A smarter, more capable Flippy robot knows how to fry 19 things
The restaurant industry has never been for the faint of heart, what with the razor-thin profit margins and continuous churn of employees. Combine that with the economic devastation wrought by the COVID-19 pandemic and, well, it’s no surprise that tens of thousands of eateries across the country have shuttered permanently over the last seven months alone. However, Miso Robotics (makers of Flippy, the burger-flipping robot chef) argue that the path back to financial stability for America’s restaurants will require an autonomous revolution.
Buck Jordan, founder and CEO of Miso Robotics, points out to Engadget that the switch from full-service dining to exclusively take out and delivery has many restaurants “operating a loss and just trying to hang on,” especially when delivery apps like UberEats and DoorDash take upwards of a 30 percent cut out of each order.
“You can see why closing up shop is really the only option for many once your revenue is coming from delivery orders,” he continued. “There’s just so much pressure on the bottom line.”
That’s why the company announced on Tuesday that it is making the latest iteration of Flippy available commercially across the globe. But this isn’t the Flippy that we saw at White Castle in July. This one is smarter, more capable and armored against airborne infection.
To start, the company has streamlined its proprietary ChefUI software, to “assist kitchen workers with operational interactions and workflow,” according to a Tuesday blog post. With this system, food workers will be able to see not only which orders are queued up for cooking but, thanks to the integration of Intel RealSense depth cameras, can also monitor the food’s internal temperature, guard against undercooked foods from making it out the kitchen, and even adjust the queue to ensure that everything in an order finishes cooking at the same time.
For example, Jordan notes, it only takes about 3 minutes for fries to cook, compared to the 10 minutes for a burger so rather than have the fries sit under a heat lamp for the 7 intervening minutes before the burger is done, Flippy is smart enough to know to only put the fries in their oil 3 minutes before the burger finishes cooking. Now if there were only a way to keep those fries piping hot during the 20 minutes it takes for the delivery driver to get them to you.
To date, Flippy can cook 19 different items — anything from corn dogs to chicken tenders — but that number is sure to expand now that the robot is trainable, thanks to an upgraded machine learning system developed in-house by Miso. The company first gathers around 1000 shots of the dish in various states of doneness. “One of the challenging things about food is that food changes states so this brown enough to change of shape,” Jordan said. “It’s not uniform.”
The company then feeds those initial photographs into a computer. “What it’s doing is constantly permuting those pictures to create, like, 1 million pictures of chicken nuggets and then uses that to train the algorithm to identify things with four nines accuracy.” All of these permutations are done in order to ensure the system understands what it’s looking at under any available lighting conditions.
And to help at least partially mitigate the dangers of contracting Covid while sweating over a hot griddle shoulder to shoulder with the rest of the line cooks, Flippy can be outfitted with a glass partition shield to help keep the human staff’s exhalations separate from the food prep. What’s more, Flippy has obtained its NSF safety certification to ensure “that you know [what the robot’s] materials are made out of, you know they won’t cause harm when cooking something and certainly we don’t want that any parts falling in the oil,” Jordan remarked.
Interested investors and restauranteurs can head over to invest.misorobotics.com for more information on pricing and availability.
https://www.blogsmithmedia.com/www.engadget.com/media/feedlogo.gif?cachebust=true Engadget RSS Feed https://www.engadget.com/rss.xml