Darwin’s Devices

Chapter 8 SO LONG, AND THANKS FOR ALL THE ROBOTIC FISH[171]

I’VE SEEN THE FUTURE, AND IT’S FULL OF FISH. NOT REAL FISH, of course. We’ve eaten most of those. I mean the robotic kind. Robotic fish are taking the place of the real ones. At the London Aquarium in 2005 you could have seen three bejeweled robotic fish, built by Professor Huosheng Hu and his team at Essex University, swimming on display. A swimming coelacanth, built by Mitsubishi Heavy Industries, made news in 2001 by making the rounds in the AquaTom in the Fukui Prefecture.[172] The French company RobotSwim has their new Jessiko autonomous robotic fish, ready for your swimming pool, local aquarium, or exploration and marine-monitoring mission.[173] Maurizio Porfiri, an engineering professor at the Polytechnic Institute at New York University, is building a robotic fish to herd real fish away from danger.[174] His NYU colleague, Professor Farshad Khorrami, has created a company, FarCo Technologies, that builds biomimetic robotic fish for industrial and defense-related situations.[175]

Why all the fuss about robotic fish? What’s in it for you and me? Will a robotic fish become your best friend, save your life, or overthrow an evil dictator? Maybe. For certain, robotic fish will help us do what we can’t naturally do: be underwater. As extensions of our hands and eyes, robotic fish are embodied-brain tools we primates will use to probe the aquatic depths. And as we’ve seen throughout this book, robotic fish are built with representations of bits and pieces of real fish, sometimes to learn more about fish, if you are a crazy biological cognitive scientist, or, most of the time, to build better machines, if you are an engineer with a job to do.

We all fall prey to the notion that evolution by natural selection is a better engineer than a hominid with a PhD from MIT. The implicit basis for this romantic view of nature as an engineer is that evolution perfects: “Evolution is a slow but sure process of perfecting design to give a life-form a natural advantage in a competitive environment.”[176] Steve Vogel, in his book Cats’ Paws and Catapults, explains the counterargument to this perfection bias beautifully: “Nature does what she does very well indeed. But—and here’s the rub—why should she do so in the best possible way?”[177] Indeed.

Think about PreyRo. Even when that population of Evolvabots appears to have evolved a mechanically optimum tail stiffness, that doesn’t mean that an average of 5.7 is the perfect solution forever and for everywhere. It’s only the best solution—relative to others in the population—at that moment and place in the adaptive landscape. As we talked about in the last chapter, the adaptive landscape usually contains multiple peaks. In the face of such rugged terrain the best that evolution can hope to do is find the closest peak, the “local maximum” in the mountain lore of mathematicians. Even with evolution by natural selection in full hill-climbing mode, it can be pushed off course by random forces and prevented from getting underway by the historical constraints of the population’s genetic history.

Instead, evolution suffices. It may provide just-good-enough solutions that aren’t quite in time, but it doesn’t even have to do that. (Nice work if you can get it, eh?) Selection, that judgeless judging environment with which an individual ceaselessly and unknowingly interacts, plays a strong role in choosing the breeders who make the next generation. But by the time the next generation is on the scene, the world may have changed, creating a new adaptive landscape that the previous, and different, selection environment could not anticipate. Nothing in the rule book for the game of life says that the playing field has to be level or even has to stay the same. In fact, except in unusual places such as the abyssal zone at the bottom of the ocean, the adaptive landscape for any population is better thought of as an “adaptive seascape,” as suggested by Professor David Merrill in his eponymous book.

Given all of this complexity and contingency in the adaptive seascape, do we really want to assume that we can look to the living world to have solved all of our engineering problems and to have done so perfectly?

No, we don’t. But neither do we want to pretend that we have nothing to learn from nature. Engineers understand this, and they understand that they need more in their toolbox than just the dry goods—stiff steel, flexible plastic, compressive concrete, and resilient rubber—with which we’ve built the constructed world around us. Phil Leduc, associate professor of mechanical engineering at Carnegie Mellon University, works as a nanoengineer, manipulating proteins within living cells, linking the mechanical behavior of one to the biological function of the other in order to design bioinspired nanofactories. Kenneth Breuer, professor of engineering and director of the Fluid Dynamics Laboratory at Brown University, collaborates with Sharon Swartz, professor of biology at Brown University, to study the complex anatomy and behavior of flying bats and their extremely flexible wings as part of a larger project with engineers from the University of Michigan to build to a bat-inspired micro-air vehicle for the US Air Force Research Laboratory. Melina Hale, associate professor of organismal biology and anatomy at the University of Chicago and a fish expert working on robotic fish, offers this insight: “with all the tremendous work that has gone into designing and building robots, we are still far from having one that functions as well as a fish.”[178] It holds for any bioinspired project.

What don’t our robots do that fish do? Just about everything: fish have ways of sensing, navigating, and moving underwater, according to Maarja Kruusmaa, professor of biorobotics at Tallinn University of Technology in Estonia, that none of our robots have.

Maarja’s team—an international one, comprising biologists and engineers from the Italian Institute of Technology, Riga Technical University in Latvia, the University of Verona in Italy, and the University of Bath in the United Kingdom—is designing a biomimetic lateral line and a robotic fish to take advantage of the new sensory capabilities the lateral line creates. We’ve discussed some of them: identifying the position of swimming companions in a school of fish at night, the location and proximity of an approaching predator, and the efficiency of self-generated water movement. All of these functions are beyond the scope of any single sensor currently made by humans. By building a biomimetic lateral line, Maarja and her team are betting that their robotic fish will be able to teach us land lubbers about physical structure in the aquatic world that we can’t comprehend ourselves.

