Biorobotics teaches machines to sense and react to the world

The integration of biology and robotics would allow robots to sense, smell or hear in order to react to changing environments, opening up a wide range of applications.

Our senses allow us to interact with our environment. They are also the basis of learning. In In order to make machines more efficient, researchers are looking at one of the best qualities of living beings: adaptability. The goal is to make robots capable of dealing with unexpected situations in the real world.

Our technologies can already sense the environment far beyond human capabilities, from atoms to the most distant stars. However, this type of equipment often requires large installations and complex data interpretation. “Robot senses should be functional in lightweight, human-scale mobile artifacts, operating in varying scenarios with affordable computing power and power supply requirements,” said Calogero Maria Oddo, head of the touch neuro-robotics laboratory at the BioRobotics Institute in Pisa, Italy.

To achieve this goal, scientists are inspired by nature. In turn, new robotic components and systems can contribute to a better understanding of how nature works. “In traditional terms, the engineer invents while the scientist discovers, but biorobotics makes it possible to establish fertile loops between the drivers of discovery and technological developments. Oddo noted.

Sensation of touch and smell

Touch has been an essential sense for human survival. It can allow us to understand and interact with the world as well as feel danger in the form of pain. In 2021, the Nobel Prize in Physiology or Medicine went in search of molecular receptors that allow us to detect temperature and touch.

Oddo’s research group studies how our nervous system processes touch in order to replicate it in biorobotic systems. This could have applications such as restoring some degree of touch to amputees wearing prostheses or developing sensor robots that can be operated remotely.

A challenge in reproducing the sense of touch is that it requires a large number of individual sensors. Researchers at the Technische Universität München in Germany have developed a electronic skin hexagonal sensing compound modules, which can be attached to a variety of surfaces. The system works with low energy consumption, reproducing a natural mechanism: instead of continuously sending information to the brain, the receptors in the human skin remain inactive until they detect a change.

It could also be useful for robots to feel pain, as undetected injuries could lead to serious accidents. To achieve this, a team from RMIT University in Melbourne, Australia, developed artificial skin that reacts to painmimicking the body’s ability to provide immediate feedback when pressure, heat or cold reaches a certain threshold.

Electronic skin prototype developed at RMIT University

In this way to equip robots with warning signals, another group of researchers from the Chinese University of Hong Kong created artificial skin that can change color for fake bruises. They achieved this by using a molecule, called spiropyran, which changes color from pale yellow to blue-violet under mechanical stress.

Scientists need to better understand our sensory systems to reproduce them with biorobotics. The sense of smell is probably the least understood. This system is so complex that scientists are still unable to predict how a specific smell will be interpreted by our brain. Yet researchers are trying to recreate it in robots

Scientists from the CEA Tech Institute in France have developed an artificial nose and integrated it into a robot capable of detecting survivors during rescue operations. The nose was made using biosensors with odor-binding proteins. He can sense a human victim under the rubble and indicate if the person is still alive.

Artificial olfactory systems could be very useful for odor assessment in the food and perfume industries, as they can be more accurate than humans. Animals such as dogs can also perform this task, but require extensive and expensive training.

A research team from the Hebrew University of Jerusalem has created an optical nose capable of detecting odors using carbon nanotubes. The device uses machine learning to detect unique odor patterns and distinguish between aromas of red wine, beer, and vodka, among others.

super robotic sense

In recent years, scientists have used biological components to improve robots’ abilities to interact with humans. According to Oddo, the bio-hybrid compounds can replicate useful traits of biology, such as self-healing, replicating the structure of neural connections, or working their way through our bodies.

For example, a research team from Duke University in the United States has designed a dragonfly-like robot with wings made of a self-healing hydrogel. Changes in the surrounding pH cause the hydrogel to break or heal, causing it to move away from acidic environments. This allows the robot to find oil spills and clean up the contaminant by soaking it in sponges that the robot carries under its wings.

Another curious example of biorobotics is the odoricopter, an autonomous drone that can avoid obstacles by using a moth’s live antenna to navigate to scents. Developed by researchers at the University of Washington, this robot could be used to access dangerous places to detect gas leaks, explosives or survivors.

“Butterfly antenna cells amplify chemical signals”, said Professor Thomas Daniel. “Moths do this very effectively – one odor molecule can trigger many cellular responses, and that’s the thing. This process is super efficient, specific, and fast.

Smellicopter University of Washington
Researcher attaching butterfly antenna to odoricopter

This approach could also be used to reproduce the sense of hearing. Scientists at Tel Aviv University have connected a cricket’s ear to a robot, taking advantage of the insect’s ability to detect sound. When the researchers clap once, the cricket’s ear responds to the sound and the robot moves forward; when the researchers clap twice, the robot backs off.

Oddo’s research group studies the use of human skin cells such as fibroblasts and hair cells to develop touch-sensitive skin devices. However, further research is needed. “Engineering artifacts with cultured cells present technical complexities, for example related to viability and biosafety, that must be resolved before these technologies are applied in real operational scenarios,adds Oddo.

Nature is full of information. However, our senses can only grasp a tiny part of it. “There are many things happening in our world that our own human sensors do not detect, such as magnetic fields, electromagnetic signals beyond the visual spectrum, very low and very high sound waves beyond our ability to hear, and more. There are countless examples of animals feeling things that we don’t. said Bradley Nelson, professor of robotics and intelligent systems at ETH Zürich in Switzerland.

For example, some animals have amazing abilities to detect explosives, drugs, or disease. Others can detect earthquakes or sonar signals. MIT researchers have already created a robot that uses radio waves to locate and grab objects hidden in plain sight.

“All areas of detection need to be improved”, said Nelson, who believes artificial intelligence will be key to achieving higher biorobotics perception to make decisions about how to react to the environment.

“Force sensing is important for physical interaction in the world, but smart skin still needs a lot of work. Computer vision has made recent progress with improvements in object recognition through deep learning, but really understanding the “story” of an image is still rudimentary Speech recognition has advanced, but we all know how difficult it is in noisy environments.

The boundaries between machines and living beings are collapsing. “In my opinion, future biorobotic systems will advance the current idea of ​​artificial intelligence. Compute functions will no longer be the result of a series of operations, but of architectural design and connectivity between all subsystems involved, including sensors and effectors,” Oddo said.

Cover illustration by Anastasiia Slynko. Images provided by RMIT University and the University of Washington.