Sheepish physics: How sheepdogs control noisy herds

Controlling multi-agent systems remains a fundamental challenge across fields, from robotics to biology, where complex global behaviors emerge from simple local interactions. The difficulty escalates when agents, like sheep, unpredictably switch between following and ignoring control signals. In this talk, I will use the age-old art of sheepdog trials to illustrate strategies for controlling such noisy, indecisive agents. By leveraging agent-based models, we will explore how individual sheep behaviors—ranging from panicked fleeing to indifferent grazing—can be harnessed for control. The insights offer applications beyond herding, including robotics, ecological systems, biochemical reactions, and even social dynamics.