Birds do it, bees do it. Even ants and fish in the sea do it. When certain individuals group together, they create a “swarm intelligence”— a collective brain capable of solving complex problems which would be insurmountable for an isolated individual. In the world of artificial intelligence, swarm engineering allows us to make robots that work in large numbers (>1000), and tiny sizes (<1 cm). Using strategies that are either inspired by nature (ant colonies, fish shoals, and bird flocks), or automatically discovered using machine learning and crowdsourcing, researchers have demonstrated how swarms of flying robots can be deployed to create outdoor communication networks, how coin-sized robots can form structures and explore their environment, and how nanoparticles can be designed for cancer treatment. In this SFI Community Lecture, computer scientist Sabine Hauert explores how individual actions give rise to swarm behaviors, and the challenges researchers face when engineering swarms for desired applications.
Sabine Hauert leads the Hauert Lab for swarm engineering at the University of Bristol (UK), where she is Assistant Professor in Robotics. Her cross-disciplinary research bridges engineering, mathematics, robotics, and the life sciences. Hauert is also an experienced science communicator and President and Co-founder of Robohub.org, a non-profit dedicated to connecting the robotics community to the world.
Watch the video
Watch some of the highlights:
06:20 The characteristics of self organization – birds, ants, bees, etc.
08:24 Swarm engineers use two tools, bio-inspiration and exploration, to understand swarm behaviors
09:30 Flying robots as an example of the rules of bio-inspiration and swarming
14:20 Using nanoparticles to direct swarm behavior in delivering drugs
22:10 Crowdsourcing the design of nanoparticles with a game called Nanodoc
25:40 Simulating the movement of nanoparticles in artificial tumors on a chip
29:19 Designing nanoparticles with specific properties in a virtual system
31:10 Using augmented reality to get nanoparticles to communicate
32:30 Getting robots to behave like nanoparticles – the kilobot
35:00 Rules for scaling up from nanoparticles to kilobots
36:30 Possible applications of nanobots
38:10 Developing algorithms for nanosystems
43:00 Kilobots that self-organize
45:44 Exploration using machine learning – artificial evolution and behavior trees
50:05 Possible uses of robot swarms in the real world