Self-Replicators
Arguably the essence of life is its capacity for self-renewal and self-propagation. Attempts to imitate this capacity with silicon and metal represent one of the most ambitious areas of robotics. Useful self-replicating robots would represent not only a godlike achievement in an abstract, philosophical sense; they also would be extremely useful, transforming currently unfeasible giga-tasks into relatively easy ones. Need to dig an undersea tunnel from Brazil to Africa? Or construct a 30,000-mile-high space elevator? Or mine the ample minerals available on the moon? Or “terraform” Mars? In principle, robots that can make copies of themselves, using relatively simple parts-or even simple raw materials such as sand and iron ore-would be able to do such things. Humans would need only to get the process started by supplying one or two such bots, which would proliferate and do the rest.
Hopkins mechanical engineering Professor Gregory Chirikjian ’88 has been one of the pioneers in this field. In 2003, he and two students published the first description of an autonomous self-replicating robot: a several-feet-wide thing that could make functioning copies of itself-one copy every 135 seconds-from Lego building blocks, magnets, and simple optical sensors. Since then, Chirikjian, who is also the editor of the influential journal Robotica, has been overseeing the design and construction of larger and more elaborate self-replicators, which can assemble copies of themselves out of simpler materials.
His current opus, he says, “occupies a large part of my small laboratory.” About 10 feet long and 3 feet high, it looks a bit like a scaled-down shipyard gantry crane. It includes tracks, two towers, and a grasping system that can pick up parts and place them accurately in three dimensions.
Principally constructed by recent PhD student Matt Moses, and undergraduate Hans Ma ’12, this ambitious device is essentially a learning tool. Engineers use such machines to better understand the key principles of useful robotic self-replication- so that they know where they’re going and how far they’ve come. These principles include versatility (the range of tasks that such a bot can perform), relative complexity (the ratio of the finished bot’s complexity to the complexity of its parts), robustness (tolerance to errors it makes), and efficiency (how fast and how cheaply it can reproduce). The ideal self-replicating robot would be highly versatile, made up of relatively simple parts and materials, highly error-tolerant-and, of course, much more efficient than humans, who need two complementary specimens to get the reproduction process going, and even then take at least a couple of decades to produce new working versions (at great expense, if these new versions are to be of much use).
The dream-or nightmare, if you prefer-of self-replicating robots that draw on soil and water and sunlight for sustenance, and breed faster than insects (a robot that reproduces once per hour could go from one copy to 17 million copies within a day), is still decades off. But Chirikjian and his colleagues already have made surprisingly long strides toward this goal. “The robots we’re making now can manufacture some of their constituent parts by squirting liquids into molds, having the parts form, and then picking up the parts and assembling them into complex structures,” he says.
In the near term, he sees self-replicator bots contributing in small but important ways, such as the self-repair of buildings or bridges. Farther out he sees at least one giga-scale goal as doable. In a 2002 paper in the IEEE/ASME Transactions on Mechatronics, Chirikjian and two students, Yu Zhou and Jackrit Suthakorn, described in detail a bot-based lunar mining scheme. “When self-replicating robotic factories take hold,” they wrote, “the moon will be transformed into an industrial dynamo. The resulting refined materials and energy that will be produced on the moon will then provide capabilities for the exploration and colonization of space that could never exist otherwise.”
Little Self-Assemblers
Bots that self-assemble down at nanoscale-as DNA and some proteins do-could be at least as useful. In the laboratory of Rebecca Schulman, assistant professor of chemical and biomolecular engineering, these nanobots don’t just take their inspiration from DNA, they are made from DNA-synthetic DNA strands that have been strung together like sausage links from their constituent nucleotide molecules. “In this context, we think of DNA not as parts of living things but as parts of useful things to build with,” says Schulman.
One of the key properties of single-stranded DNA is its ability to bind to a counterpart, “complementary” strand. Cytosine (C) bases on a DNA strand are complementary to guanines (G), and adenines (A) are complementary to thymines (T); thus a strand with the code CGAT would bind tightly to any GCTA strand. It’s a simple evolutionary trick that enables DNA-based organisms to replicate efficiently and with high fidelity. But for Schulman’s purposes, DNA’s tunable complementarity makes it a versatile nanoscale Velcro.
The basic idea here is to use scaffold structures of double-stranded DNA that have sticky, naked (single-stranded) DNA segments in certain places, for example at each end. The basic structures of these DNA “tiles” can be simple, two-ended lengths of DNA, folded-over triangles with three sticky segments, four-ended crosses, and so forth. The right mix of these structures and their sticky bits-when they’re all placed in water and stirred-will act as a sort of self-assembling jigsaw puzzle and arrange itself into a pre-designed shape. “In this way, we can assemble complicated structures with just a few kinds of simple building blocks,” says Schulman. “And in principle, these structures will reassemble again even if torn in half, provided that their building blocks are available.”
Like Chirikjian’s bulkier self-replicobots, these DNA-based structures so far are mainly tools for learning more about the processes of self-assembly. But another of Schulman’s projects is closer to finding real-world use. In this case, instead of trying to make complex structures with DNA-based elements, she and her students are using them to self-assemble simple wirelike strands that connect two fixed points, for example on a circuit board. The strands could then become a scaffold for attaching electric-current-conducting metal atoms or other useful elements. “This would be a way, for example, to self-assemble circuits that are smaller than can be made with current computer-chip lithography techniques,” Schulman says. “Or we could use it to connect things that we don’t otherwise know how to connect; all we’d have to do is put DNA-bindable labels on the start and end points, and add the building blocks.”