys ashore."
Micro Air Vehicles
The Army and DARPA, meanwhile, plan to test a family of largely autonomous, squad-level Micro Air Vehicles made by Allied Aerospace. The five-year, $31-million advanced concept technology demonstration (ACTD), scheduled to begin last month, will examine electrically powered, diesel-powered and hybrid, electric/diesel vehicles. They weigh from 6 to 10 pounds (2.7 to 4.5 kg), with an EO/IR payload of under 8 ounces (227 grams).
These 9-inch (22.8-cm)-diameter, ducted-fan UAVs, intended for urban warfare, take off vertically, dip over and fly horizontally, land on a building ledge and transmit images. The electric MAV can hover and fly at up to 68 knots–almost twice the speed of Dragon Eye, though only for 10 minutes–and "perch and stare" for 24 hours. The "E-MAV" has a service ceiling of 17,000 feet. The diesel MAV’s target flight time is 90 minutes and perch time is 10 hours, with a service ceiling of 15,000 feet. Its target speed is 81 mph.
DARPA envisions a three-phase, overlapping schedule, running the three MAV versions. Testing of the electrically powered vehicle was to begin last month, followed by the diesel version in January 2003, and the hybrid MAV in October 2003. The system is autonomous, says Sam Wilson, ACTD program manager. "It will land on the edge of a building with no human involved."
The first phase of testing also will look at differential GPS and will use "vision engine" technology to map the territory to land on. Each test cycle will involve a period with the Army’s 25th Infantry Division at Schofield Barracks, Hawaii. The end result: an "optimum MAV" for war gaming with the U.S. Pacific Command in early 2005. Afterwards the Army will have two years to further experiment with 25 MAV systems, before deciding whether to buy production units.
The exercise will employ EO, IR and a combination of the two sensors in the first, second and third rounds of the trial, respectively. A lightweight, uncooled IR camera, 360-degree EO camera and ultra wide-band (UWB) radars also are being developed for this UAV class (September 2001, page 34). Imagery will be updated 30 times per second, but a frame will be transmitted only if change has been detected. In the trials the UAVs will fly vignettes such as reconnaissance orbits, transmission from a prepositioned, forward location, and change of watch.
The MAVs are intended to operate with a "land warrior" type of interface–a futuristic infantry ensemble, including body armor, wearable computer, radio, video head-up display and small unit wireless network. DARPA hopes to use the wearable computer in the second phase of the ACTD and the small unit local area network in the third phase.
…And Long Term
The Navy’s Autonomous Intelligent Networks and Systems (AINS) project envisions a drone armada, tens to thousands strong, as part of a decentralized command and control structure of the future. This swarm takes a high-level human order, identifies and executes the necessary tasks, and makes decisions based on what it observes, so that its missions can continue when lines of communication with battle commanders are broken. The armada is a sensor and communications network, a real-time, fault-tolerant "Internet in the sky," including multiple airframe and sensor types. Whether or not the concept is ever fielded at full scale, the research is attracting high-level interest and will spin off advancements in dynamic wireless networking, vehicle intelligence and adaptive control.
Picture a group of unmanned air and ground robots tracking and pursuing a threat vehicle. This scenario was demonstrated by researchers at the University of California at Berkeley last year. The AINS program planned to run a hybrid simulation in October 2002, including humans, actual unmanned vehicles and virtual UAVs, to test the accuracy of multivehicle teaming and clustering algorithms. Researchers are expected to show certain key capabilities each year, says Allen Moshfegh, science officer at the Office of Naval Research (ONR), which manages the program.
Low-cost swarm vehicles, flying at 30 to 20,000 feet, would be part of a larger, multilayered architecture, also including:
High-altitude pseudo satellites (80,000 to 120,000 feet),
Medium-altitude, high-resolution camera platforms (50,000 to 60,000 feet), and
Unmanned ground and sea vehicles.
The project has harnessed an army of researchers, tackling prickly problems. How, for example, to avoid interference between many UAV antennas, guarantee acceptable error rates, and transmit large volumes of data while flying at high speeds in heavy jamming? A special radio is being developed under the AINS program that will combine multi-input/multi-output (MIMO) and orthogonal frequency division multiplexing (OFDM) techniques.
"Through radio concepts we can predict our link environment," says Moshfegh. "We do periodic learning of our transmission environment, model it, and based on that, we manage our modulation and coding techniques."
