Reviews and recommendations are unbiased and products are independently selected. Postmedia may earn an affiliate commission from purchases made through links on this page.
A burger-flipping, risotto-stirring robot ‘chef’ may seem like a science fiction dream come true. But the reality is, cooking can be challenging for robots, according to researchers from the University of Cambridge. Today, some restaurants use robots to power fry stations, load dishwashers and cook pasta. Robotic tools are projected to play an even more important role in the kitchens of the future, but their cheffing skills still lag behind humans.
tap here to see other videos from our team.
People learn to cook by watching others, whether virtually or in person, the researchers highlight. They trained their robotic chef to learn as many have in the internet age, by watching cooking videos.
“We wanted to see whether we could train a robot chef to learn in the same incremental way that humans can — by identifying the ingredients and how they go together in the dish,” Grzegorz Sochacki from Cambridge’s Department of Engineering, the paper’s first author, said in a statement.
The researchers developed eight “simple” salad recipes and filmed themselves making them. After programming the robot with their cookbook, they trained it to identify the ingredients and actions used to prepare each of the salads: fruits and vegetables such as apples, bananas, broccoli, carrots and oranges, and objects including the cook’s arms, face, hands and the knife they used for chopping.
The robot was able to correctly identify the recipe 93 per cent of the time, though it was only able to detect 83 per cent of the human cook’s actions, the researchers report in the journal IEEE Access.
It was also able to identify the difference between a new recipe and a variation — such as when the human cook doubled the quantities or made a mistake. After watching a video demonstrating a ninth recipe not in its programmed “cookbook,” the robot correctly recognized it as new, added it to its collection and recreated it.
“It’s amazing how much nuance the robot was able to detect,” said Sochacki. “These recipes aren’t complex — they’re essentially chopped fruits and vegetables, but it was really effective at recognizing, for example, that two chopped apples and two chopped carrots is the same recipe as three chopped apples and three chopped carrots.”
As it stands today, the style of cooking videos most popular on social media wouldn’t suit their robot chef’s learning style, Sochacki added. It would have trouble making sense of the rapid pace and partially concealed ingredients — a potato held in a closed hand while being shaved on a mandoline, for instance, instead of being held up to the camera so the robot could clearly see the whole vegetable.
“As these robot chefs get better and faster at identifying ingredients in food videos, they might be able to use sites like YouTube to learn a whole range of recipes,” said Sochacki.
Programming robots to make a variety of dishes is expensive and takes time, the researchers said. Autonomous cooking robot Nosh promises to whip up more than 120 dishes and costs 50,000 rupees ($814) to preorder; the Moley Kitchen, is programmed to make 5,000 recipes and do the washing-up, and costs at least $412,200, the Guardian reports.
Their results, the researchers note, “demonstrate how video content can be a valuable and rich source of data for automated food production and could enable easier and cheaper deployment of robot chefs.”
Recommended from Editorial
Shrinkflation is getting worse in Canada — and it could mean paying more taxes
This Montreal restaurant is Canada's greatest, according to annual list of 100 best