When I visited Maarja’s biorobotics laboratory in spring 2009 I was struck by the ambition of the FILOSE (robotic FIsh LOcomotion and SEnsing) project.[179] The biomimetic lateral line will be placed on a self-propelled robotic fish (Figure 8.1) that controls its swimming performance by varying both neural control variables, like the frequency of its tail beat, and the mechanical properties of the body, such as stiffness.[180] The robotic fish will have to learn how the patterns of flow sensors distributed around its body relate to the external flow patterns and the detection of underwater objects; the FILOSE engineers have used a digital simulation to show that this is possible. This work is funded by the European Commission under their Seventh Framework Programme, which pursues the “European Union’s Lisbon Strategy to become the ‘most dynamic competitive knowledge-based economy in the world.’”[181]

In the United States the National Science Foundation is funding work on fish to uncover novel ways to create highly maneuverable underwater vehicles. Malcolm McIver, professor of mechanical engineering at Northwestern University in Illinois, leads a team of engineers and biologists, including George Lauder of Harvard, studying the precision maneuvering of the black ghost knife fish.[182] Knife fish are so called because of their tapered and stiff body. Unlike other fish, knife fish don’t bend their bodies to swim; instead, they ripple a long, thin fin that runs along their belly. This fin enables them to do something unusual among fishes: move vertically without having to point upward. By building a biomimetic robot, Malcolm’s team has shown how this happens and that the function, once understood as a mechanical principle, can be used by engineers building underwater vehicles, even if those vehicles are otherwise unfish-like.

Using a process he calls “biologically derived design,” James Tangorra, assistant professor of mechanical engineering and mechanics at Drexel University, builds robotic bluegill sunfish with George and Melina. Interested in efficient and novel propulsion, they focus on fins and their internal structures—thin, bifurcated rods called rays. In each of the sunfish’s two pectoral fins, fourteen fin rays are independently controlled to bend, cup, and curl the fin. This level of structural control, which occurs dynamically during swimming, is unprecedented in engineering systems. By building fish-inspired robotic structures, the team has learned something new about real fish fins in the process: they bend using a novel structural mechanism involving the shearing of two parallel struts connected at one end.[183] With any flexible, animated structure, like a fin or a whole undulating body for that matter, neural control of its motion becomes the next frontier, as Melina explained when asked about design challenges facing her team and others building robotic fish: “we need to incorporate some sort of realistic understanding of sensation and sensory processing—taking in the sensory input and processing multiple inputs to, in a sense, decide on a motor output.”

The FILOSE fish, the robotic knife fish, and the robotic sunfish fins projects show how fish inspire engineers. Engineers are also building robotic fish in China, France, Indonesia, Japan, Korea, Singapore, and the United Kingdom.[184] Among most of these fish-inspired projects we see three different approaches for getting started: (1) identify a function of a fish or function within a fish that is novel to engineering and can be reduced to a mechanical or mathematical principle, (2) identify a behavior of a fish or system of fish that is novel to engineering and can be reduced to an algorithm, or (3) identify a structure of a fish that is novel to engineering and can be duplicated in other materials. Clearly, novelties in aid of applications are what roboteers fish for.

For Frank Fish—who we met in the last chapter and who, funded by the US Office of Naval Research, heads a multi-institutional building of a robotic manta ray—nautical engineers’ interest in fish has been constant and longstanding. “It all comes down to four words,” he explained, “speed, maneuverability, efficiency, and stealth.” But there’s more to the story than just that: when I asked Frank why so many engineers were building robotic fish, without hesitation he said, “Because the Navy is funding them.”[185]

The US Navy, through the Office of Naval Research (ONR), has been funding work on fish and dolphins since the office’s inception in 1946, leveraging academic research to help engineers understand aquatic propulsion.[186] Having had two ONR research grants in the 1990s with my mathematics collaborator from Lafayette College, Rob Root, I’ve benefited from the Navy’s interest in fish and robotic fish.

Rob and I conducted experiments and built mathematical models of swimming fish with much help from students like Craig Blanchette, Nick Boetticher, Hayden-William Courtland, Vynette Haultaufderhyde, Wyatt Korff, Nicole Lamb, Nicole Librizzi, Matt McHenry, Karen Nipper, David Paul, William Shepherd, Eamon Twohig, and Stephanie Varga. We were able, with additional help from our collaborators Peter Czuwala, Lena Koob-Emunds, Tom Koob, and Chuck Pell, to develop theory about the importance of body stiffness in swimming fish.

The punch line: yes, you can tune a fish. Fish tune their bodies by changing stiffness. Because stiffness is proportional to the speed[187] at which a wave travels or vibrates, using muscles to stiffen the body will drive the fish’s flexural waves faster. When you make undulatory body waves, you perform the mechanical work of transferring the body’s momentum to the water, and when the body’s flexural waves move faster, you are generating more power.

When we had some evidence that real sunfish were using their muscles to alter body stiffness, ONR gave us a chance to build our first self-propelled robotic fish.[188] Matt McHenry and I were anxious to make Nektors in the image of our study species, pumpkinseed sunfish; Chuck Pell, one of the Nektor’s co-inventors, kindly agreed to share his invention and his talents. He created a mold using a dead sunfish, from which he fabricated five identically shaped sunfish-Nektors. Each sunfish model was made out of a slightly different formulation of PVC rubber, which is how Chuck varied the models’ material stiffnesses over a range that included the stiffness values we had measured in real sunfish.

