Robots with stomachs

Fred Hapgood

Popular articles on robotics usually focus on high tech features like manipulators or vision or bipedality, but the people who actually have to build them worry a lot more about power.  There are very few attractive options for machines that are supposed to run around untethered, especially if they have to do so outdoors.  Batteries are heavy, solar power arrays are cumbersome, internal combustion engines are noisy and need to be refueled, and electrical outlets are few and far between.

This fact of life explains the unusual chain of associations Chris Campbell catalyzed with his exploding beer bottles.  In 1996 Campbell was a graduate student in Mechanical Engineering at the University of South Florida and a enthusiastic amateur brewmaster.  One day he forget to vent some of his bottles properly and they blew up.  Later that day he mentioned his experience to his professor, a roboticist named Stuart Wilkinson.  "Aha!", Wilkinson recalls thinking.

What Wilkinson saw was that brewery energetics could be applied to solve the robot power problem.  You could feed sugars to yeast, producing carbon dioxide (and alcohol), and then use the pressure of the CO2 to turn wheels, moving the robot along.  "We called it the 'flatulence engine'," Wilkinson says, referring to a comparable process that takes place during human digestion.

The point and virtue of the flatulence engine was that sugar exists naturally, in all vegetation, and vegetation is abundant, accessible, free, and available 24 hours a day, rain or shine (at least outside of cities).  A robot that could keep its fermenter fed and in good health would have no more power worries than a goat.  One obvious application would be a whole suite of gardening robots -- mowers, trimmers, weeders -- that would draw their power from digesting their own work product.  Farmers might become entirely energy-independent, running all their machinery directly from a small fraction of their crops.  Mobile sensor platforms could be dispatched to roam through wilderness areas, gathering data for forest management, environmental monitoring, and ecological research.  There is a long line of people who might be interested in ocean-going machines that could feed themselves by filtering algae or plankton from the water, like baleen whales, or that could sit on the ocean bottom and live like clams.  Military applications would include robot snipers -- guns that live in the woods -- and automated reconnaissance patrolling.  Free-ranging robots could be used to watch national borders and fight poachers.  A bit further out, the machines might become the antibodies of an ecological immune system -- equipped to detect and eat the ecological invaders out of favor this season, like kudzu or Asian cheatgrass.  Far from least, artists building garden sculptures and outdoor installations would be all over the technology. 

Eventually Campbell and Wilkinson were forced to conclude that while the flatulence engine was a breakthrough conceptually, it was not really practical.  A useful robot would almost certainly have to have computers, a radio, navigation and warning lights, probably cameras and possibly many other types of sensors.  All these devices run off electricity, and that the flatulence engine, being mechanical, could not provide.  (Mechanical energy can be converted to electrical, but the inefficiencies involved were depressing. )

Still, the idea focussed his research, and soon Wilkinson stumbled on the work of a crusading British electrochemist named H. Peter Bennetto.  Bennetto, now retired, had worked at King's College in London on means of bringing electrification to the third world that were environmentally sound and did not lead to dependence on oil imports.  As part of this work he had become an authority on biologically-produced electricity.  (Ultimately, all living things run off electricity.) Organisms generate their electrical power chemically, by making electrically charged atoms, or ions, inside cells.  These ions then organize electron flows throughout the cell, bringing power to wherever it is needed. 

Bennetto discovered a chemical that could penetrate cell walls, absorb some of these electrons, and then exit.  Once outside the cell, the chemicals could be stimulated to shed the diverted electrons, creating an electrical gradient or potential.  He also found that many species of bacteria can lose up to 80% of the electrons they make without suffering any obvious ill effects.  These facts combined into a grand vision of networks of sustainable power plants distributed throughout the emerging nations, harvesting electricity from bacterial cultures fed with local vegetation.  His work comes close to setting a record for the greenest possible form of power generation. 

While the microbial fuel cell (as the device is called) has not drawn much interest from the third world power sector, it definitely rang a bell in South Florida.  The discovery of Bennetto's research made Wilkinson confident enough to commit his career to the study and design of gastrobots -- robots with stomachs.  Today Wilkinson has a simplified evaluation device -- basically a train with three 7" x 11" cars that is fed on refined sugar-- running around a track in his lab.  He calls it the 'gastronome'.  (The Gastrobotics Institute is funded by the Tampa Electric Company, apparently more forward thinking than most utilities.)

The engineer says he gets a flow of mail from citizens worried that he is throwing away our leverage over robots, handing them the freedom they need to become the dominant species.  It's bad enough that they will be able to feed themselves; suppose someday they develop a taste for meat? The energy content of animal protein is far higher than that of vegetable protein.  (There actually is a team in Reading, England, building a gastrobot that will live on animal protein, specifically garden pests.  They call their machine the 'slugbot'.) Suppose gastrobots took over the factories -- which would no doubt be automated -- that make them? Where would we be then? Driven into the dark corners of the world by tireless, implacable, flesh- eating machines.

There is no question that there is at least one really bad movie to be made about this idea, but there is this consolation: the gastronome is not going outdoors any time soon.  Gastrobotics is a brand new idea in mechanical engineering, and it raises a long list of issues.  An internal environment has to be designed that will keep the power microbes happy for months on end.  The machines will need the ability to distinguish between foods that will keep their microbes happy and that will kill them.  The personal sanitation question is very complicated (how do you clean the cleaning surfaces?) but essential, since a little dirt or stray fiber in the wrong place can waste lots of energy.  Wilkinson has discovered that bacterial colonies yield 20% more power if they are fed continuously rather than by eating at different times during the day, which suggests that gastrobots are going to need precise senses of hunger and satiation.  Ways must be found to prevent the electron transporting chemical from being excreted when the gastrobot relieves itself, which means that the machines will need a kidney.

In fact the whole problem of waste is quite involved, even on the level of environmental policy.  Chemically speaking robot excrement would be biological, but it would be from a robot, and people might think of it differently than waste from a bird or a rabbit.  Perhaps it should have a special signature shape or color or odor, so everyone would knew that a robot was near.  Or perhaps not.  In any event at some point somebody is going to have sit down and give some hard thought to the question of the proper design of robot poop, which shows you just how new this whole territory can be. 

Still, it does seem inevitable.  The gastrobot combines two strong trends in modern technology: finding ways to harvest power from local sources (solar, wind, geothermal) and borrowing design ideas from biology.  Today engineering institutes are building submarines that swim like fish, airplanes that get about by flapping their wings, and land vehicles that move by walking.  The two most promising approaches in artificial intelligence are neural nets, which were originally based on explicit analogies to the brain, and genetic algorithms, inspired by sexual recombination.  The nature of progress in robotics and automation can be understood as de- domesticating machines; allowing them to operate on their own, eventually even to have original ideas.  Gastrobots are another illustration of this deep trend towards blurring the ancient distinction between biology and technology.  Perhaps someday we will have wild machines roaming around in the ecology, self-sufficient, fully autonomous, answerable to no master but themselves and fate.