Researchers at the University of California at Los Angeles (UCLA) plan to build an 8-by-8 radio system–eight transmit and eight receive antennas–capable of 1-Gbit/sec transmissions, says Babak Daneshrad, a professor of electrical engineering. The radio will exploit MIMO techniques, which allow different signals to be transmitted on different antenna elements at the same frequency and the same time. "We can undo the interference and deliver clean signals at the receiver by employing advanced signal processing," Daneshrad claims.
ONR plans a simulation in the last year of the UCLA program. "We’ll deploy 50 to 100 vehicles and scale to 10,000 nodes in a simulated environment," Moshfegh says. The exercise will be used to validate underlying theories and concepts.
The radio also will employ OFDM modulation, which allows the transmission of different signals on a single antenna through the use of multiple, overlapping frequency subchannels.
With MIMO, "you excite each antenna element with a different signal at the same frequency, linearly increasing throughput," Daneshrad explains. OFDM, on the other hand, "allows you to generate many, many small channels and transmit a different signal on each subchannel."
MIMO is one step beyond phased arrays, he says. "Phased arrays transmit the same signal on different antenna elements, giving you gain, whereas the MIMO concept transmits different signals on different antennas, increasing throughput." Parameters such as power and range can be traded off against each other, as long as something, such as quality of service, remains constant.
Using both techniques together, "you get a two-dimensional signaling space–spatial separation plus frequency separation," Daneshrad says. "Throughput increases and outage decreases–exponentially." The hybrid technique is suited to urban warfare, where buildings hamper radio operation and congestion can be severe. The combined approach is the most efficient way of transmitting data in limited spectra, he argues.
More Power Needed
While these concepts are well understood in communications theory, they have not been demonstrated together. The main challenge is signal processing. If designers used conventional digital signal processors (DSPs), they would need thousands of these chips to achieve the required 1 teraflop (1 trillion floating point operations) of power, Daneshrad says. He hopes to achieve this with four or five application-specific integrated circuits (ASICs). So far, the UCLA team has a 1-by-1 radio. Signal processing algorithms have been identified, and ASIC design will begin next year.
Another problem is communications delay. Messages can’t be stranded on a router-UAV, as they are on the Internet. The swarm will require an elaborate message priority scheme, plus at least 128-bit addressing techniques, including GPS location, altitude and team/node IDs. The communications architecture would be multilayered, like today’s Internet. But requests from high-level mission software to the underlying hardware won’t have to trickle down through multiple protocol layers, and the hardware layer’s answers won’t have to climb slowly back up again to tell the application whether a request can be executed or how it must be constrained. "The goal is to adjust all these layers simultaneously, so we can send time-critical information that guarantees quality of service, latency and priority-based access," Moshfegh says.
Another challenge is network intelligence–how to make the vehicles solve not just immediate problems, but remember and execute longer-range goals, such as getting to a certain place at a certain time? A team at SRI is focusing on these problems, which haven’t received much study, says program manager Charlie Ortiz. "We are trying to develop systems that can react quickly to problems and consider the longer-term consequences of their actions."
On the air vehicle side, ONR is using relatively low-cost UAVs designed by Advanced Ceramics Research (ACR), in Tucson, Ariz. The 4-foot (1.2-meter) wingspan aircraft uses off-the-shelf avionics, including differential GPS, says President and CEO Tony Mulligan. These fixed-wing vehicles can fly about 65 miles per hour, at 1,000 feet with a range of up to 150 miles (241 km) and a payload of up to 4 pounds (1.8 kg). ACR expects to quadruple the range in the coming year. One ground station can fly up to 10 planes simultaneously. The current price of $20,000 per unit–with a $7,000 avionics package–could drop to $2,000 if the vehicles are purchased in volume. The company also is working on a bigger drone, weighing 75 to 100 pounds (34 to 45.3 kg) and exploring the launch of tiny UAVs from a larger unmanned vehicle. Current sensors are commercial color video and low-light cameras.
Although realization of the AINS concept may be years away, near-term applications will surface, Mulligan contends. ACR is developing a "client/server interaction" concept that would allow several unmanned vehicles to be controlled by the driver of a Humvee or jeep. The idea is to increase situational awareness and see what’s coming up beyond a soldier’s view. When the truck turns, the UAVs turn, and when the truck stops the UAVs circle overhead. ACR planned to begin demonstrating the concept in October, at Camp Roberts, Calif., starting with a single vehicle.
The AINS program also is developing "pseudo GPS" navigation techniques, using pseudo satellites and ground sensors to triangulate position. Passive landing techniques were demonstrated in August 2001, as an unmanned helicopter landed on a truck-mounted platform, or "deck." ONR plans a demo in February, where the truck will move and the deck will heave to mimic different sea states.