With help from Vassar Professor Bob Suter, Matt and I developed a way to swim the sunfish model in a flow tank—kind of a treadmill for fish. For a given and constant frequency, each sunfish model swam upstream in the initially still flow tank. We increased the flow speed until the sunfish model could just swim in place, moving neither upstream nor downstream, balancing the drag and thrust. After doing this for all of the models and a bunch of frequencies, we found that stiffer sunfish swam faster.[189]

Recently the Navy’s biorobotic research efforts have been joined by those of another federal agency, DARPA, the Defense Advanced Research Projects Agency.[190] In addition to an ONR-like goal to improve naval technology, DARPA’s role in national defense ranges more broadly, across the services, as shown by their mission statement: “DARPA’s mission is to maintain the technological superiority of the U.S. military and prevent technological surprise from harming our national security by sponsoring revolutionary, high-payoff research bridging the gap between fundamental discoveries and their military use.”[191]

DARPA funds numerous robotics projects that use animals for inspiration. One such DARPA-funded project was led by Robert Full, professor of integrative biology and director of the Poly-Pedal Laboratory at the University of California at Berkeley. Bob, an expert in biomechanics and animal locomotion whom you may remember from Chapter 6, teamed with engineers at Boston Dynamics, Inc., the Illinois Institute of Technology, and the Robotics Institute at Carnegie Mellon University to build a six-legged robot inspired by insects. Called RiSE (Robots in Scansorial Environments), the 3.8-kilogram robot successfully scaled a three-story vertical concrete wall, untethered.[192]

RiSE is one of many successful robotics projects that DARPA has funded. In an effort to propel the development of autonomous battlefield vehicles, they created the DARPA Challenge, open to any team. In 2004 a robotic car, Stanley, the creation of Stanford and Volkswagen engineers, used its superior path planning, object detection/avoidance, and navigational skills to run the fastest time in a 138-mile course across the desert. Three years later, in the DARPA Urban Challenge, Carnegie Mellon University’s autonomous car, Boss, won the $2 million prize when it was the fastest robot to navigate a cityscape, filled with real pedestrians and human-driven cars, without violating the California driving laws. Although these well-publicized challenges featured automobiles, DARPA continues to fund research into animals and bio-inspiration.[193] Last time I checked, it looked like you could make pitches to DARPA for bio-inspired and robotics research to the officers running the following programs: “Biologically-driven Navigation (solicitation number DARPA-SN-11-07),” “Deep Sea Operations (DARPA-BAA-11-24),” and “All Source Positioning and Navigation (DARPA-BAA-11-14).”[194]

With all this potential federal funding for defense-related, bio-inspired robots in the United States, I guess it shouldn’t come as a surprise that other countries have robotic warfare programs. But I was flabbergasted, I admit, to learn just how many. According to robotics engineer Ron Arkin, professor of engineering and renowned roboticist at Georgia Technological Institute, fifty-six nations are developing robotic weapons.[195] And that forces an uncomfortable reality check: even though Arkin doesn’t discuss underwater robots in his book on robotic warfare, I too am working—at least indirectly—for the military.[196]

Growing up, I drove my mom crazy because not only did I play war incessantly with my buddies, Fritz and Karl von Valtier, but I also read about war. I read everything I could get my hands on, and I remember how impressed with myself I was when I had been through every book on war in the library of North Hills Elementary School. This was the 1970s, the middle of the Cold War with the Soviet Union, and after exhaustive scholarship, eleven-year-old John decided, to my mother’s horror, that nuclear weapons had taken all the fun out of war. The Cold War, with tanks and troops lined up just looking at each other, was boring—nothing happened. What would General Patton say, I asked? I just couldn’t see how we could be “at war” without any action.

My mother, who had marched against the Vietnam War, told me that if I got drafted, she’d shoot me in the foot. I’m not kidding. She said so repeatedly and passionately. During what my family called my “Spock phase,” I fought her passion with my deadpan version of Vulcan logic. I explained, without emotion, of course, that I wasn’t interested in going to war to die but rather to fight the bad guys and protect our country. When someone attacks your ship or your planet—I mean country—it was only logical to defend yourself. For some reason this line of argument failed to penetrate. I guess she was Bones to my Spock: illogical.

World War II, I told her, was my favorite war. She nearly died when I said that. How could her son—or any one for that matter—have a “favorite” war? “Well, mom,” I remember explaining, “World War II had action, on multiple fronts, with different enemies who used different strategies and tactics.” It was modern, with incredible technological developments in rocketry, aviation, and naval engineering. Best of all were the aircraft carriers and the strategic battles of the Pacific Theater! I bought books on carriers and carrier warfare. When I sat at the dinner table and explained to my older sister, Ann, that she needed to know that our pilots loved the Grumman F4F Wildcats, even though they were too slow to duel the Mitsubishi Zeroes head to head, because they had self-sealing fuel tanks and armored cockpits, I think that my mom got what I was really into: technology. Boys and their toys. She backed off—and didn’t mention shooting me in the foot again.

