Israeli robot learning to smell
An Israeli technological breakthrough is enabling researchers to teach robots to detect and identify odours.
A Tel Aviv University robot being taught to smell. Photo by Tel Aviv University
Tel Aviv University researchers made it possible for a robot to use a locust’s antenna to detect and interpret smells with a sensitivity 10,000 times higher than existing electronic devices. The findings, recently published in the peer-reviewed Science Direct, suggest the technology will eventually be able to sniff out explosives, drugs, diseases and more.
The researchers connected a desert locust’s antenna to an electronic system and then used a machine-learning algorithm to teach it.
“Man-made technologies still can’t compete with millions of years of evolution. One area in which we particularly lag behind the animal world is that of smell perception,” said Dr Ben Maoz, one of the project’s lead researchers.
“An example of this can be found at the airport where we go through a magnetometer that costs millions of dollars and can detect if we are carrying any metal devices,” D Maoz explained.
“But when they want to check if a passenger is smuggling drugs, they bring in a dog to sniff him. In the animal world, insects excel at receiving and processing sensory signals. A mosquito, for example, can detect a 0.01% difference in the level of carbon dioxide in the air. Today, we are far from producing sensors whose capabilities come close to those of insects,” he said.
The researchers point out that human and animal sensory organs, such as the eye, ear and nose use receptors that identify and distinguish between different signals. The sensory organ translates these findings into electrical signals, which the brain decodes as information.
According to Prof. Yossi Yovel, another lead researcher, teaching the robot to identify smells was simply a methodical process.
“We connected the biological sensor and let it smell different odours while we measured the electrical activity that each odour induced. The system allowed us to detect each odour at the level of the insect’s primary sensory organ,” Prof. Yovel explained.
“Then, in the second step, we used machine learning to create a ‘library’ of smells. In the study, we were able to characterise eight odours, such as geranium, lemon and marzipan, in a way that allowed us to know when the smell of lemon or marzipan was presented. In fact, after the experiment was over, we continued to identify additional different and unusual smells, such as various types of Scotch whiskey,” he said.
Prof. Yovel added, ” A comparison with standard measuring devices showed that the sensitivity of the insect’s nose in our system is about 10,000 times higher than the devices that are in use today.”
The researchers say their next step is to give their robot the ability to navigate in order to localise an odour’s source.
Dr Maoz noted that their research can also be applied to the sense of sight and touch.
“Some animals have amazing abilities to detect explosives or drugs; the creation of a robot with a biological nose could help us preserve human life and identify criminals in a way that is not possible today. Some animals know how to detect diseases. Others can sense earthquakes,” he said. “The sky is the limit.”
