Abstract
Swarm systems exhibit complex behaviour that emerges from simple local interactions, yet deciphering these underlying rules remains a challenging task. This talk presents a method that extracts human-readable behaviour tree controllers directly from video demonstrations of swarm behaviours. By bridging the gap between emergent collective behaviour and local agent rules, this approach empowers engineers to design and control swarm systems while providing biologists with automatic ways to generate testable hypotheses from observed natural swarms. From simulated robots and real robots to ant colonies, we elevate "seeing" swarm behaviour to reveal deeper insights and facilitate practical design and control.
About the speaker
Khulud Alharthi holds a Master’s degree in Computer Science and earned her PhD in Swarm Robotics from the University of Bristol, where she conducted research in the Hauert Lab. She is currently a postdoctoral researcher at the University of Bristol’s Data Science Lab. Following her postdoctoral work at the University of Bristol, she will return to Saudi Arabia to take up her role as Assistant Professor in the Department of Computer Science at Taif University.