I forgot about World War II, and Spock, for good measure, thanks in no small part to junior high school and the rage of the hormones. Studying primate social behavior and mammalian reproductive biology became the pursuit of the neighborhood boys. On the latter subject, we found our school library to be woefully inadequate. Fortunately, both of Fritz and Karl’s parents were physicians, and the Drs. von Valtier kept their medical library at home well stocked and unlocked. We read widely and voraciously, appreciating the fact that anatomical texts were illustrated, and we thought it best not to trouble the good doctors with the knowledge of our library visits, even though I’m sure they would’ve been proud of the fact that we were inspired and independent scholars.

I was proud, for my part, of learning to drive. I impressed my instructor, and myself, when I was the only student in drivers’ education who could handle a manual transmission. My instructor, however, was already married, so I needed a different means of drawing the attention of the eligible girls in my class. Where facility with a stick shift failed, the opportunity provided by a license and the family car worked. I turned sixteen, got the keys, and fell in love.

A year later I was still in love, dating the same girl, and was now the owner of a 1169 cc, four-speed, three-door hatchback ’74 Honda Civic that my Uncle Pete had kindly saved for me in his garage. I also had a job, washing dishes at Mateo’s Pub, to support my bliss. With such happiness, imagine my surprise when, at age seventeen, I started thinking about war again, and this time for real. I wanted to be an officer in the Navy. It was simple. I wanted to go to the Naval Academy at Annapolis, get a four-year college degree there, and then, I don’t know, go fly Grumman F-14 Tomcats on the aircraft carrier USS CVA(N)-65 Enterprise.

Mom went, as they say in the military, ballistic. Still unaware of the irony, she again discussed the confluence of ordnance and my distal appendage as a deterrent to my service in the armed forces. But, I countered, this wasn’t the draft, the press gangs that she decried in years of old. That, and the Vietnam War, were long over. I was choosing to go into the Navy to pursue my education (and maybe to mess around in boats too). Then I’d have a degree, a steady job, and after twenty years as an officer I could retire and start my own marine biology business, like Jacques Cousteau. A man, a plan, a canal: Panama!

My working-man argument failed to gain purchase with Mom. Let’s sum her position up: war is bad because I would be trained to kill. I needed to switch tactics. Okay. Hmmm. Too bad I couldn’t find a way to be trained to save instead. Wait! The Coast Guard, which I immediately dubbed, for marketing purposes, the “humanitarian branch of the military.” Mission: search and rescue, drug interdiction, aids to navigation, and, my favorite, iceberg patrol.[197] Mom was on board.

The Coast Guard maneuver gave me some sea room, so to speak. I was allowed to take the admissions test for the Navy’s Nuclear Power School (passed) and applied for and won an NROTC (Navy Reserve Officer Training Corps) scholarship to the School of Engineering at the University of Michigan. Mission accomplished? Not quite. It turns out that as I learned more about the Coast Guard, and the Coast Guard Academy in particular, I decided I didn’t want Navy—I wanted Coast Guard. Only one problem: I didn’t get into the academy.

I was crushed. My boyhood dreams and now teenage ambitions were dashed upon the rocks of reality. Tormented by the Sirens, I resigned myself to become a naval officer through Michigan’s ROTC program. Mom would just have to deal with the Navy.

But she didn’t have to. About four weeks after I’d received the Academy’s letter stating that I was not a principal appointee, I received another saying that I had been named as an alternate, and a position had become available. Would I still be interested, they asked, in joining the “hard core about which the Navy forms?” By gum, an officer and a gentleman, I would be!

After high school graduation I sailed off, with Mom at the helm in her Chrysler K-car land cruiser, to the Coast Guard Academy in New London, Connecticut. Next day, my hair was shorn, I bravely kissed my mother good-bye, and then I joined the military. But not for long. After the six-week boot camp, I found a way to get out, honorably. And I was glad to do so. But not for the reasons you might think.

I didn’t mind the endless push-ups. I loved the seamanship training and messing around on boats and ships. I hated having to learn how to waltz. (I’m not kidding. Remember: an officer and a gentleman.) The food was okay, but I wasn’t allowed to look at it (I’m not kidding about that, either). Even being hazed as a lowly “swabby” by the upperclassmen was tolerable. For example, by proclamation in our cadet handbook, we were commanded to know the correct time at all times. But it also stated that the correct time was impossible to know. With this paradox in mind, I was trained to spit out, rapid fire, the one and only proper response to the simple question, What time is it, mister? “Sir, I’m greatly embarrassed and deeply humiliated that due to unforeseen circumstances over which I have no control, the inner workings and hidden mechanisms of my chronometer are in such great inaccord with the sidereal motions by which time is generally reckoned that I cannot with any degree of accuracy state the correct time, sir. Sir, I can state without fear of being too far in error that the approximate time is [glance at watch and state the time in military time].”

With all of this fun playing war, what would cause me to abandon ship? Ronald Reagan. I entered the Academy in 1982 as one of 232 cadets, sir. What I learned, after swearing in, was that my class, the class of 1986, was smaller, by exactly 100 cadets, than the class of 1985. The Coast Guard, for historical and strategic reasons under the Department of Transportation rather than the Department of Defense (see: humanitarian branch of the military!), was subject to drastic budget cuts under Reaganomics. A survey was conducted amongst the swabbies regarding preference for majors. Not interested in marine engineering, ocean engineering, mechanical engineering, electrical engineering, physics, or prelaw for any of my three choices, I thrice checked, with a regulation number-two pencil, the small square boxes of my destiny: marine science. The next day it was announced: the Academy would be cutting the major in marine science. Hit! You sunk my battleship!

