How artificial intelligence is transforming the produce industry
Artificial intelligence software is everywhere these days. It’s used to inform smart assistants such as Alexa and Siri.
It’s used to pilot self-driving cars and inform marketing chatbots. It’s used in transcription software and manufacturing robots.
AI is especially advantageous when it performs tasks that humans find less than desirable. With the addition of artificial intelligence, robots can even play a role in food processing plants to swiftly address issues such as high staff turnover and labor shortages.
AI is not innate intelligence but rather the ability to solve problems by learning through experience. Deep learning software enables robots to quickly adjust to new inputs and to perform human-like tasks at incredible speed. In the food industry, robots can help processors and food handlers optimize quality, reduce waste and improve flexibility.
The following three companies provide examples of how AI can be used to improve processing tasks.
Micropsi Industries’ MIRAI increases flexible automation
Micropsi Industries recently launched new artificial intelligence-based software — called MIRAI — that allows robotics companies to add hand-eye coordination that improves robotic handling of product. MIRAI is compatible with a range of robots, including those made by Universal Robots and FANUC. In a recent interview, Dominik Bösl, managing director at Micropsi Industries, explained how the technology works.
Conventional robotic solutions have always been challenged by variance, such as recognizing differences in shape and size. MIRAI solves this problem.
“We can deal very well with variants in space, in shape, and also in position, so recognizing objects, finding objects, trying to grasp objects, that is where — not just MIRAI, but AI in general — will really unleash a completely new level of robotics,” he said. “Traditional robotics was falling short in that respect.”
Bösl said there has been much demand for these types of robotic solutions in the food processing sector where shape and size vary, and the ability to grasp objects has been challenging. Using its camera technology, MIRAI could be used, for example, to differentiate between ripe and unripe fruit, or to help line carrots into packaging.
What makes MIRAI most interesting for robotics companies is that skills are trained rather than programmed. The robot learns from humans who demonstrate a given task by manually guiding the robot by the ‘hand.’ The accompanying camera takes 15-25 images per second to capture that movement until it eventually becomes a skill.
“We can teach the robots in hours what others cannot program in weeks or months,” said Bösl.
Bösl believes AI has to the power to enable completely new applications in the world of automation, especially where automation has been traditionally limited by flexibility. Flexible solutions, he said, will bring robotics and automation closer to objects, particularly in farming and food processing. It also makes the technology more feasible and usable for smaller companies which may not have seen the value in investing in robotic solutions before now.
“The faster the retraining becomes, the more flexible you can be in your application,” he said. “Today, the robot could be sorting bagels and doughnuts, and tomorrow, it could help you in cutting carrots.”
Apera AI develops software that tackles flexibility and variability
Vancouver-based Apera AI has been working on similar software to guide robots in doing specific movements, including identifying objects and then handling or manipulating those objects in a specific way. Eric Petz, Apera AI’s head of marketing, said they’re working with integrators who want to use the software in processing in the food and beverage space.
For instance, Apera AI has been working with a company that wants to be able to use a robotic solution to select and place a number of blueberries onto a processed salad. The software allows the robot to first make sense of disorder and then complete the set task.
The company has clients interested in finding solutions for handling apples to onions, as well as product packaging. Apera AI works closely with companies such as Universal Robots to develop these solutions.
Where the software still struggles is handling malleable objects, such as a bag of frozen peas. Robots can’t yet be programmed to handle these types of products without the risk of damage. But where the software excels is speed. Currently, Apera AI offers one of the fastest solutions currently on the market. One vision cycle, which includes identifying the object, sending instructions to the robot and performing the task, takes 0.3 seconds.
“That’s like the world’s fastest right now,” Petz said. “That’s state of the art.”
To give that some context, the standard marker for most robotics companies is 2,000 picks per hour, which means a vision cycle of 1.8 seconds. Apera AI software can be six times faster than that standard.
Petz said Apera AI clients — whether they’re in the food and beverage, automotive or aerospace industries — all face major challenges when it comes to finding labor. He sees the company’s software as a critical solution to these shortages, as it allows robotics and automation to continue where humans left off.
“When you’re talking about dull, dangerous or dirty jobs that they otherwise can’t fill, automation is the natural choice to replace human labor in those cases,” he said.
Afresh uses AI to cut food waste in supermarkets
Supermarket chains have an obligation to their customers to deliver fresh produce, at least, if they want to keep those customers. But keeping up with demand, especially amid seasonal shifts and market disruptions, makes this a complicated task.
Through the adoption of AI, San Francisco start-up Afresh aims to simplify that task — and reduce waste in the process.
Co-founder and COO Nathan Fenner said Afresh was started on two foundational principles. The first is that fresh food is the future of grocery. The fresh sector, he said, has experienced significant growth, driven in part by health-conscious consumers and e-commerce. It was and still is the critical strategic differentiator for how supermarkets distinguish themselves from other retailers both in their retail outlets and online.
“We really came to the conclusion that in order to win going forward, the most strategic areas for grocers were their fresh departments,” said Fenner.
The company’s second foundational principle is to create a solution specifically designed for the fresh market, especially when considering supply chain and inventory management software. That technology, he noted, was built for the non-fresh side of the business.
“A lot of the technology out there had actually been developed as a horizontal solution that was made to serve a bunch of different retail subverticals,” Fenner said, adding that the same software was being used for auto parts, clothing and electronics.
“Unsurprisingly, when you bring technology that’s built for hard goods that come that have barcodes into the hyper-perishable world of fresh, it doesn’t perform very well,” he continued.
In addition to being perishable and sold at random weight, fresh products are often processed in-store. These complexities make it far more difficult for a rigid traditional system to handle accurately.
“What we saw was that this most strategic part of the business was being essentially governed by manual intuition and gut instinct,” Fenner added. “And not surprisingly, that drives a lot of inefficiency in the supply chain, particularly in a labor market where there’s super high turnover.”
Enter Afresh’s store-level replenishment solution for produce, a tool that optimizes how store level produce managers order produce into their stores each day. The aim is to not only minimize food waste, but also to ensure that product is available on the shelves when it’s needed. The start-up leverages AI technology to solve the problem of decision making in an uncertain environment.
Fenner said Afresh technology offers several benefits to retailers, including an average reduction in shrink of about 25% and optimal inventory levels.
“Afresh is doing a really good job of keeping the sales floors full but the back rooms empty,” he said. “And the net of that is less shrink, but also more sales, more turns, fresher product, a better end consumer experience.”
Today, Afresh AI software is in use in 3,000 grocery stores across the U.S., with about 7% of produce ordered in U.S. grocery stores ordered through that software, Fenner said.