Some of the most perplexing and awe-inspiring phenomena in nature involve self-assembly, where thousands—sometimes millions—of individual disorganized actors manage to form large, complex structures solely through local interactions. This can be seen when there are dozens of birds flying in a flock, massive schools of fish which all seem to swim according to some unspoken rule, or when certain species of ant come together to form bridges and rafts with their bodies.
So far, figuring out how to replicate these natural instances of self-assembly with an algorithm and some robots has proven to be challenging. There have been some successful instances with autonomous drone swarms and other microrobots, which researchers have been able to program to self-assemble to perform different tasks. This trend in ‘programmable matter’ took yet another step forward last year thanks to work coming out of Harvard’s Self-Organizing Systems Research group, which developed a “large scale robot collective” that can self-assemble into a variety of different shapes.
As detailed in a paper the Harvard researchers presented at the 2016 Distributed Autonomous Robotic Systems conference (and seen in the above video), they managed to create a self-assembling robotic system that is based on a subtractive approach as opposed to the more common additive approaches to autonomous self-assembly robotics.
Generally speaking, self-assembling robotic systems begin as a disorganized mess and then come together in complex patterns by adding more and more robots until the desired configuration is achieved. In the Harvard model, however, 725 Kilobot robots begin in a tight grid assembly and then each robot decides if it is part of the desired shape that has been entered into the system. Those robots that are not a part of the desired shape then disperse to the edge of the coordinate system (or in this case the box in which the robots are placed), leaving only the desired shape remaining.
Remarkably, the robots are able to do this using only a single overhead light to guide their motion while only being able to communicate with other robots that are up to three body lengths away, which means a given robot is never communicating with more than 36 other robots at a time. The robots gauge the distance from one another based on the wireless signal strength and are accurate to the millimeter.
According to the research report presented at DARS, “the subtractive method [seen above] leads to higher reliability and order-of-magnitude improvement in shape formation speed.” As the researchers point out, a subtractive method of self-assembly, or self-disassembly, offers a number of advantages, such as requiring a low-level of motion precision, which becomes advantageous the more robots are involved in a swarm.
“Our theoretical and experimental results suggest that such a self-disassembly algorithm can achieve a wide class of shapes with high efficiency and accuracy, making it a good candidate for shape formation in modular robots and programmable materials,” the researchers wrote in conclusion. In the future, they hope to add a component to the algorithm that allows for random movement instead of programming the bots to either move toward or away from the light, which will allow for a more dynamic range of possible shapes.
Get six of our favorite Motherboard stories every day by signing up for our newsletter.