Whereas my inner child wished for exciting wars and my young adult pined for a seafaring career, my middle-aged adult, looking back, is mortified with both desires. How easy it was to romanticize every aspect of war when you aren’t in it. It wasn’t until I transferred to the College of the Atlantic (COA), where I earned my undergraduate degree, that friends and mentors helped me challenge my implicit acceptance of the Roman poet Horace’s call to battle: Dulce et decorum est pro patria mori (“To die for the fatherland is a sweet thing and becoming”[198]).

One mentor was Dr. Ted Grand, an expert in the functional anatomy of mammals at the Division of Zoological Research at the National Zoological Park (NZP) in Washington, DC. His colleague and my COA adviser, Sentiel, a.k.a. “Butch,” Rommel, had persuaded him to take me on for an internship. Butch, a former naval officer and expert in navigation of vessels and people, had course corrected my life by helping me reckon, after leaving the Coast Guard, the heading back to the ocean. At the water’s edge on Mount Desert Island, Maine, Butch had me studying the biomechanics of marine vertebrates. Apprenticing at the NZP, reasoned Butch, was staying the biological course because Ted would help me improve my skill as a functional anatomist and my ability to ask and answer evolutionary questions with a quantitative approach.

In the beginning what neither Butch nor I appreciated was that each time Ted led me through the dissection of an animal, he was also guiding me on multiple levels, challenging me to make my implicit assumptions explicit and justify my positions. I don’t recall if it was over a wildebeest or a hippo, but he quickly learned of my interest in war and my background in the Coast Guard. I’m sure he must have understood my romantic/heroic naïveté, because he quickly took me to several used book stores to expand my military library. I soon learned that my elementary school’s reading list hadn’t included the likes of Wilfred Owen, the British soldier and poet who wrote from the World War I trenches of a fellow infantryman gassed by a German shell:

I hadn’t read Michael Herr’s “Dispatches,” a first-person account of Vietnam from what we would call, today, an embedded reporter:

These days, Ted points out that although I have left behind the facile romanticism of war as the hero’s journey, what really appears to have me bothered, he says, is the secretiveness that is essential to waging war. He might be right. I hate telling and keeping secrets, as my children will tell you, because someone always gets hurt when you withhold information. That parental stance informs my professional life too.

In the summer of 1999 Rob and I, at ONR’s request, went to the Unmanned Untethered Submersible Technology (UUST) conference hosted biannually by the Autonomous Undersea Systems Institute. With us was Peter Czuwala, the engineer working in my lab, who presented on analytical models of swimming fish that he and our students, Craig Blanchette and Stephanie Varga, had spearheaded. Any moral concerns that our students voiced about working for the Navy I addressed by explaining that all of our work was available in the public domain.[201] I was proud, I said, to have the Department of Defense supporting nonproprietary work on fish. No secrets.[202]

For his part, Rob’s line in the sand was making weapons. Toward the end of this UUST meeting, our program officer, the person running the ONR’s Bioengineering Program and overseeing our research grant, sat all of her grantees down in a room. We had heard rumors of the pressure that she and other ONR administrators were getting from the admiralty to justify all of this academic, basic research. “In future research proposals to ONR,” she said, “we want to see explicit reference to how your work will help us make better weapons-delivery platforms.” Rob and I looked at each other and said, “We’re done.”

But we’re not done. As long as we work on fish and robotic fish, even if we publish openly, we are part of a new arms race, a race among fifty-six countries working to weaponize robots.[203] Given that everyone is building robots for war, we’d be wise to heed DARPA’s mission: “prevent technological surprise from harming our national security.” One way to avoid surprise is to consider the obvious. By “obvious” I mean all that is available to you and me, those of us without special clearances, the not-secret information that, when you look at it from a different perspective, may allow you to guess about what is happening in secret. We’ll take a look from our new perspective of evolving robots.

Although most of the military robotic systems I know about appear to be remotely controlled, with a human in the control loop, some are semiautonomous. It won’t be long before fully autonomous robots are in operation because they can operate faster and more accurately than humans.[204] Their improved performance on the battlefield will drive innovation in that direction. Speed kills.

After that the logical direction, as I see it, is to move toward robots adapting their behavior as the battle wages. Behavioral adaptation is what makes rag-tag rebels so hard to beat in a protracted war. As part of the rebel alliance, you may be outmanned and outgunned, but every enemy has a weakness; if you can figure it out and take advantage, then you have a chance. For example, the US military is vulnerable to attack from improvised explosive devices, simple but deadly weapons that disrupt vehicular patrols in Iraq and fuel delivery in Afghanistan. The fastest way to adapt is through learning, and if you are a robot this means getting feedback about your performance that changes your onboard software. Behavioral adaptation is already well established in robots, with multiple methods for learning. One such adaptive learning algorithm is called an “adaptive neural control chaos circuit,” invented by Poramate Manoonpong, professor of physics at Gottingen University in Germany, and his colleagues for rapid and reversible learning in changing environmental conditions.[205]

