The robot chef learns to “taste as you go”

A “chef” robot was trained to taste food at different stages of the chewing process to evaluate whether it was ripe enough.

Working closely with home appliance maker Beko, Cambridge University researchers trained their chef robots to evaluate the salinity of a food at different stages of the chewing process, mimicking a similar process in humans.

Their discoveries may be effective in developing automatic or semi-automated food preparation, helping robots learn what is good and what is not, which can make them cook better.

When we chew our food, we notice changes in texture and taste. For example, biting a fresh tomato at the height of summer will produce juice, and when we chew, both saliva and digestive enzymes are secreted, our perception of tomato taste will change.

The robot chef, who was previously trained to make omelettes based on human taste responses, tasted nine different variations of a typical dish of scrambled eggs and tomatoes in three different stages of the chewing process and created “taste maps” of different foods. .

Researchers have found that this ‘taste as you go’ method significantly improves the robot’s ability to quickly and accurately assess the salinity of a dish compared to other electronic testing techniques, which only test a sample. The results are published in the journal Frontier 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; The saliva produced during chewing helps transport chemical compounds from food to taste receptors, primarily through the tongue; And signals from taste receptors are transmitted to the brain. Once our brains are aware of taste, we decide whether we like food or not.

The taste is also very unique: some people prefer spicy food, others have sweet teeth. A good chef, whether amateur or professional, depends on their sense of taste and can balance the different flavors of a dish into a well-formed end product.

“Most home chefs are familiar with the concept of tasting on the go – testing a dish through the cooking process to see if the taste balance is right,” said Gorgezger Sochaki of the engineering department. The first author of the article was from Cambridge. “If robots have to be used in certain areas of food preparation, it is important that they are able to ‘taste’ what they are cooking. A

“When we taste, the process of chewing constantly responds to our brains,” said Dr. Arsene Abdullali, co-author of the engineering department. “Current electronic testing methods only take a single snapshot from a homogeneous sample, so we wanted to replicate a more realistic chewing and testing process in a robotic system, which should be a tasty end product.

Researchers are members of Cambridge’s bio-inspired robotics lab, led by Professor Fumia Eider of the Department of Engineering, which focuses on training robots to solve so-called last-meter problems that humans find easy, but robots find it difficult. Cooking is such a task: previous experiments with their “chef” robot created a passable omelette using feedback from human tasters.

“We needed something cheap, small and fast to add to our robot so that it could taste: it was cheap enough to use in the kitchen, small enough for a robot and fast enough to use while cooking,” Sochaki said. .

To mimic the process of human chewing and tasting in their robot chefs, researchers have attached a conductance probe, which acts as a salinity sensor in a robotic arm. They make scrambled eggs and tomatoes, the number of tomatoes and the amount of salt in each dish are different.

Using the probe, the robot ‘tastes’ the food like a grid, returning readings in just a few seconds.

To mimic the change in texture resulting from chewing, the team then puts the egg mixture into a blender and the robot examines the dish again. Different readings at different “chewing” points map the taste of each dish.

Their results show a significant improvement in robots’ ability to assess salinity compared to other electronic testing methods, which are often time consuming and provide only one lesson.

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

“When a robot learns to cook, like any other cooking, it needs feedback on its effectiveness,” Abdullali said. “We want robots to understand the concept of taste, which can make them cook better. In our experiments, the robot can “see” its differences as it chews food, which improves its ability to taste. A

“Beko’s vision is to bring robots into a home environment that is safe and easy to use,” said Dr. Muhammad W. Chughtai, Chief Scientist of Beko plc 6 “We believe that the development of robotic chefs will play a big role in the future of busy homes and living facilities. This result is a leap forward in robotic cooking, and using machine and deep learning algorithms, mastication will help robot chefs adapt. Different foods and flavors for users. A

In the future, researchers want to improve the robot chef so that it can taste a variety of foods and improve its sensing ability so that it can taste sweet or fatty foods, for example.

The study was funded in part by Beco PLC and the Engineering and Physical Sciences Research Council (EPSRC), part of the UK Research and Innovation (UKRI). Fumia Ida is a Fellow of Corpus Christi College, Cambridge.

Leave a Comment