This article reviews that Defense Advanced Research Projects Agency’s (DARPA) biomimetic programs is a robotic lobster that is under development at Northeastern University in Boston. This crustacean look-alike may someday ply river and sea bottoms, at depths to 40 feet, seeking underwater mines and other military prey. The robotic lobster will have to operate for hours, accommodate irregular rivers and sea beds, maneuver at various depths, adapt to rough and tumble surf, handle changing currents, distinguish between rocks and mines, and send out a sonar alert when it detects a mine. The sensors, which are the width of a human hair, are fabricated using an internally developed process called NUMEM (for Northeastern University Metal Micromachining), which builds up the devices through a sequence of metal deposition, patterning and selective etching, and plating on a silicon substrate. Signals from both the antenna and the hair sensors are processed in the lobster’s microprocessor brain and used to control the bionic leg muscles.
Creatures that creep, crawl, skitter, or fly can infiltrate the most intensely fortified military defenses much more deftly and secretively than any highly trained humanoid. Unfortunately, creepy crawlies don’t train very well. So instead of conscripting beetles, bees, cockroaches, or scorpions, the Defense Advanced Research Projects Agency hopes to mimic them.
Emulating Mother Nature—never an easy task—is especially difficult when it involves an entire creature, not just a particular sensory function. It means melding multiple disciplines such as mechanical and electrical engineering, physics, mathematics, and biology, to name a few, into a science called biomimetics, that has as its goal the creation of a replica that can perform the same tasks as the original.
A beetle that seeks burnt wood on which to lay its eggs has the remarkable ability to detect smoke and infrared emissions from a forest fire 70 kilometers away. By mimicking the beetle’s sensor capability, its robotic counterpart may be able to detect chemical or infrared emissions with greater sensitivity than is now possible.
Even bees and other insects are fodder for emulation. But attempts to build micromechanical insects has required that scientists puzzle over questions such as, “How does an insect in the low Reynolds number regimen get off the ground?” and, “How does a fly manage to take off backward, fly sideways, and land upside down?”
“Our program has resulted in some fundamental discoveries about rotational lift,” said Alan Rudolph, the program manager at DARPA’s Defense Sciences Office in Washington. Until very recently, engineers were confounded as to just how the aerodynamics in small insects worked.
One of the more advanced DARPA biomimetic programs is a robotic lobster that is under development at Northeastern University in Boston. This crustacean look-alike may someday ply river and sea bottoms, at depths to 40 feet, seeking underwater mines and other military prey.
“Though the present program won’t deliver a lobster ready for military deployment,” Rudolph said, “we do expect to demonstrate significantly greater functionality than now exists for mine hunting and surveillance.”
Copying A Crustacean
The robotic lobster will have to operate for hours, accommodate irregular river and sea beds, maneuver at various depths, adapt to rough and tumble surf, handle changing currents, distinguish between rocks and mines, and send out a sonar alert when it detects a mine—all while being perceived as nothing more than an innocuous critter scurrying across the bottom. The price: $300 per robot.
Achieving this ambulatory and sensory miracle is the responsibility of principal investigator Joseph Ayers, director of Northeastern University’s Marine Science Center, who is directing 22 scientists around the world in reverse engineering of the lobster.
Ayers and his students have analyzed thousands of hours of videotapes of real lobsters, converted the movements of legs, claws, abdomen, and tail into their mathematical components, and correlated the movements with the nerve signals that actuated them.
“Analyses have shown that to maneuver in rocky, turbulent surf, lobsters control their legs and posture so they don’t get tossed about or tumbled against rocks,” Rudolph explained.
Sensing the Environment
A lobster moves its legs in response to feedback from its antennae and from the sensitive hairs growing in surface indentations on its claws. The antennae bend upon contact with an obstacle, while the hairs are bent by water flowing past the claws as the lobster moves.
To sense bending or force, engineers typically rely on one of three categories of strain gauges: thin-film metal foil, piezoresistive, or capacitive array. While these devices are highly accurate and capable of measuring a continuous range of strain, they demand a continuous power source during operation. Since the robotic lobster probably will be battery-driven, these sensors could present a substantial drain on resources. To reduce power consumption, researchers at Northeastern are turning to switch-based sensors because switches don’t draw current until they are activated.