But learning or evolving software only takes you so far. The next step will be to have adaptation of the body, the hardware itself, by setting up selection to act on a population of Evolvabots. “Hod Lipson is the state-of-the-art for mechanical adaptation,” explained former student and now colleague Josh de Leeuw. I agree. Hod, associate professor of many things (engineering, computer science, robotics) at Cornell University, takes a bio-inspired approach in order to engineer machines that can build other machines. Hod’s lab has designed and built embodied robots automatically from digitally evolved scenarios and, along with his colleague Josh Bongard, assistant professor in computer science at the University of Vermont, created robots that sense self-injury and respond by altering their body and behavior.[206] Bongard takes an approach that he calls “artificial ontogeny,” allowing a robot to combine learning over its lifetime while it is embedded as an individual in a population of evolving robots.[207]

At first, the idea of machine-making machines—replicating themselves, reproducing novel offspring, or making offspring for someone else—may seem far-fetched. But as Adam Lammert, one of the inventors of Tadro (see Chapter 3), pointed out to me recently, self-replicating machines were shown to be feasible as early as 1957.[208] Lionel and Roger Penrose, at the University College, London, demonstrated that recognition systems and subunits needed for replication could be built into the body of a physically embodied model organism. These models, made of plywood, were “creatures” with levers that prevented joining other creatures unless the proper mechanical signature was present. Once conjoined, two creatures, or a creature and subunits, could undergo fission to create two replicants.[209] Today Hod uses rapid-prototypers, machines that create three-dimensional parts of nearly any size and shape from software instructions.

Thus, the final step in making military robots will be to have them make their own robotic offspring as part of evolution on the battlefield. Robots will evolve their brains and bodies in response to the dynamic chaos of the moment in order to carry out their longer-term mission. Evolutionary adaptations will occur in a population of military Evolvabots with feedback from the battlefield environment using a fitness function the battlefield commanders generate. That fitness function might reward performance such as rate of target detection, rate of target hit, probability of survival, and robustness, or the ability to continue to operate once damaged.

When I presented this fitness idea to Chuck Pell, he disagreed: “The basic principle for robots in war should be this: [robots should be] unmanned, expendable, and cause maximum damage.” When I pointed out that a simple fitness function of maximum damage doesn’t leave room for evolving very complicated robots, he said that was the point. Chuck’s perspective is that complicated robots are expensive, and that’s a problem. With expensive robots, he explained, commanders won’t want to lose them, and those in charge will alter their tactics accordingly. The inescapable consequence, which I’ll call Pell’s Principle, “is that robot capabilities are proportional to cost, and cost is inversely proportional to expendability.” The logical conclusion of Pell’s Principle is to build and use swarms of simple, small, expendable robots (Figure 8.2).

“You can’t stop all the little robots,” said Chuck, “robots like MicroHunter.”[210] MicroHunter was a microAUV (Autonomous Underwater Vehicle) built using the same cycloptic helical klinotaxis system that we used to build Tadro (Figure 8.3; MicroHunter was introduced in Chapter 4). But whereas Tadro is a surface swimmer, MicroHunter swims underwater, spiraling its way to a light source from anywhere you could put it in an Olympic-sized pool. MicroHunter, funded by a DARPA contract to Chuck as an employee of Nekton Research (see Chapter 7), and Hugh Crenshaw, then professor at Duke University, caught their program officer’s eye when the duo reported that it had a 100 percent success rate finding the target.[211] “Nothing is ever that good,” said Chuck, “so they sent statisticians and a former Navy SEAL to investigate.”

SEALs, the US Navy’s special operations experts, are world renowned for their abilities underwater.[212] So even though Chuck and Hugh knew that MicroHunter could hit the three-meter light target when unchallenged, they figured that a group of four MicroHunters, swimming slowly, wouldn’t stand a chance with a special ops SCUBA diver in the pool. To everyone’s surprise and the SEAL’s chagrin, the MicroHunters and the SEAL played to a draw, with the MicroHunters hitting the target 50 percent of the time in six three-minute trials. That’s seriously excellent performance for a piece of embodied intelligence with but a single sensor and a single degree of freedom on the motor output side. Now try to imagine defending against not just four MicroHunters but a swarm of fifty. Your only defense may be evolution.

We’ve had enough theory and practice of evolution in action in this book that I’m guessing you’ll be able to guess what I’m about to say. Here we go: any military that evolves robots on the battlefield will likely do so using the following principles:

* Robots are expendable. Given Pell’s Principle (see page 223), the only way to get large numbers of robots on the battlefield is to have them be expendable, and that usually means inexpensive. Large numbers also ensure that sufficient variation is in place to allow selection to act on the population. Large numbers also serve the tactical advantage just explained with regards to overwhelming the adversary with a swarm of simple autonomous agents.

* Robots are simple. This, too, follows from Pell’s Principle. The way to make robots expendable is to make them simple. Simple also usually means inexpensive to make and quick to produce. Employ the KISS principle in your design. Find the minimum brain, body, and behavior needed to seed your population. Choose which characters evolve.

* Robots evolve quickly. Given that generation time is one factor that limits the rate of evolution, make the generation time short. Short generation time will minimize the response time between a change on the battlefield and the change in the behavior and hardware of the population of robots.

* Robots evolve in small cohorts (small and genetically isolated subpopulations). Given that evolutionary change occurs rapidly in small, isolated populations, create many small companies of robots rather than a single large battalion. Keep in mind that random influences will dominate if the population is too small. Note that this may, at first, seem to run counter to principle 1 and using large numbers. You can have multiple populations or companies in simultaneous operation.

* Robots diversify in generational time. Given that evolutionary change will be rapid with principles (3) and (4), then allow the companies of robots to speciate, to evolve along different evolutionary trajectories. Diversification will allow more and different kinds of adaptation to occur simultaneously, thus increasing the chances of both tracking changes in the environment and finding the best solution at a given time and place.

