Watch This: Squishy Robots Evolve for Speed

By Breanna Draxler | April 24, 2013 11:08 am

With names like Incher, Jitter and Wings, you know these aren’t your run-of-the-mill robots. Through computer-simulated evolution these flexible robots have developed unique gaits. The result is a new source of ideas for how robots could move—and a bizarre but irresistible visual.

Researchers at Cornell University and the University of Wyoming wanted to see evolution in action, so they designed a soft robot simulator. They provided the computer program with four materials and one rule. The materials resembled the basic components of our own bodies: bone, soft tissue and a couple kinds of muscle. The only rule governing the system was that faster creatures would reproduce more. Essentially, the researchers incentivized forward motion, so the faster the robot, the more successful it would be in the evolutionary race.

Then they sat back and let natural selection do the rest.

Over the course of 1,000 generations, this simple set of components yielded some truly kooky concoctions. The robots jiggle, flap and jump at various speeds. They look ridiculous, and may seem counterintuitive, but these squishy robots get the job done. In fact, for comparison, the researchers put scientists to the test, to see if they could design a better robot from scratch, using the same parameters as the simulator. Not a single participant (engineer or not) was able to produce soft robots that outperformed those evolved in the computer simulation.

Evolution, it seems, takes the cake yet again.

By combining evolutionary algorithms with biological materials, the researchers say soft robotics will go beyond the limits of human intuition and engineering to produce the complex artificial life forms of the future.

Watch the wiggly galloping robots strut their stuff in the video below.

Video courtesy of Cornell University and the University of Wyoming.

CATEGORIZED UNDER: Technology, top posts
  • TexCIS

    Only ONE rule? Baloney. I’m a computer programmer, and I guarantee you there are thousands of lines of code behind that simulation. Some obvious ones (for non-programmers): 4 rules for what types of materials, 2-3 rules for each of them, i.e. muscle = contract, expand and sequence. Tissue = support + type of support. Another rule: each type of tissue has to start as a square.

    This is pure intelligent design. Your “evolution” is a baloney sandwich.

    • Elias Chan-Sui

      Would you have preferred the words ‘overriding’ and ‘permanent’ instead of one and only?

    • Ronny Hugo Hansen Warelius

      I take you seriously in direct proportion to the amount of energy your brain spent before arriving at your conclusion. if you think about the problem “How could self-replicating molecules come from the 118 elements on the periodic table?” for 6 hours, 365 days a year, for 60 years, your brain would have spent 11.5 gigajoules thinking about the problem. If I put 11.5 gigajoules into the brains of whales or monkeys or dolphins or ANY known brain not human, in the form of food, we don’t get the correct theory of the universe. What then, makes you think otherwise about the human brain, other than you being a human brain that don’t want to feel badly about your own abilities?

      • Andrew Kiener

        “Energy expended” doesn’t equal “quality of output”, as this post shows.

        • Ronny Hugo Hansen Warelius

          If you thought for a decade about what comment to write, then you would have said something half-clever, instead of something utterly predictable. I say he is wrong (and imply everyone is wrong), you assume I am also, wrong. Brilliant piece of intellectual work.

    • Scott McKie

      Only one rule for the genetic algorithm’s fitness function. That’s all they were referring to – not the whole simulation.

    • Andrew Kiener

      For a programmer, you have surprisingly little understanding of how evo simulations work.

  • Guest

    I agree, shame on Discover for trying to promote EVILUTION with this horsefeathers and hogwarsh. It’s clear these fancy pants Cornell folks are simply trying to destroy the fabric of small town America with their ungodly jello cube monsters. But seriously, I wonder

    • Guest

      Obvious troll is obvious.

  • Ronny Hugo Hansen Warelius

    They look no more clumsy than humans.

  • Bobareeno

    The minute you introduce a “reward,” you influence the direction of the development and give the implication that “speed” has a desirable purpose. The programmers guided the results to achieve what THEY wanted. A more amazing development would have been if the program wrote itself.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!


Briefing you on the must-know news and trending topics in science and technology today.

See More

Collapse bottom bar