Honeybees guide their hive mates to new food sources by dancing; guppies negotiate leadership with their mates; carrier pigeons take shelter when falcons strike. Scientists have been studying social interactions like these since the advent of animal behavior research. Today, their research has a new twist – not only with live animals, but also with robotic animals. Under the control of the researchers, the robotic animals interact with live, flesh-and-blood animals in experiments. Scientists hope that experiments like this will reveal new insights into what factors make guppies so socially adept? How do bees communicate to their companions? And, what other characteristics of the social life of animals?
This may sound unusual, but in fact, related research has already begun. Advances in robotics and computing power mean that engineers can create robotic animals that are lifelike enough for animals to recognize and respond to one another. The level of realism that is “realistic enough” varies from subject to subject. Sometimes the robot has to look realistic, sometimes it needs to smell the same, sometimes it just needs to move.
Robotic animals have one big advantage over live animals: They do tasks exactly the same way over and over again, as instructed by researchers. This gives scientists a degree of control over their experiments that would be difficult or impossible to achieve any other way. “If you can make a robotic animal, and it’s an eyeliner, it’s mixed into a group of animals, and the animals can accept the robot as a member of the group, and then you can make the robot perform tasks, see See how real animals react,” said Dora Biro, an animal cognition researcher at the University of Rochester in New York.
With robotic animals, researchers can separate factors, such as the size of the fish and its experience, that cannot be separated in real animals. They can speed up experiments by subjecting real animals to the exact same stimuli over and over. Sometimes, they can also accomplish their goals while avoiding the animals facing real predators or potential invasive species.
Below are five biomimetic robotic animals that researchers are using to study and control the social interactions of real animals in specific situations.
Robot bee sneaking into the hive
| Robotic bees enter the hive |
The famous bee “swing dance” has been known for more than 60 years – returning worker bees move rapidly in a specific route near the hive entrance and vibrate their wings and bodies, in this way they tell their companions the location of food. But researchers still don’t know how their hive companions decode this information. “What is the meaning of the signal? Which parts of the dance really convey the message? Which are just unintentional movements?” asked Tim Landgraf, a robotics expert at the Free University of Berlin. These are the tasks the robot bees have to accomplish, he believes.
Landgraf built a model the size of a real bee — a rough plastic ball in the shape of a bee with a single wing — and connected it to a mechanical drive system to control where and how the model moved and vibrated. After placing the robotic bees in the hive, Landgraf found that the robotic bees could indeed guide real bees to food sources, even ones they had never used before. This result is a strong proof of the theory.
But the success of robo-bees is not reliable. “Sometimes the bees follow within seconds, but sometimes it takes days, and we can’t explain why,” Landgraf said. He realized there was another problem he hadn’t considered— – How the bee chooses which one of the dancing worker bees to follow, and when to follow. He wondered if potential followers were actively seeking out information about food sources, or did the dancers have to convince them somehow? Are all the specific signals understood by only a few experienced worker bees?
To answer these questions, Landgraf and his team are developing an upgraded version of the robotic bee—one with a more realistic scent and more reliable wing-vibration mechanism that can enter a uniquely marked bee hive— And hopefully it will be able to trace the experience of living bees. Despite the delays caused by the new crown epidemic, they finally began to test the system, but the results are not yet available. Still, Landgraf said: “I reckon we’re likely to have some new discoveries.”
| Robot Falcon Hunting |
How would a flock of pigeons react when a falcon strikes? The classic “selfish herd principle” assumes that each pigeon will try to squeeze into the middle of the flock when attacked, causing predators to take away other unfortunate companions. But the idea is not easy to test. Every falcon’s attack is different – the height and angle of the starting position can be different, and all these changes affect the pigeon’s response. So Daniel Sankey, a behavioural ecologist at the University of Exeter in the UK, started using robotic falcons.
”This controllable approach is very useful for research,” says Sankey. “You can make sure that when the pigeon is released, the falcon is always exactly 20 meters behind, and the operation can be repeated.” Furthermore, he points out, the robotic falcon is very useful to the pigeon. safer. “As far as I know, there has been a case where a trained falcon wiped out a flock of pigeons.”
A falcon enthusiast helped Sankey with his robotic falcon. Aside from an extra propeller that drives it, the robotic falcon looks lifelike. Sankey repeatedly “attacked” a flock of carrier pigeons while tracking the location of each pigeon via GPS. He found that, contrary to the “selfish herd principle”, the pigeons did not make an effort to crowd the flock when attacked.
The robotic falcon “attacks” the pigeon flock.
In addition, Sankey’s analysis shows that these pigeons mostly try to fly in the same direction as their mates, so that the flock can evade attacks in unison without any pigeons falling behind into the hands of predators. “This shows that they can work together to escape predators without sacrificing any of their companions,” Sankey said. Although there is no conclusive evidence, this suggests that the flock may be cooperative and not selfish.
| Robot fish in school |
Which fish in the school is most likely to be the leader? Most studies have shown that larger fish tend to have a greater impact on the direction the school is heading. Larger fish are generally older and more experienced, and they also behave differently than smaller fish in a school. But which of these factors has the greatest impact on being a fish leader? This is hard to test with real fish. “How can you make a big fish behave like a small fish? These can only be tested with robotic animals,” says Jens Krause, an animal behaviorist at Humboldt University in Berlin. He co-authored a review of robotic animals in behavioral research in the 2021 Annual Review of Control Methods, Robotics, and Autonomous Systems.
