Environews Innovations

Environmental Health Perspectives • VOLUME 112 | NUMBER 8 | June 2004 A 487

Imag

e So

urce

, Pho

todi

sc, M

att

Ray/

EHP

L obsters have a keenly

developed sense of smell,

which they use to detect and

trace the odor of their food to

its source in the ever-turbulent

ocean. Scientists working in a

new field known as biomimet-

ic robotics believe that humans

can solve real-world problems

by dissecting this and other

forms of animal intelligence,

and then using that knowledge

to design, build, and program

autonomous machines with

similar superhuman capabili-

ties. Eventually, such robots

could be used to track and pin-

point underwater sources of

pollution, detect and locate

mines and other unexploded

ordnance, and even troll the

ocean’s depths for thermal vents

and other locations offering

untapped natural resources.

First, the ScienceNeurobiologist Frank Grasso, an

associate professor of psychology

at Brooklyn College, is one of a

growing number of biomimetics

researchers. Grasso’s principal aim

is to advance scientific knowledge

in the field. “We’re not really in

the business of developing new

technologies,” he says. “We’re in

the business of developing new

understandings of the organiza-

tion of controlled behavior.”

Grasso works with lobsters in

order to learn how they are so

remarkably adept at tracking an

odor to its source in a turbulent

Innovations | RoboLobsters

RoboLobstersThe Beauty of Biomimetics

A 488 VOLUME 112 | NUMBER 8 | June 2004 • Environmental Health Perspectives

Innovations | RoboLobsters

Arie

l Cha

ouat

and

Nad

av S

hash

ar/In

teru

nive

rsity

Inst

itute

of

Eila

t

environment, and how to translate what islearned into new capabilities that humanscan utilize. “Our understanding of fluidmechanics and the natural environment ispoor compared to what the lobster is able todo with the kind of information that is outthere,” Grasso says. “The animals are capableof taking olfactory information and makingdecisions with it at a much faster time scalethan we’re able to understand right now.”

By closely observing the lob-sters as they sense and track largeplumes of chemicals to theirsources under controlled laborato-ry conditions, Grasso and hiscoworkers develop theories as tohow they might be doing so.They incorporate those hypothe-ses into design parameters andprogramming for their roboticlobsters. “Then we test the robotunder the same conditions as theanimal, as a way of validating ourtheories as to how the animal isdoing the tracking,” he explains.

The autonomous robots onlynominally resemble actual lob-sters. The arrangement of sensorsand the locomotion apparatus areintended to physically support thedevices’ programmed strategies,just as the animal’s body and sen-sory organs are inherently interfaced with itsneurological information processing anddecision making.

Although physical resemblance itself isnot a goal, a principle that Grasso calls “bio-mimetic scaling” is a critical factor in theexperimental framework. Biomimetic scalingdictates that a robot need mimic the animal’sbiological characteristics only to the degreethat those features support the concept beingtested. For example, in a set of experimentsreported in the 31 January 2000 issue ofRobotics and Autonomous Systems, Grassoand colleagues designed and programmed arobot with known lobster values for bodysize and shape, sensor arrangement in space,speed and pattern of locomotion, and tem-poral and spatial resolution of the sensors inthe robot’s hardware and software.

The resulting robot is somewhat reminis-cent of a lobster, but only insofar as thosefeatures allow the confirmation of the theo-ries behind the mimicry. Evening the experi-mental playing field in this manner results inongoing, interactive refinements of thehypotheses and the robots themselves asthey improve in their ability to mimic reallobsters’ behavior.

A Sense of ProgressA lobster’s antennules, the extraordinarilysensitive chemosensory organs that protrude

from its head, are responsible for the crus-tacean’s superb sense of smell. “We’ve spenta lot of effort characterizing how the anten-nules move and how they sample in space,”says Grasso. “You can observe them direct-ly, and we’ve got a pretty good idea ofwhat’s going on there. The more importantquestion is what’s going on once the anten-nules’ information passes through the brain,and how it’s processed.”

The very fact that lobsters can do whatthey do—overcome turbulence and effi-ciently track underwater odors to theirsource—proves that such an extremely com-plex problem can be solved, and Grasso isdetermined to find out how so that humanscan duplicate the skill. He explains, “It’s amatter of being able to ask the right ques-tions of the lobsters and of any animal westudy, to be able to constrain the modelsthat we implement in the robots to be real-istic in describing the transformations thatthe nervous system of the lobster produces.”Ultimately, Grasso hopes the work will leadto plume-tracking systems that exceed eventhe lobster’s innate abilities: “If we movetoward the lobster’s solutions, then alongthe way we may discover a turn that thelobster was evolutionarily unable to take,that we could exploit and use to developeven better systems.”

Although they are far from ready tobe deployed as marine pollution fightersor mine detectors, Grasso’s robotic lob-sters have demonstrated that the conceptis valid and worthy of further investmentof research funding and effort. His firstrobot, called RoboLobster, successfullytracked plumes from more than 32 feetin laboratory conditions, where the flowof the turbulence and plumes could becontrolled.

In 2002, Grasso took a pair of second-generation robots to the Red Sea for afield trial. “We took the algorithms, thestrategies that we had put into the robots,and tested them under uncontrolled con-ditions to see whether the intelligence wehad extracted from the lobsters and putinto the robots could actually bear up inthe natural environment,” says Grasso.“And remarkably, the robots didn’t fall

apart—they were able to trackthe plume with a comparableefficiency to what they had donein the laboratory.” Although henotes that success in the Red Seatrials does not mean the robotswere tracking exactly as theircrustacean counterparts would,Grasso says the tests were amajor validation of the operat-ing principles at that point.

