Chile stuff 5 month 7 Daily news , In recent days, , Researchers at the University of Cambridge in the UK trained a chef robot to constantly “ chew ” And taste the food , To imitate the cook's cooking process .
The birth of a delicious dish , Rely on the chef to taste continuously during the cooking process 、 Increase or decrease seasoning , In the process , Taste is an important index for people to judge the taste of food .
So that robots can “ Cook while tasting ”, Although it sounds absurd , But the first author of the paper 、 From the Engineering Department of Cambridge University Grzegorz Sochacki say ：“ We hope robots can understand the concept of taste , This will make them better chefs .”
The researchers developed a device with a conductance based taste sensor UR5 Experimental device of mechanical arm , Simulate chewing by mixing food 、 Electric current conduction repeats the taste of salt , Help robots taste food “ taste ”.
▲ The researchers' experimental device
therefore , Trained robots “ The cook ” You can taste the salinity of food at different stages of the chewing process , The salinity information is generated and sent to the computer , Then generate a visual taste data image .
The title of the paper is Mastication-Enhanced Taste-Based Classification of Multi-Ingredient Dishes for Robotic Cooking（《 Classification of multi-component dishes based on chewing to enhance taste for robot cooking 》） Has been in 5 month 4 Published in Swiss open access publisher Frontiers Its robotics academic journal Frontiers in Robotics & AI（《 Robot and artificial intelligence frontier 》） On .
link ：10.3389 / frobt.2022.886074
01. Stir and taste , Simulate human cooking process
Cooking is one of the difficult problems in the field of robot automation , Several researchers participated in the study of different automation links of cooking , Including controlling the time of frying sausage by the robot through visual feedback 、 Remotely operated robots decorate cakes 、 Loading dishwashers, etc. with a mechanical arm .
Researchers also use “ Electronic Tongue ” To help robots detect meat and so on , But this process often requires shredding 、 Complex processes such as separation or mixing with alcohol , Can make “ Electronic Tongue ” Play a role , The biggest difference between robot chefs and human chefs in the cooking process is that human chefs can “ Taste and make ”.
therefore , Existing solutions for robots “ The cook ” Not in time for . The co-author of the paper 、 Doctor of engineering, Cambridge University Arsen Abdulali say ：“ Current electronic testing methods only take a snapshot from homogeneous samples , Therefore, we hope to replicate a more real chewing and tasting process in the robot system , This should produce a more delicious end product .”
in fact , When people chew food , You will taste the changes in taste and texture , for example , In summer, when we take a bite of fresh tomatoes , Tomatoes release juice , Plus the saliva and digestive enzymes released by human chewing , Will change our perception of tomato flavor .
The researchers are the first author of the paper 、 Cambridge engineering department Grzegorz Sochacki say ：“ Most home cooks are familiar with the concept of eating while eating —— Check a dish throughout the cooking process , To check whether the taste balance is correct . If robots are to be used in some aspects of food preparation , The important thing is that they can ‘ Taste ’ What they are cooking .”
Researchers at the University of Cambridge have found that cooking while tasting can improve the ability of robots to quickly and accurately assess the saltiness of dishes . therefore , The researchers trained their robot to taste different scrambled eggs with tomatoes , Taste nine different scrambled eggs and tomatoes at three different stages of the chewing process , Then the taste data images of different dishes are generated .
▲ Experimental process
Their findings may help develop automated or semi automated food preparation , It can help robots learn what tastes good , What tastes bad .
02. 9 Try each dish 3 Time , Generate visual taste image
Reproducing human chewing process can also extract more information during chewing . Researchers said , In several states of food processing, tasting can significantly improve the classification performance of foods with different quantities and the same components .
In order to prove the above conclusion , The researchers modeled the human tasting process , measurement “ chew ” In the process, the taste of food at different stages and generate data . Chewing is the process of crushing and grinding food , Its main purpose is to reduce the average size of food particles , meanwhile , Smaller particles can also provide greater surface area for digestive enzymes to function . therefore , Chewing plays a very important role in the tasting process , The researchers set up a mixer to simulate the process .
