Have you ever wondered how a farmer and a single sheep dog can herd sheep? A recent paper in Journal of the Royal Society Interface explains just how. Using GPS data from collars researchers have developed a computational model which "reproduces key features of empirical data collected from sheep–dog interaction". The model has two simple rules:
"The first rule: The sheepdog learns how to make the sheep come together in a flock. The second rule: Whenever the sheep are in a tightly knit group, the dog pushes them forwards." (BBC)
The movie below (which accompanies the paper) shows some simulation runs with different numbers of agents. The shepherd (blue) approaches and rounds up the agents/sheep (black dots) and then proceed to herd the group toward the target.
Strömbom. D. Mann, R. P., Wilson, A. M., Hailes, S., Morton, A. J., Sumpter, D. J. T., King, A. J. (2014) Solving the shepherding problem: Heuristics for herding autonomous, interacting agents. Journal of the Royal Society Interface, 11: 20140719.
Thanks to @Badnetworker for drawing my attention to this.