What happens when the leader of a herd of animals is a robot?

Probably nothing special.
Is there a leader in a group of animals moving in perfect coordination, and can a robot be taught to imitate his behavior and lead a group?

Israel Benjamin Galileo

Schools of fish or birds are one of the most spectacular sights in nature: hundreds, thousands or even millions of individuals move together, create dynamic three-dimensional shapes, change direction suddenly and in coordination, and sometimes appear as one giant creature - it is easy to find impressive video clips online that demonstrate this (try searching "bird flock" or "fish shoal").

When a scientist observes such a phenomenon, he may think not only "How beautiful!" but also to ask questions like:
"How does that happen?"
"How can such a phenomenon be imitated?"
"How can you use the behavior of the bands to achieve a desired goal?"

Bands and group decisions
Is there a leader or leaders leading the pack? If not, how does it happen that the band moves in a coordinated way? If there are leaders, how are the decisions passed from the leader to all the individuals in the pack, many of whom cannot see the leader? Of course the answers are different for different species, but it turns out that there is a lot in common between all of them and even for humans, as we will see later.

In the XNUMXs, a mathematical model began to take shape that describes the coordinated movement of a flock without the need for leaders, although it will be seen later that in many species the leader also has an influence. The model assumes that each individual obeys three laws:

1. Avoid collision with close details.
2. Move in the average direction of movement of your neighbors in the pack.
3. Keep close to your neighbors in the pack (as long as they are beyond the range of the first law).

The first law works at short distances, causing an individual to move away from other individuals if they are very close to him. The second and third laws operate at longer ranges, but still at distances much smaller than the size of the entire herd: in any case, the individual does not need to be aware of the entire herd, but only of the "neighboring" individuals that surround him.

We note here that these laws, despite their success in matching many observations in the field, certainly do not cover the entire range of behavior of the herd or herd. As we know, the group may move in a specific direction or route, and often there are leaders who influence the direction of the movement and the structure of the group. As we will see later, these facts also fit nicely into the model.

where are the fish
In this model, despite its great abstraction, there is a biological logic, and indeed measurements have shown that it predicts well the behavior of flocks. For example, in October 2007, a computer analysis of the movement of flocks of starlings, known for their impressive acrobatic displays, was published (and see link at the end of the column). Studying this movement is a complicated problem that requires fast stereoscopic photography to reproduce the movement of each bird in three-dimensional space, and a combination of techniques from the fields of statistical physics, computer vision and optimization. It was found that the model is suitable when defining for each starling its "neighbors" as the 7-6 birds closest to it, regardless of their distance (previous models assumed that the neighbors are all individuals within a fixed radius around the bird).

On top of that, a preference was discovered for referring to the neighbors that are on the right or left of the bird (that is, not straight ahead), a fact that is not surprising considering the location of the starling's eyes on the sides of the head. The independence of defining neighbors by distance explains the ability of flocks of birds to reunite after they split up (as happens when attacked by a predator).

One of the new studies in this field incorporated another law, which represents temperature preferences, and was able to explain the pattern and migration routes of the capelin fish (a small salmon-like fish) around Iceland. This species migrates hundreds of kilometers between breeding and grazing areas, and has an important role in the marine ecosystem as well as in the fishing industry. In recent years, the population of this fish around Iceland has decreased, so the ability to predict the location of the schools is important for proper control of the fishery.

After the model was tested against data from previous years, each of which had a different migration route, the researchers used it to successfully "predict" the migration route in the 2008-2007 season. The prediction was retrospective: the results were indeed known to the researchers, but not to the software. In the season in question, this route was very different from the typical route, so much so that it seems as if the population "disappeared". Because of this, fishing was stopped in January 2008 and resumed only two months later, when the shoals were found again. If the model had been implemented at that time, it would have somewhat eased the economy of Iceland (which, as you know, went through a severe crisis during this period).

Computer simulation of bands
Such models, which draw their inspiration from living creatures, are also used in the field called "Artificial Life", among other things in the process of creating videos and animations that describe the movement of a flock, swarm or herd. When the simple rules described here are incorporated into such software, the result seen on the screen is a collection of complex behaviors that appear realistic and believable. This technique was used, among other things, by the Disney animators in the segment of the wildebeest herd's panicked flight in the movie "The Lion King".