That’s the offensive view of military Evolvabots. What about defense? How would you stop an army, navy, or air force of evolving robots? Keep in mind that even if it’s your militia of Evolvabots, you will need ways to constrain and control them too. This is starting to sound like nearly every story and movie on robots in which they rise up and break the shackles of their creators to take over the world. Although we aren’t making a movie, let’s run with the plot line anyway. To avoid military defeat or robotic overthrow, here’s what you do:

* Limit initial production prior to battle. Control numbers and types of robots. Constrain raw materials. Limit, reduce, or eliminate energy sources. If the population is yours, you may want to design in hard production or run-time limits to avoid the enemy co-opting the group.

* Limit reproduction on the battlefield. Because reproduction is key to the evolutionary process and usually involves some vulnerable moments, like finding mates and creating offspring, you should target these situations. Also, target the machine-making machines, because they need to be on the battlefield to keep generation times short. Limits to production (see item #1, above) can also be employed on the battlefield by cutting supply lines.

* Limit repair on the battlefield. Injury provides another vulnerable situation. If robots are self-repairing, their function will be impaired. Capture low-performance agents preferentially. If the injured robots are being repaired by other agents, target the repair teams.

* Evolve predatory robots. If you or the enemy employ Pell’s Principle, you’ll need to be prepared to capture or destroy swarms. For starters, you’ll need to let your evolving predators have the capacity and capability of filter feeders like baleen whales. Consider behavioral adaptation first in your predators because the shorter generation time of the prey will limit opportunities for hardware evolution in the predators.

* Make complicated robots. That’s right. Want to control your own robots? Make them complicated. Because complicated usually means expensive, you are likely to have only a few of them. You will also hesitate to send them into harm’s way, as Chuck was predicting. Furthermore, complicated robots will never take over because the laws of probability virtually guarantee their failure. If every component has, let’s say, a 99 percent chance of not failing on a given mission, that sounds pretty good, right? But what if you have two such components in your robot? By the law of independent probabilities, we take the product of the two: 0.99 x 0.99 = 0.98. Not bad. A 98 percent chance of the system, composed of those two components, not failing. Now give your system two thousand components, not unreasonable for some of the more sophisticated fish robots I’ve seen. That’s 0.99²⁰⁰⁰ = 0.0000000002. You’ve got no chance—your robot will fail! The way we keep complicated machines like airliners and space shuttles in business is by building components with lower failure rates (0.99999), engineering redundancy into the machine’s critical systems, and inspecting and replacing parts before they fail. Bottom line: to ensure control of your Evolvabots, make them out of many crappy components.

My Hollywood alarmism about evolving robots on the battlefield may have you thinking that all of this warfare stuff is just fantasy nonsense. Maybe it is. But let’s pretend that you are an admiral and you have to make that call: are military Evolvabots nonsense or good sense? Will some other military surprise us with Evolvabots in battle? This matters because you, as admiral, need to make practical decisions that have long-term consequences. Do you put your limited resources into an offensive Evolvabot development project? Or do you put resources into Evolvabot defensive countermeasures? Keep in mind that if you do use resources on Evolvabots, you have to cut the budget of other tactical systems. How can you be sure that Evolvabots will ever be a serious risk and worth the cost of development or countermeasures?

You can take DARPA’s tack and examine feasibility. Presented with the idea of commanding a fleet of evolving robotic fish, for example, you might want to assess one of the most important aspects of any battle: communication. If no one can figure out how to communicate with a swarm of underwater robots and adjust plans as the battle commences, then you probably don’t have much to worry about.

Communication before and during battle is paramount for the simple reason that, in the words of the Helmuth von Moltke the Elder, chief of staff of the Prussian Army for thirty years, “No plan survives contact with the enemy.” Any battlefield is chaotic swarm intelligence in action. For a battle plan to adapt, each agent has to know the purpose of the mission, understand their part in it, have the ability to communicate on the battlefield to update their tactical knowledge of the enemy, coordinate with other agents and adjacent units about their positions and disposition, and make decisions quickly as information deteriorates and change accelerates.

The first element of communication and decision making on the battlefield starts prior to engagement, and it’s called the commander’s intent:

In the best case the commander’s intent is known and understood by all sailors or soldiers so that as the plan deteriorates in battle, individuals can use adaptive behavior to advance the mission. “The commander’s intent,” suggests Chuck, “should be embodied, embedded, in the warriors.” That means that design of the military Evolvabots has to involve Command and Control during design because intelligence and intent, as we’ve seen throughout this book, are part of not just the programmable nervous system but also the type, arrangement, and quality of sensors, motors, and chassis.

After talking to Chuck I was wondering if the commander’s intent (CI) itself could serve as the fitness function for military Evolvabots. Whatever the CI—cause maximum damage to target X or guard squadron Y or rescue fleet Z—the ongoing performance of individual Evolvabots can be judged relative to it. Because the performance of each individual is compared to that of others in its population, the feedback about what works is relevant automatically. It turns out that engineers working for the US Army have already tried, in digital simulation, the idea of using the CI as the fitness function: “Evolution continues until the system produces a final population of high-performance plans which achieve the commander’s intent for the mission.”[214] Did you get that? Tactical plans, which are extremely complicated themselves, can be evolved using genetic algorithms that use the CI as the fitness function.

