How do you like your eggs in the morning? If your answer is “whipped into a reliably good omelet by a robotic arm” then boy, do we have a story for you.
A team of engineers at the University of Cambridge, UK, have trained a robot to prepare an omelet, as reported in the journal IEEE Robotics and Automation Letters. You might be wondering why on Earth, beyond feeling peckish, anyone would bother to create such an automaton but, as the researchers point out, assessing whether a robot has successfully cooked a meal is an interesting source of debate.
“Cooking is a really interesting problem for roboticists,” said lead researcher Dr Fumiya Iida from Cambridge’s Department of Engineering in a statement. “As humans can never be totally objective when it comes to food, so how do we as scientists assess whether the robot has done a good job?”
The researchers used machine learning to try and teach the robot both how to complete the challenging manipulations involved in cracking, whisking, and crafting an omelet, but also how to account for complex matters of taste. While robots have been used in production to create food items such as cookies and pizza, they are not used in the process of actually cooking, which is subject to many variables and therefore much harder for artificial intelligence to master.
In order to learn about taste, they used a statistical tool called Bayesian Inference to incorporate feedback from omelet taster data samples and use this to tweak the algorithm, which would dictate how the robot altered its cooking technique. The process essentially takes qualitative information regarding how good an omelet was and turns it into quantifiable data that the robot can use to hone its craft. However, this in itself caused some problems.
“Another challenge we faced was the subjectivity of human sense of taste – humans aren’t very good at giving absolute measures, and usually give relative ones when it comes to taste,” said Dr Iida. “So, we needed to tweak the machine learning algorithm – the so-called batch algorithm – so that human tasters could give information based on comparative evaluations, rather than sequential ones.”
The result? “The omelets, in general, tasted great – much better than expected!” said Dr Iida.
The benefits of going through the rigmarole of teaching a robot arm how to cook is that once the engineering has been perfect the machine learning can be applied to a host of robotic chefs, churning out consistent results on a large scale. Who knows, maybe when lockdown ends and restaurants reopen, we’ll find ourselves being served by robot chefs. It would still be less weird than sitting at a table full of diners sporting these...
1