In the measurement of taste , The robot reproduces the taste of salt through a conductivity sensor , Salinity increases with ion concentration 、 Ion mobility and ion charge increase .
Considering that there are multiple receptors on the surface of the human tongue , Researchers will taste at multiple points in the experiment and express the taste as a series of measurements , In order to imitate the human process of chewing and tasting by robot chefs , The researchers attached a probe similar to a salinity sensor to the robot arm , The sensor can be moved to multiple positions , The location and data of the sample finally generate an image containing taste data .
▲ 9 Differences in salinity measurements before and after dish mixing
The researchers are ready to 9 A tomato scrambled egg with different salinity and tomato content , The robot then uses a probe to “ Taste ” dishes , And return the reading in a few seconds .
The reading is also used as the data information of taste to generate an image . The image is based on 2 Parameters , They are the number of test points and the size of the plate , The test points are distributed in a square grid .
▲ Conductance measurement histogram of dishes with different additives in different mixing stages .
During the experiment , The mixing process cannot be controlled. Every dish is exactly the same , therefore , During the experiment, the researchers asked the robot to taste each dish 3 Time , But the researchers used only the first and last taste to classify , To improve the repeatability of the experiment .
The first taste is on unmixed food , then , The robot mixes the samples for a few seconds and tastes them again , This measurement is for visualization purposes only . Finally, the robot is at the largest RPM Mix it again 60 Seconds later , Try this dish again , Chewing produces different readings at different times to further enrich the data information of taste map .
03. The salinity difference of unmixed ingredients is obvious , But we can't distinguish homogeneous samples
It can be seen from the taste map generated by the experiment , The unmixed sample will show a very obvious conductivity reduction area , That is, where there is no salt , These areas will have a very clear boundary with the egg . The data conductance distribution of the last sample is relatively uniform , Its conductance value is between that of tomato and egg .
▲ The taste mapping of the same tomato scrambled egg after three different mixing stages
Researchers said , Each stage of chewing produces significantly different taste data , This also provides additional information for the experiment .
According to this study , Compared with other electronic tasting methods , The ability of robots to assess saltiness has been significantly improved , These methods are usually time-consuming , And can only provide one reading .
The researchers say , By imitating the process of human chewing and tasting , Robots will eventually be able to produce food that humans like , And can be adjusted according to personal taste .
Abdulali say ：“ In our experiment , Robots can ‘ notice ’ Differences in when food is chewed , So as to improve its taste ability .”
Household appliance manufacturers cooperating with the project Beko Senior scientist of Muhammad Chughtai say ：“ We believe that the development of robot chefs will play an important role in busy families and assisted living families in the future . This result is a leap forward in Robot Cooking , By using machines and deep learning algorithms , Chewing will help robot chefs adjust their tastes according to different dishes and users .”
Even if some experimental results look intuitive , But in some homogeneous samples , The performance of the device is poor . in fact , Mixing different amounts of salt and tomatoes will eventually produce the same average salinity , Due to the specific structure of the sensor , It doesn't distinguish between... With the same chemical composition 2 dish .
Besides , The temperature of the dish 、 Whether the dishes are solid or liquid will affect the tasting process of the robot in cooking .
04. Conclusion ： Robot automatic cooking , It needs to be compared with human taste
before , Robot applications usually focus on process 、 In the fixed operation task ,“ Taste ” This task is often full of great uncertainty , Different people are different because of their own characteristics , It will also have different feelings about the taste of food , Enough to prove the difficulty of robot automation in the field of cooking .
In addition, human beings will be affected by the chemical composition in saliva and food temperature in the process of tasting 、 Personal preferences, etc , So the researchers say , Future work will focus on the study of saliva , Chemicals that may add lipase and amylase to human saliva for robots .
There are still deficiencies in the research of robot automation in the field of cooking , But turning information such as taste into visual data can promote further research . Researchers said , They will establish a standard based on the classified data , As a benchmark for comparing the psychophysical research of robot taste and human taste , And continue to expand the extension of this concept in the future .