The fundamental communication issue on the battlefield is that small groups and isolated individuals have to make decisions without checking with their commanders. As Lieutenant Colonel Lawrence G. Shattuck, professor, Behavior and Sciences and Leadership at the US Military Academy, West Point, has written, the pace of events on the battlefield often precludes direct contact with superiors even if communication channels are open.[215] Once communication ceases, for whatever reason, soldiers need to know the CI to help frame their decisions, getting inside the commander’s head to know how she would be making the decision, according to Shattuck.

The central technical challenge is this: get autonomous robots working, communicating, self-repairing, reproducing, and evolving in the wild, without help from humans. Proximal challenges, in addition to the ones already discussed in this chapter, include the following:

* How do we embed the fitness function in a population of freely roaming Evolvabots?

* How do we have that fitness function, which is imposed by humans, be an automatic part of the world in which the Evolvabots are working?

* Or do we let the fitness function be unspecified but emerge from the survival of the robots in the world?

* In any of these scenarios, how do we monitor and control Evolvabots in the wild?

If these issues can be solved, then everything in this chapter is feasible.

But wait. Just because we can do something, does that mean that we want to—or should? The central ethical challenge, framed by Ronald Arkin in his research on the topic for the US Army Research Office, is this: get autonomous robots to behave “within the bounds prescribed by the Laws of War and Rules of Engagement.”[216] “The advent of autonomous robotics in the battlefield,” writes Arkin, “as with any new technology, is primarily concerned with Jus in Bello [acceptable limits to conduct in war], that is, defining what constitutes the ethical use of these systems during conflict, given military necessity.”[217] Arkin’s goal—which I support wholeheartedly—is to have our military robots outperform our human soldiers in terms of ethical conduct. Evolving robots need a conscience.

Predicting the future is even easier than understanding the past. That is the fundamental tenet (tenet number 1), as I see it, of the art movement created by Lloyd Dunn, known as Retrofuturism (Figure 8.4). As you’ve seen in this chapter, I’ve been able to predict, with virtually no data to support my arguments, a scenario in which we are at the beginning of a new kind of military arms race. Evolving robots, I claim, will alter the way we fight wars and defend ourselves. For retrofuturistic completeness I should also predict the exact opposite (tenet number 2), namely that evolving robots are a trivial sideshow in the growing field of robotics and have nothing to tell us about the future of warfare.

I don’t really think that the second prediction is true. Too bad. In all seriousness, this chapter has been a bummer, right? Who wants to talk about war and autonomous killing machines when we can talk about studying the evolution of the first vertebrates instead? I don’t. But the reality is that evolving robots are and will be created for academic, industrial, and military purposes. This means that we should all become students of robots of any kind, whether they be evolving robots, nonevolving autonomous robots, or semiautonomous and remotely controlled military robots. We need to understand robots so we can proceed with due caution and deliberation. No secrets. No surprises.

Now for an apology. In this book I’ve covered just a tiny sliver of the world of robotics: evolving robots. And I haven’t even done that little bit justice. I’m sorry. For example, I’ve talked mostly about the work done by myself and my collaborators, referring just occasionally and superficially to the great researchers who have inspired us: Ronald Arkin, who is creating the field of robot ethics after unifying behavior-based robotics together; Barbara Webb, who created biorobotics; Stefano Nolfi and Dario Floreano, who created evolutionary robotics; and Valentino Braitenberg and Rodney Brooks, who cocreated the field of behavior-based robotics that jump-started all of the above. To help overcome my guilt for giving all these masters short shrift, I’ll tell you that they all have written great books on their subjects, and you should read them.

I should make a parallel apology for the world of evolutionary biology. With a head start of over a hundred years on robotics, evolutionary biology and my omissions of name are more difficult to characterize and recognize. I can tell you that I’ve largely ignored the fascinating world of EvoDevo, the evolution of ontogenies and the constraints and possibilities that developmental systems give to the species they construct. Sean Carroll is the place to start reading there. Vertebrate paleontologists like David Raup, Steven Stanley, Robert Carroll, and Michael Benton ought to feel slighted because they have carried the torch and blazed the trail with their excellent textbooks. The great biomechanicists, McNeil Alexander, Tom Daniel, John Gosline, Mark Denny, Paul Webb, Andy Biewener, get nary a mention. You have to leave out a lot, I’ve learned, when you write a book.

And finally, here’s a parting shot, a reminder of one of the concepts that I consider most important and most often misunderstood: evolution. I’ve been using the word “evolution” throughout this book not in its causal, on-the-street sense of directed, progressive, optimizing design but rather in its scientific sense: a change in a population of agents over generational time, in a given environment, caused by random genetic processes coupled with selection, where selection results from the interaction of individual agents with and within their physical environments.

We’ve talked about, designed, experimented with, and analyzed the behavior of two types of robots: (1) Evolvabots, which change in ways that we can’t predict with certainty when they constitute a population with genetic inheritance and selection pressure; (2) Evolutionary Trekkers, which we create to be of a certain form in order to help us understand how extinct or never-existing forms may have behaved in a given environment. Both Evolvabots and ETs are built to test ideas about evolution as a process (how things can evolve), specific evolutionary events (selection pressures), or specific evolutionary situations (big, four-flippered animals).

Although I’ve focused on fish and aquatic vertebrates in this book, just keep in mind that you can use Evolvabots and ETs to model any kind of critter.

So long, and thanks for all the robotic fish.