Pollution Patrol“Dr. Grasso’s research may proveto be very important in the not-too-distant future for fightingpollution in our rivers, lakes, andoceans,” says Roger Quinn, a pro-fessor of mechanical engineeringat Case Western Reserve Univ-ersity. “Conceivably this controlsystem can be used to guide

robots as they patrol bodies of water to findpossible pollutants and determine theirsource before great harm is done to theenvironment.”

With most of his funding coming fromthe U.S. Navy, Grasso’s principal charge isto develop the autonomous agents to detectunexploded mines, but he sees applicationof the technology to determine pointsources of pollution as a natural extensionof the robots’ ability to sniff out and homein on the source of any chemical that leavesa trail. “[The ability to detect] the source ofan odor could be applied to finding sourcesof pollution coming out of plants,” he says.“You could even imagine tracking the wakeof a leaky trawler or some other oceangoingvessel that had been polluting. The onlylimitation will be how smart we can makethe robots, and how well we can make thesensors that are attuned to the things wewant to track.”

The next hurdle Grasso wants toaddress with biomimetics is how to track amoving target that generates a turbulentwake, such as that same leaky trawler. Thatadds several layers of difficulty to the chal-lenge, and the lobsters and their roboticdoppelgängers may not be up to it—because a lobster crawls along the oceanfloor chasing stationary targets, itsapproach is basically two-dimensional.

Robotics 101. Frank Grasso delivers a briefing onshore as students preparefor an underwater trial of the RoboLobster.

Environmental Health Perspectives • VOLUME 112 | NUMBER 8 | June 2004 A 489

Innovations | RoboLobstersA

riel C

haou

at a

nd N

adav

Sha

shar

/Inte

runi

vers

ity In

stitu

te o

f Ei

lat

Fish, on the other hand, perform thatkind of three-dimensional tracking routinelyas they detect and chase their prey, and theyrepresent the next generation of roboticsand study for Grasso’s group. In this case,

however, the researchers are building robot-ic fish that will allow study of the cues thatpredatory fish use to track fish. By mimick-ing the hydrodynamic and chemical distur-bances made by a moving “bait” fish, theresearchers hope to find out how predatorstrack those cues to their sources. A firstreport from this work was published in the19 June 2001 Proceedings of the NationalAcademy of Sciences.

It’s a more complex undertaking, because“the signal used in tracking a chemical to itssource could be a mixture of the mechanicaldisturbance that’s produced by the object

itself—the source of the odor—and thechemicals that are streaming off of it,” saysGrasso. The clear question to be asked withthis fish system, he says, is whether the fishare actually tracking chemicals or whether

they are following the disturbance left in thewake. “It’s most likely a ratio of the two,” headds, “and that leads to three-dimensional,multisensory integration, which is a reallyinteresting and challenging problem.”

A Biomimetic ZooIt appears that almost any animal hasunique qualities worth mimicking. Grasso,aside from his robotic lobsters and fish, isalso working on modeling and buildingrobotic octopi. The boneless tentacles of theoctopus function like so-called soft actua-tors: the muscles alternately supply force

generation and mechanical support to moveobjects along the length of the tentacle.

“If you’ve ever noticed,” Grasso ex-plains, “on the underside of an octopus,there are long rows of suckers, which theanimal can rotate once they’ve graspedahold of something, and they can walk theobject along the length of the arm one wayor the other. So there’s a whole set of actua-tors that can be produced that are soft, flexi-ble, and somewhat intelligent, because theycan actually taste and feel the things thatthey’re manipulating, and dynamicallyadapt to them.”

Grasso’s group is one of several involvedin a large project designed to build bio-mimetic soft actuators capable of operatingin a cluttered environment, maneuveringover and around objects, and then grabbingan object of interest and manipulating it.Eventually, a robotic octopus could be usedto move through rubble to rescue earth-quake victims, among other uses.

Pursuing a wide variety of practicalapplications as well as the scientific knowl-edge they yield, other researchers in bio-mimetic robotics have produced mechani-cal mimics of lampreys, cockroaches, crick-ets, flies, snakes, dolphins, scorpions, andmore. Quinn’s group, for example, hasdeveloped small robots mimicking thecockroach’s locomotive system. Theirdevices use “whegs”—cockroach-like com-binations of wheels and legs—to climbover obstacles. Quinn says Grasso’s work“is of particular interest to us because weplan to develop highly mobile underwaterrobots based on our ‘whegs’ technologythat could be outfitted with [a chemicalsensing and tracking] system.”

As the field progresses, and the manygroups working in biomimetic roboticsincreasingly collaborate and combine theirspecialized technologies, it seems likely thatin the near future we will routinely seemachines exploiting for human use many ofthe advantageous features nature hasbestowed on animals over millions of yearsof evolution.

Ernie Hood

Sugge s t ed Read ingAyers J, Davis JL, Rudolph A, eds. 2002. Neurotechnology for Biomimetic Robots.

Cambridge, MA: MIT Press.

Blazis DEJ, Grasso FW, et al. 2001. Proceedings of Invertebrate Sensory InformationProcessing: Implications for Biologically Inspired Autonomous Systems, 15–17 April2000, Woods Hole, Massachusetts. Biol Bull 200(2):147–242.

Vogel S. 2001. Cat’s Paws and Catapults: Mechanical Worlds of Nature and People.New York, NY: W.W. Norton & Company.

Webb B, Consi TR, eds. 2001. Biorobotics: Methods & Applications. Menlo Park, CA:AAAI Press.

Crustacean demonstration. RoboLobster trials are conducted in the field at a depth of5 meters (right; robot is in center of photo). Dye marks the substance to be sensed.Students score robot behavior by tracking its movements relative to the dye (above).