The sensors, which are the width of a human hair, are fabricated using an internally developed process called NUMEM (for Northeastern University Metal Micromachining), which builds up the devices through a sequence of metal deposition, patterning and selective etching, and plating on a silicon substrate.
“Most developers do polysilicon micromachining,” said Nicol McGruer, an associate professor of electrical engineering. “We build up layers of silicon dioxide, copper, and gold, then remove the copper, leaving the gold suspended above the substrate. The whole process takes less than 200°C compared with the 800 to 900°C for the polysilicon micromachining. Because it is compatible with IC postprocessing, we can easily build the MEMs device on top of a chip.”
The antenna sensor is a normally open switch consisting of a gold cantilever and contact pad mounted on an extremely flexible silicon substrate. When the antenna encounters an obstacle, the substrate bends, closing the switch. The closer the object, the greater the bending. Since each cantilever detects only one curvature, multiple switches are needed to detect a range of curvatures.
The sensor fits inside an artificial antenna fabricated from a pair of 0.015-inch-thick PVC sheets sandwiched together. Before the sheets are joined, each is milled with a 0.005-inch slot to hold the sensor and its associated flexible circuitry.
“The challenge is to create a design that will bend with the antenna without developing fatigue failure,” McGruer noted. “Once we achieve that, we have to be able to seal it so water won’t get in.”
Sometimes the sealing process itself can turn a working device into a nonworking one. For example, materials heated during the sealing process may outgas and deposit on the contacts.
The hair, or flow, sensor uses a gold cantilever paddle bent perpendicular to the substrate and the flow path. The pressure of the flow against the paddle causes the cantilever to close. The flow rate that will close a switch depends on paddle area and thickness, contact location, and contact gap, so several sensors in different sizes are needed to detect various rates of current.
Flow rate sensitivity is manipulated primarily by varying the paddle area. The smallest paddle—150 microns long, 80 wide, and 8 thick—will detect a flow of 5.4 meters per second. The gap between the contact points on the cantilever and substrate is 2,500 angstroms, or a quarter of a micron.
The switch is a unidirectional device and to determine flow direction four switches are required, one mounted at each of the four compass points. Since each switch is an On/Off device, it delivers only a single bit of information. Four bits can be produced using arrays of switches or by building multiple contacts into one device.
Walking the Walk
Signals from both the antenna and the hair sensors are processed in the lobster’s microprocessor brain and used to control the bionic leg muscles.
“Think of extending your arm straight out,” explained George Adams, a professor of mechanical engineering at Northeastern. “You would lift it by contracting the biceps muscle. Likewise, lobsters move their legs by contracting their muscles. We mimic these muscles with nitinol wire that contracts when heated by an electrical current and moves the lobster leg upward, producing locomotion.
Nitinol is a nickel-titanium, shaped- memory alloy that contracts by as much as 10 percent when heated to 150°F. Once cooled, it returns to its original shape. By alternately heating and cooling the nitinol wire, researchers can replicate a more natural leg motion than is possible with any arrangement of motors and gears.
The same response is being tapped in a companion DARPA project that has not advanced as far: the development of a robotic lamprey to search for floating mines offshore. The natural lamprey uses hundreds of muscle bands to swim in an eel-like fashion. Its robotic counterpart combines a plastic backbone with short overlapping segments of nitinol that are alternately heated and cooled four times per second to reproduce the muscle contractions that propel swimming.
A major challenge still facing the lobster program is development of a power pack to drive the sensors, microprocessors, and moving parts. Rudolph feels that any pack must supply power that is adequate to perform useful work for four hours.
As exciting as robotic critters may be, don’t expect them to go marching into the sea any time soon. However, you may find some near-relatives traipsing across your living room by Christmas.
“Robotics is capturing some sophisticated functionality and doing it in a cost-effective way,” Rudolph remarked. “The entertainment industry has picked up on this, and I expect by next Christmas we will see a plethora of intelligent robots headed our way.”