Krautzer and his colleagues developed the robotic fish—a 3D-printed model of a guppy mounted on a magnetic base, powered by an electric mechanism beneath the tank. Two cameras linked to a computer allow the robotic fish to respond instantly to the actions of its companions.
The school of fish sees the robotic fish as one species.
They found that as long as the model had eyes and a color pattern roughly similar to that of a real fish, the guppies would treat the model as a normal live fish and respond. This allowed the researchers to swap out larger or smaller versions of the robotic fish, all else being equal, to study the effect of body size alone. They found that guppies did tend to follow larger robotic fish leaders. The team also used the robotic fish to study how the travel speed of individual fish affects group behavior.
The team also discovered another surprising fact about shoal leaders: Social etiquette is important. An early version of their control program would cause the robotic fish to get too close to their companions, forcing the real fish to back away to keep their distance. “We had some robotic fish chasing live fish,” recalls Krause. Later, the team tweaked the robotic fish so they could maintain a polite social distance. And the new version of the “social master” robot fish has proved that they have an absolute advantage in attracting followers.
| Robot termites gather in swarms |
Previous studies have introduced robotic animals into groups of live animals to stimulate a response from the live animals. Also, there is a way to parse animal behavior using robotic animals alone: program a group of robotic animals to follow what you know is the behavior of real animals, and see if they can replicate the behavior of real animals.
That’s the approach used by Harvard collective behavior researcher Justin Werfel. Wayfair wanted to understand how termites build such complex anthills, which are known for their array of grooves at the entrance. He focused his research on one specific step in the construction process—how termites choose a dumping site for soil excavated from an anthill. This simple judgment determines the complex shape of the anthill entrance.
Werfel and colleagues have some evidence that termites may dump soil at the junction of high-humidity areas inside the mound with dry air at the surface, a good indicator of the boundaries of their homes. But they didn’t know if the termites’ dumping behavior was also influenced by other factors.
So they made a colony of robotic termites. Since robotic termites don’t have to interact with real insects, they don’t have to appear lifelike. Instead, these robotic termites are brick-sized carts that transport and unload colorful blocks on a flat surface. Each “termite” carries a humidity sensor, which is programmed to carry the building blocks when humidity is high and remove them when humidity drops. At the same time, as each “termite” moved, a hamster pipe was constantly dripping water to ensure higher humidity in the area occupied by the robotic termites.
Scientists have used robotic termites to study termite mound and soil dumping behavior.
”We know that the robotic termites only care about humidity because that’s what we’re giving them,” Wefel said. As it turns out, that’s enough. Because eventually, the robotic ant colony unloads the blocks at the entrance to their “anthill”, and they even seal off the entrance on windy days, just like real termites. Of course, Wefel points out that while this doesn’t prove that termites build their anthills based solely on humidity rules, it does suggest that such rules are sufficient for them to get the job done.
|The “nemesis” fish lurks in the water |
Not only can biomimetic robotic animals reveal the secrets of animal behavior, they may soon be used to manipulate live animals in specific fields.
The mosquito fish is native to the southern United States and is now one of the top 100 invasive species in the world. Behavioural ecologist Giovanni Porverino of the University of Western Australia decided to control bionic robotic animals in an unusual way.
Scientists are trying to scare the invasive mosquito fish with a robotic largemouth bass.
Polverino and his colleagues built a robotic fish designed to look like a largemouth bass. In their native waters, largemouth bass are their main predators. Scientists have programmed robotic fish to swim aggressively towards mosquito fish, hoping to scare invasive species without affecting Australia’s native species. (Many wild animals exhibit persistent fear responses.)
And here’s what they saw: As little as 15 minutes a week of contact with a robotic predator, the mosquito fish lost less body fat and more energy to escape. , and the energy used for reproduction is correspondingly reduced. “Its impact on mosquito fish is huge, but other species are not at all afraid because we’re replicating a predator that doesn’t exist in Australia,” Polverino said.
Polverino still has a lot of work to do before dropping his artificial Predator into the real world. “Our robotic fish performed well in the lab,” he says, “but this was achieved with a computer nearby, a webcam above the tank, and a battery.”
Even so, he A decision was still made to negotiate with a national park in Queensland, where two endangered species of fish live in small, clear pools that have recently been occupied by mosquito fish. The pool is small, which may provide a good environment for the first field test. “It’s not ready yet,” Porverino said, “but it’s something that can be tried.”
Of course, when researchers try to integrate robotic animals into real animal populations, many things can go wrong—sometimes , the reason for the failure is not complicated at all. For example, when Biro tried to build a robotic pigeon to study the collective decision-making of homing pigeon flocks, the robotic pigeon couldn’t keep up with the real flock. Still, the opportunity to test animal behavior in new ways leaves her hopeful, and she hopes to try it again someday. “If we can get this all working, then there’s all sorts of interesting things waiting for us to do,” Biro said. “That’s what I’ve been looking forward to.