Sheep dogs follow simple rules – may be used to develop ‘sheep robots’
We can all be fascinated by skilful sheep dogs’ abilities to lead sheep where they want them. But how do they do it? A multidisciplinary Swedish–British research group consisting of biologists and mathematicians have now studied this, showing that the dogs follow very simple rules.
The study, which is published today in the journal Interface, provides information which makes it possible to develop robots that could be used for instance in herding, evacuation and environmental disasters.
Using a sheep dog to gather and herd a flock of sheep is a well-known example of when an individual (the dog) can collect and move a group of other individuals (the sheep) to a given place, even though the members of the group themselves don’t have any incentive or will to move to that place. However, no one really knows how the dog manages this on its own.
The researchers behind the new study have investigated which rules or ‘instructions’ the dog might be following when it collects a scattered flock of sheep and moves them to a certain place. From Uppsala University, a group of mathematicians have contributed to the study. By constructing an algorithm and running computer simulations, they found that if the dog follows two simple rules, a) gather the flock if it is scattered and b) if the group is collected, drive it towards the goal, the dog can manage to gather and move groups of several hundred individuals to a given target.
The British co-authors have previously conducted and analysed experiments in Australia where a sheep dog collected and moved 46 sheep to a pen in a larger meadow where the sheep had been grazing. High-precision GPS devices had been strapped to the dog and all the sheep so they could see how they moved through the experiment, from the dog being sent off to collect the sheep, until the sheep were gathered in the pen.
‘When we compare computer simulations of our algorithm for group size 46 with the experimental data, we see striking similarities. Both visually in animations and in quantitative measurements of where the dog tends to be located in relation to the flock of sheep’, says Daniel Strömbom, main author of the study.
The same is true for how far the dog tends to be from the flock of sheep’s gravitational centre, and on which side of the flock the dog tends to be relative to the target. This shows that the rules that the researchers have used in their algorithm correlate well with experimental data, and thus that these rules may be those actually used by the sheep dog.
Since the algorithm has the capacity to gather and transport groups much larger than 46 individuals, and with group individuals with varying properties, there are numerous possible applications where it may be useful to implement in a robot. For instance for cleaning up oil spills, handling groups of livestock and assisting with evacuation of dangerous places such as smoke-filled buildings.
‘The advantage of using a robot instead of humans to carry out these tasks is primarily improved safety, for people and animals. In many cases it is probably more efficient and economical to use robots if possible’, says Daniel Strömbom.
The study is a collaboration between Uppsala University, the University of London, the University College of London, the University of Cambridge and Swansea University. Uppsala has primarily contributed knowledge in mathematics and modelling, while the British universities have contributed biological expertise.