A Swarm of Biologically-inspired Little Underwater Explorers
The natural world abounds with self-organizing collectives, where large numbers of relatively simple agents use local interactions to produce impressive global behaviors. Fish schools are particularly impressive – collectives of thousands migrate long distances, search for resources, and even form dynamic shapes like flash expansions or bait balls to evade predators or capture prey. Even more inspiring are the fish schools that move within coral reefs, navigating together in complex cluttered environments. These biological collectives exhibit several properties that are highly desirable from an engineering perspective: they are decentralized, providing robustness to failure of agents, and they rely primarily on local sensing and nearest neighbor interactions, exhibiting high degrees of scalability and adaptability.
The goal of the BlueSwarm Project is to develop a novel 3D swarm testbed inspired by reef fish schools: An underwater robot collective, with 30+ fully-autonomous miniature (~10cm) robots, that use purely local communication and sensing to demonstrate complex global 3D coordination, inspired by the kinds of complexity that coral reef fish schools achieve.
This new project has three main thrusts: (a) The development of an underwater robot swarm platform, with miniature (~10cm) but highly maneuverable underwater robots. (b) The development of algorithms and programming methodologies to create complex global-to-local 3D collective behaviors using implicit coordination. (c) Using BlueSwarm robots to understand fish biomechanics and schooling. See our recent publications and movie links below for current progress.
As an example of complex 3D behavior, imagine an underwater swarm that starts at a base station, then forms a dispersed network to sense a wide area (“shoaling”), then detects a most-profitable direction and collectively migrates in that direction (“schooling/migration”). When it finds a target (e.g. a coral-destroying lionfish), it creates a circular moving perimeter to surround the target (“milling” or “bait ball” behavior). Or migrates back to the base to recharge, without getting lost in the obstacle-rich 3D environment (“localization”). Applications of such a swarm include environmental monitoring in sites of high ecological sensitivity such as coral reefs, gathering of data about ocean acidification and climate change, sampling plankton populations, or inspections of submerged wreck sites for search and rescue operations. The underwater domain poses many challenges, such as inherent sensory and communication deficits, and a bio-inspired collective approach can address these challenges in new ways. Our research will generate a new body of fundamental knowledge about underwater collective robotics -- in particular, minimal sensing, communication, and AI algorithms needed for achieving meaningful collective behaviors.
PEOPLE: Florian Berlinger, Melvin Gauci, Jeff Dusek (now faculty@Olin), and collaborators.
FUNDING: Wyss Institute, Amazon ML Research Award, ONR Science of Autonomy
MOVIES: SSR Youtube Channel