Waste Not
Princeton, Microsoft Co-Develop Wireless Solution to Combat Produce Waste

When it comes to fresh produce making it from farm to market, unfortunately the old adage holds true: One bad apple can indeed spoil the bunch — as a lot of good, edible fruits and veg are often thrown out with the bad.

But researchers from Princeton University and Microsoft Research may have developed a solution — a fast and accurate way to determine fruit quality, piece by piece, using high-frequency wireless technology.

The new tool promises to help cut food waste by giving suppliers a way to sort fruit based on fine-grained ripeness measurements — enabling distributors and retailers to save good fruits and veg from being thrown out with the ugly or spoiled items deemed not fit for market, and move ripe fruit to the front of the line.

According to the new Princeton-Microsoft study — which was presented at the recent 2023 ACM MobiCom conference — conventional methods to determine ripeness of produce tend to be unreliable, overly broad, too time-consuming or too expensive to implement at large scales.

As Yasaman Ghasempour, assistant professor of electrical and computer engineering at Princeton and co-author of the study, explained: “There is no systematic way of determining the ripeness status of fruits and vegetables. It is mostly random visual inspection, where you check one fruit out of the box on distribution lines and estimate its quality through physical contact or color change.”

But this kind of visual inspection often leads to poor estimates, she said. Rather than rely on how the peel looks or how it feels to the touch, wireless signals can effectively peek under the surface of a piece of fruit and reveal richer information about its quality.

According to recent ReFED research, the US produced 91 million tons of surplus food in 2021 — 15 percent (13.8M tons) of which was due to spoilage. This uneaten food represented about 38 percent of the country’s total food supply that year; and of that, about 36 percent went to landfills — where it releases potent greenhouse gases including methane, which exacerbate climate change. The United Nations estimates that 13 percent of food produced worldwide is lost between harvest and retail, while an estimated 17 percent of total global food production is wasted in households. The new study’s authors assert that inefficiency at this scale is only seen in the food industry and that automated, noninvasive and scalable technologies can help greatly reduce that waste.

“When we look at the global challenges around food security, nutrition and environmental sustainability, the issue of food waste plays a major role,” said Ranveer Chandra, Managing Director of Research for Industry and CTO of Agri-Food at Microsoft. “If we could reduce food waste, it would help feed the population, reduce malnutrition, and help mitigate the impact of climate change.”

The team, led by Ghasempour and Chandra, developed a system for determining ripeness using wireless signals in the sub-terahertz band that can scan fruit on a conveyor belt. The sub-terahertz signals, between microwave and infrared, can pick up minute details about the fruit — including, for example, the amount of sugar and dry matter content.

As fruit continues to ripen after harvest, its physical, chemical and electrical properties also change — bananas turn yellow; grapes wrinkle; avocados darken — and once food makes it home, innovations such as Makro's Life-Extending Stickers can guide consumers on how to use different items at different stages of ripeness. But for a lot of fruit, it is hard to know how those outward markers correlate to actual ripeness or quality — as anyone who has bitten into a perfectly shiny, red apple only to find it mealy and dry can attest.

When a sub-terahertz pulse hits a piece of fruit, some frequencies are absorbed, others get reflected and a lot of frequencies do a little of both with varying intensity. The reflection creates its own signature; and by procuring data from these signatures, the researchers were able to accurately determine a fruit’s ripeness.

As Ghasempour explained, fruits’ seeds, pulp and skin; as well as variations in size, thickness, orientation and texture added complexity to the issue, so “It was really challenging to develop a model for this. So, we performed some wave modeling and simulations; and then augmented those insights with the data that we collected.”

For the experiment, the researchers used persimmons, avocados and apples. Fruits with smooth skins are easiest to measure — the uneven texture of an avocado peel reflects a weaker signal and produces unwanted effects. But the researchers found ways to get around this and say that with enough data, the method can be applied to most fruits.

They believe this tool can also be used to other types of foods, including meats and beverages, by using different kinds of physiological markers — which could be a game-changer for food-safety monitoring and consumer choice.