How robots can teach us about animal evolution

Download Flash Player to view this content.


First aired on Quirks & Quarks (28/4/12)

What's the difference between animals and robots? For starters, how they are made. Animals are made of water, fat, bone and protein, while robots are made of polymers and metals. They are also built differently. Animals reproduce themselves and grow and develop, while robots are assembled from parts in laboratories. But, arguably, the biggest difference is how they evolve. Animals are a product of evolution, the generations-long process of mutation, variation and natural selection. Robots are a product of engineering, drawings, computer programs and sophisticated manufacturing. However, if Dr. John Long, a professor of biology and cognitive neuroscience at Vassar College in New York State, has his way, this may change. Dr. Long's work is in evolving robots. But don't worry, he's not trying to create a new breed of robots that will take over the world. He's attempting to model the early evolution of vertebrates using simple robots, and he chronicles this process in his new book, Darwin's Devices: What Evolving Robots Can Teach Us About the History of Life and the Future of Technology.

Long turned to robots after running out of traditional tools that biologists use to study evolution. Fossils of the first vertebrates, which are 500 million years old, offered insight into what the animals looked like, but offered no clues as to how they behaved. "You need behaviour and you need those ecological interactions to understand the process of evolution," Long explained to Quirks & Quarks host Bob McDonald in a recent interview. So he did what many scientists before him have done: he turned to engineers, applying their basic premise, "if you can build it, you understand it" to biology.

This is important, because evolution is so difficult to model and understand. Even the evolution of the simplest species (or robot) depends on a variety of factors, the external environment, the random effects of mutation and mating, the process of natural selection and more. And, as Long points out, characteristics evolve independently. "They can evolve together sometimes and independently other times," he said. "Evolution is in many ways unpredictable. You're going to get something different every time you do it."

Long is specifically interested in the evolution of vertebrates -- a group which mammals, including humans, are part of -- and "what were the selection pressures that drove the evolution of these first vertebrates." Long believes that understanding this biological feature "is key to understanding our own evolutionary history." So he and his team built a group of simple "fish-like" robots, called Tadro3, with randomly varied physical attributes, such as the stiffness and length of their backbone, "just like mutation in real bodies." The team then placed the robots -- which are independently behaving and are not controlled by Long's team in any way -- in an environment modelled after the real environment the first vertebrate survived in 500 million years ago and then evaluated their success according to an established set of fitness criteria.

The experiment quickly proved a success and it yielded surprising results. Long's hypothesis was that the robots with stiffer backbones would be better swimmers, and that the better swimmers would be better at finding food. That's not what they discovered. Instead, robots with more flexible tales were better at getting food. This result is not only an important discovery for better understanding evolutionary processes, it validated Long's use of bio-robots in his research. "This is a powerful tool that we have here that we can refute evolutionary hypotheses."

Other, more complex, experiments followed, where Long introduced other elements to the robots' environment, such as predators. He'll continue with building bio-robots and modelling evolution and believes biology can learn a lot from engineering and vice versa. When a robot doesn't work as anticipated, the team can go back to their understanding of the animal's biology and of the robot's biomechanics to uncover the root of the problem."There's this wonderful iterative process going back and forth between biology and the engineering that drives both of those fields."

And if Long has his way, this is a relationship that will only continue to evolve.

Watch Long's evolving robots in action in the video below.