A new framework has been designed to push forward swarm intelligence, the branch of AI that mimics the group behaviors of birds, fish, and bees.
The coordinated movement of robots could improve search-and-rescue operations and wildfire detection.
The collective intelligence found in nature is a wonder of efficiency and coordination. Birds flock to forage. Fish school as a way to avoid predators. Bees use swarming as their method of reproduction.
However, replicating this self-organizing behavior in artificial swarms has been a major challenge for researchers.
“One of the great challenges of designing robotic swarms is finding a decentralized control mechanism,” said Matan Yah Ben Zion, an assistant professor at Radboud University’s Donders Center for Cognition, in a press release Monday from New York University (NYU).
“Fish, bees, and birds do this very well—they form magnificent structures and function without a singular leader or a directive. By contrast, synthetic swarms are nowhere near as agile—and controlling them for large-scale purposes is not yet possible,” added Ben Zion, the study author.
A simple rule for complex behavior
The study focuses on a central challenge for robotic swarms: establishing a decentralized control mechanism.
This is a way to ensure robots can work together effectively as a group, similar to a flock of birds or a school of fish, even without a single guiding authority.
The researchers developed a set of geometric design rules for controlling swarms to overcome this issue.
These new rules are based on natural computation, like protons and electrons’ positive and negative charges.
A new quantity called “curvity” was introduced to the model. Curvity is an intrinsic charge-like quality that causes a robot to curve in response to an external force.
According to the new framework, each robot is assigned a positive or negative curvity value to control the way it interacts with its fellow robots.
“This curvature drives the collective behavior of the swarm, which points to a means to potentially control whether the swarm flocks, flows, or clusters,” said Stefano Martiniani, assistant professor at New York University.
Applications in drug delivery
In a series of experiments, the researchers successfully demonstrated this new framework. It showcased that their curvature-based criterion controls how pairs of robots are attracted to one another.
Moreover, this principle naturally scales up to control the movements of thousands of robots. Researchers discovered that it can be directly embedded into a robot’s mechanical design.
Interestingly, this geometric rule could be applied to large industrial or delivery robots and microscopic robots designed for medical treatments like drug delivery.
These new rules for swarm control are based on simple, elementary mechanics, making them easy to implement in a physical robot.
Furthermore, the new strategy could transform the challenge of controlling swarms from a complex programming problem into an issue of material science.
The development of robotic swarms is an evolving field with several recent advancements.
In April, H2 Clipper secured a patent for using robotic swarms in large-scale aerospace manufacturing.
In another study, Pennsylvania engineers developed a decentralized swarm strategy. The tiny robots follow simple mathematical rules to self-assemble into complex, honeycomb-like structures by reacting only to their immediate environment.