The first principle you learn about natural selection and evolution in general is “survival of the fittest.” The strongest organisms get the most resources and survive to pass along their genes. But that’s a gross oversimplification of the way evolution works in the real world. Social animals in particular, which include humans, work together even when they don’t benefit directly. This helps them retain membership in a group that provides protection and shares resources, ensures that close genetic relatives survive, and reduces the need for direct competition. Scientists from Georgia Tech wanted to better understand how robots could develop this kind of emergent behavior and built worm-like robots to study it.
Emergent behavior occurs when individuals in a system cooperate with their neighbors to yield results more complex than any one individual is capable of. This is the classic “hive mind” concept, which is exhibited by insects like ants. A single ant isn’t capable of complex thinking, but many ants together can accomplish remarkably sophisticated feats. It is important to note that emergent behavior isn’t the result of coordination from a central “commander,” but rather interactions between individuals. Those interactions can be surprisingly simple — the kind of behavior that could be described with just a handful of if/then statements. That is what makes emergent behavior so promising when it comes to robots. Instead of explicitly programming complex behavior, roboticists could build numerous “dumb” robots and allow them to develop emergent behavior in order to accomplish complex objectives.
These particular robots were inspired by California black worms, which will huddle together into tight balls when the weather is hot and dry. The worms on the outside of the ball will dry out and die, but the worms inside the ball retain enough moisture to survive. That emergent behavior perfectly illustrates how “survival of the fittest” is more complicated than the brawniest individual succeeding. The robots are mechanically simple, with just a couple of motors to pivot “arms.” Those arms have little hooks and each robot’s 3D-printed body is covered in mesh, so the robots can climb over each other. Their programming was extremely basic: keep squirming until I am out of the light. This caused the robots to exhibit rudimentary emergent behavior, as some robots covered the others to block out the light. In a sense, the top robots sacrificed themselves to protect the robots on the bottom, similar to the California black worms protecting their brethren from the heat. This specific emergent behavior isn’t particularly useful, but it is a stepping stone on the path to more complex cooperation between robots.