Resilient and Adaptable Robot Swarms
Swarming insects work together in large numbers. Swarms are highly adaptive, robust in the face of individual error, and can vary in size—making them an attractive model for robot swarms. With this CAREER award, Kirstin Petersen, Electrical and Computer Engineering, is looking beyond current models of autonomous robot swarms, in which robots work in parallel and operate on the basis of centralized processing or control architectures, and toward swarm intelligence, in which the whole is more than the sum of its parts. This type of distributed coordination is gaining in popularity but is still poorly understood, especially as a design tool for engineering swarms that achieve biological levels of resilience and adaptability.
Natural swarms achieve their scalability, adaptability, and error tolerance by integrating information into—and propagating information through—their shared environment. This research will characterize biological model systems, such as honey bees, and will result in a model of swarms that can function within dynamic environments and can integrate, diffuse, decay, and filter information. This project will extend the concept of environmentally mediated coordination to investigate models of robot swarms that can work in dynamic environments with realistic levels of error and hardware failure.
Adoption of distributed coordination and swarm intelligence in robot swarms can complement existing control architectures. The result could be systems that, though less efficient than those with centralized control architectures, are less costly to deploy, more easily scaled up, more resilient to individual failures, and can adapt to changing tasks or environments.