The emergence of complex behavior from simple laws is one of the most interesting and surprising areas in the study of systems consisting of a large number of autonomous individuals. Understanding the phenomenon is required for understanding collective behaviors in a very wide variety of fields - from the behavior of producers, consumers and investors in the economy, to the dynamics of the immune system.

Understanding the phenomenon can also help in creating a new generation of simple and cheap robots that cooperate with each other to carry out their tasks. In such uses, it is impossible to provide each robot with enough sensing means and calculation ability to "see the big picture", and it is not desirable to subject all robots to a central calculation that will control each of them. Even if it is possible to build such a system, it will fail as soon as the central computer breaks down or the connection with it is lost. Therefore, multi-robot systems are sometimes designed according to principles taken from animal flocks, such as the way a swarm of ants patrols a wide area, identifies food sources, and brings food to the nest.

Using the herd effect
Shepherds know how to control a large flock and move the sheep as they please by understanding the behavior of the flock. One of the methods for this is the use of a small number of dogs, trained to act so that the herd's instinctive response to a threat from predators will lead the entire group, which may include hundreds of sheep, in the direction desired by the shepherd. Dolphins also seem to know the laws of movement of schools of fish: groups of dolphins have been observed using air bubbles and sound production to cause fish to gather into a ball that revolves around itself. Once this goal is achieved, each dolphin in turn dives into the ball to satisfy its hunger.

Herding behavior that can be manipulated appears in humans as well: researchers at the University of Leeds (Leeds) in England published an article in 2008 in the Journal of Animal Behavior, describing an experiment in which the human participants were asked to walk randomly in a large hall, where they had to maintain proximity to others but were forbidden to communicate with each other. Only ten of the two hundred participants received instructions about their route of movement, but this was enough to cause the others to form into groups that moved accordingly, without realizing that someone was leading them (see link). It is clear that the conditions of the experiment are artificial and different from the realistic actions of individuals in a crowd, but nevertheless the researchers expect the results to provide insights into the planning of traffic routes in public areas for the purpose of preventing overcrowding and for rapid evacuation in times of emergency.

The same idea of ​​leading a band by a "plant" leader was recently realized by Dr. Maurizio Porfiri from the Polytechnic Institute of New York University (NYU-Poly). Although many species of fish do not need a leader to lead a school, it is also known that the fish in a school tend to be more influenced by a fish in their environment if that fish is larger or more active than the others. With the help of Dr. Porfiri's main specialization, in the fields of mechanical engineering and "smart materials", he developed a robot that is the size of a palm and swims quietly - an essential feature because noisy mechanisms drive away the fish. The robot's "muscles" are made of ionic polymers that contract and release in response to electrical pulses, thus creating a movement with characteristics similar to the swimming of a real fish.

The current version of the robot is controllable only on a plane: it cannot rise to the surface of the water or descend to the bottom. Therefore the experiments are carried out in a shallow water tank. In a video clip documenting the experiment, you can see how fish that are about half the length of the robot join it and follow its movements. Recently, researchers from the University of Leeds (Leeds) in England also published a similar study, with a fish whose shape is more faithful to reality but its movement is less convincing, and even there the "leader" was able to lead the pack.

One could imagine fishermen sending robots to lead shoals of fish straight into their nets, but Dr. Porfiri has other ideas: he suggests using such methods to keep fish away from danger zones, such as the waters near nuclear power plants, and to lead migrating birds to new nesting areas when their familiar nesting areas are destroyed, and even lead people to a safe exit when a fire breaks out.

These days, scientists are expressing concern for the fate of the western population of the bluefin tuna fish. This population, which has already suffered an 80% decrease compared to its size in the sixties due to increased fishing, reproduces in the Gulf of Mexico between spring and summer, so the eggs and young fish may be severely damaged by the oil spill in the Gulf of Mexico that began at the end of April this year. Dr. Porphyry's development creates hope that in the future we can better deal with the ecological consequences of such environmental hazards by sending a small number of robots to lead large populations to safety.

The writer works at ClickSoftware developing advanced optimization methods.