"Shepherd Dog" turned out to be manipulating herds using two algorithms
There are dogs who skillfully manipulate herds of grazing pastures at wider ranches, guiding them to the intended place and stopping the damage caused by theft and natural enemiesSheep dogIt is a shepherd dog. It is a shepherd dog that skillfully controls herds and manipulates more efficiently than humans, but in reality it became clear that we used two algorithms.
The Mathematics of Herding Sheep | Motherboard
Solving the shepherding problem: heuristics for herding autonomous, interacting agents
I studied in the UK ·University of SwanseaAn international research team led by Dr. Andrew King, a biologist, was announcedpaperAt the same time it clarifies the prospect that it can solve the method and can control the behavior of people other than livestock and use it for environmental purification.
The theory which unraveled the movements of sheep making herds was also called "modern Darwin"William Donald HamiltonDr. announcedpaperIt was clarified in "Selfish herd geometry"Here, the mechanism of the action taken by the sheep exposed to the crisis by the foreign enemies is stated.
When being attacked by a natural enemy such as a wolf, the psychology happens to the sheep "Let me fall into the center of the flock to help themselves." And as all sheep take similar actions, the herds are strong towards the center of the groupcentripetal forceWill occur. This compression force makes the density of the flock very high, and the herd of a group of sheeps shows a movement like a single individual is a theory which "selfish grouping geometry" revealed It is getting.
And shepherd dog seems to be instinctively understanding the psychology and behavior of this sheep, and it is clarified by survey results that skillful use of two actions is actually used when controlling swarm It was. In the survey, a GPS sensor was attached to the shepherd dog and sheep, and the movements of all individuals were recorded in detail, the following movements were observed.
When the sheep (red frame) is in a state of being separated without making a flock, the shepherd dog (blue frame) first goes to the sheep and takes action to put all the sheeps together in one place . This is the first algorithm "aggregation (Collection) "Movement will be.
A state in which sheep has become almost one flock.
Once the herd is formed and becomes a mass, the shepherd dog stimulates one point of the flock which is the furthest away from the target point in the lower left, and creates movement in the herd. This is the second algorithm "Drive (Driving)"is.
Although it is a group of people who gradually advanced, sheep individuals that protrude from the herd as the arrows point out also come out.
Shepherd dogs then stop stimulating the herd and encourage them to return the stray individuals to the flock (Collecting).
Then, once the flock was formed again, I realized that I take the action of driving a flock aiming at the destination again.
Based on this behavior, the research team modeled behaviors and conducted simulations on computers. The situation is shown in the YouTube movie below. The left frame is the actual measurement data, the right is the "movement of the flock of sheep" produced by the simulation.
And this figure summarizes the movement of the simulation. From the upper right to the lower leftDrivingAnd remake the flock out of the courseCollectingThe two algorithms are reproduced in the form of lines.
Based on this research result, the research team will be able to automate the livestock control robot, eliminate environmental pollution due to the outflow of crude oil, and also use it for technologies to control human crowded groups We are doing it.