Robot chef learns to “taste on the go” AI & Robotics News

The robot chef was trained to taste foods at different stages of the chewing process to gauge if they were seasoned enough.

Working with appliance maker Beko, researchers at the University of Cambridge trained their robot chef to judge the saltiness of a meal at different stages of the chewing process, mimicking the same process in humans.

Their findings could be useful in developing automated or semi-automated food preparation, helping robots learn what tastes good and what doesn’t, making them better cooks.

When we chew food, we notice a change in its texture and taste. For example, when we bite into a fresh tomato at the height of summer, we release juice, and when we chew, releasing saliva and digestive enzymes, our perception of the taste of a tomato changes.

The robot chef, who had previously been trained to make omelettes based on taster feedback, tasted nine different variations of a simple scrambled egg and tomato dish at three different stages of the chewing process and compiled “taste maps” of the various dishes. .

The researchers found that this “taste” approach significantly improved the robot’s ability to quickly and accurately judge the saltiness of a dish compared to other electronic tasting technologies that only test a single homogenized sample. The results are published in Frontiers of Robotics and AI.

Taste perception is a complex process in humans that has evolved over millions of years: the appearance, smell, texture, and temperature of food affect how we perceive taste; saliva, formed during chewing, helps transport chemical compounds from food to taste buds, primarily on the tongue; and signals from the taste buds are transmitted to the brain. Once our brain is aware of the taste, we decide whether we like the food or not.

The taste is also very individual: someone likes spicy food, and someone has a sweet tooth. A good cook, whether amateur or professional, relies on his sense of taste to balance the various flavors of a dish into a well-balanced end product.

“Most home cooks are familiar with the concept of tasting on the go — checking a dish as it cooks to make sure the balance of flavors is right,” said Grzegorz Sochacki of Cambridge Engineering, the paper’s first author. . “If robots are to be used for certain aspects of food preparation, it is important that they can ‘taste’ what they are cooking. »

“When we taste, the chewing process also provides constant feedback to our brain,” said co-author Dr. Arsen Abdulali, also from the engineering department. “Existing electronic testing methods only take one picture of a homogenized sample, so we wanted to replicate a more realistic chewing and tasting process in a robotic system, which should result in a tastier end product. »

The researchers are members of the Cambridge Laboratory for Bioinspired Robotics, led by Professor Fumiya Iida of the Faculty of Engineering, which trains robots to solve so-called last meter problems, which are easy for humans and difficult for robots. Cooking is one such challenge: previous tests with their robot chef produced a passable omelette using feedback from human tasters.

“We needed something cheap, small and fast to add to our robot so it could do a tasting: it had to be cheap enough to use in the kitchen, small enough for the robot and fast enough to be able to use during cooking,” Sokhatsky said. . .

To mimic the human process of chewing and tasting in their robot chef, the researchers attached a conductivity probe, which acts as a salinity sensor, to the robot’s arm. They cooked scrambled eggs and tomatoes, varying the number of tomatoes and the amount of salt in each dish.

With the help of a probe, the robot “tried” the dishes like a grid, returning the result in just a few seconds.

To mimic the texture change caused by chewing, the team then put the egg mixture into a blender and asked the robot to test the dish again. Different readings at different chewing points provided flavor maps for each dish.

Their results showed a significant improvement in the robots’ ability to judge salinity compared to other electronic tasting methods, which are often time-consuming and produce only one reading.

Although their method is a proof of concept, the researchers say that by mimicking human chewing and tasting processes, the robots will eventually be able to produce foods that humans will enjoy and can be modified to suit individual tastes.

“When a robot learns to cook, like any other chef, it needs feedback on its work,” said Abdulali. “We want robots to understand the concept of taste, which will make them better cooks. In our experiment, the robot can “see” the difference in food as it chews, which improves its palatability. »

“Beko’s vision is to bring robots to homes that are safe and easy to use,” said Dr. Muhammad V. Chugtai, chief scientist at Beko plc. “We believe that the development of robotic chefs will play an important role in residential and nursing homes in the future. This result is a breakthrough in robotic cooking, and the use of machine learning and deep learning algorithms will help robot chefs adapt. taste for different dishes and users. »

In the future, the researchers plan to improve the robot chef so that it can taste different types of food, as well as improve its sensory capabilities so that it can taste sugary or fatty foods, for example.

The study was funded in part by Beko plc and the Engineering and Physical Sciences Research Council (EPSRC), part of UK Research and Innovation (UKRI). Fumiya Iida is a member of Corpus Christi College, Cambridge.

Back to top button