Home " News " MIT in practice: smart inventory management using paper sensors and AI by MOOS

MIT in practice: smart inventory management using paper sensors and AI by MOOS

Manually counting stock: this is not really feasible at high turnover rates. Given the current tight labor market, it is also difficult to free up employees for this purpose. There are advanced ways to automatically monitor inventory, but these are expensive solutions.

MOOS is therefore working on a more economical alternative: sensors made of special paper, which transmit data wirelessly and where, with the help of AI, it is possible to keep track of inventory remotely, in real time.

Paper has played an important role in inventory tracking for centuries. But simply noting on paper the number of products still available is, of course, of a completely different order than the paper sensors that MOOS is currently developing. This startup uses special, innovative paper for this purpose, on which a specific pattern has been printed with carbon. These are a kind of hooks through which a current is sent. When an object is placed on this paper, that pressure causes the hooks to make contact with each other, which then produces specific current signals. Then AI still has to be involved to analyze those signals and accurately trace them back to the product in question. Or rather, products. Because it involves larger quantities.

Innovative technical paper

"All of that listens pretty closely," stresses Xander Groesbeek, managing director at MOOS. "It took the inventor, Professor Koehly, who works in Canada, no less than 15 years to develop this technical paper. The three founders of MOOS, Marc Reunis, Pelle van Rees and Harm Hogenbirk, came up with the idea in 2020 to use this innovative paper for digital inventory tracking at retail companies. Manual counting costs those companies a lot of time, while they are often already understaffed and it is very difficult to find new employees in this day and age. And hanging up cameras and then having the inventory calculated via automatic image analysis turns out to be a rather expensive solution. Not only because of the equipment needed, but also, for example, because sending video images involves large data files. And RFID chips? These have to be manually stuck to products and removed again at checkout. Many human actions, and the chips themselves are also quite expensive. So in many cases that too turns out not to be an appropriate solution."

Subsidized AI exploration thanks to the MIT scheme

And sensors made of paper? The paper needed for that is relatively affordable and you can easily cut it to the size you need. In humid environments, such as refrigerators, a plastic coating is added so it can be used there as well. All low-tech yet. But the big high-tech challenge is in the area of AI. And so that was an important reason for MOOS to apply for the MIT scheme. After the requested grant was awarded in 2022, further exploration into the application possibilities of AI could begin.

More variables than thought

MOOS partnered with the Vrije Universiteit in Amsterdam. Francisco Blasques, professor of Econometrics and Data Science, then worked with his students and the data scientists at MOOS to develop a data model. Groesbeek: "That's when it really became clear to us how many variables we were dealing with. The AI system must not only be trained with information about the size and weight of products, but must also learn to take abnormal situations into account. For example, if one product leans against another, the weights of both products change. Also, correctly counting stacked products turned out not to be as easy as we thought. This is because the pressure that the second can puts on the paper is a lot higher than that of can number 25, at the very top. That's really a different curve. So with a stack, you can't simply add up the masses of all the individual products. Because of principles like that, it was really hard work to include all the variables correctly in the model. So together with the VU we spent much of 2022 working on that, and that was partly possible thanks to our participation in the MIT scheme."

Pilot phase started

Meanwhile, the technology is reliable enough to actually be applied in companies. So far, however, it has been a pilot project. Because if it doesn't work quite as it should, it is quickly solved at one location, but if MOOS then has to make the same adjustment at 99 other locations of a company, it will take a lot of time.

International interest

Meanwhile, there is no lack of enthusiasm. "I have a background in innovation and therefore always check the desirability, feasibility and viability of an innovation," says Groesbeek. "And by now it's abundantly clear to me that in terms of desirability, it's just fine. In fact, I've never experienced so much interest in a new service. And that doesn't just apply to retail companies. Hospitals, cleaning companies and hotel chains are also showing interest. So we are clearly developing a solution to a problem that many companies and organizations have. In addition, the international interest shows that good, affordable and easily scalable inventory management is not only a great desire of Dutch companies and organizations. For example, we have already had requests from Asia. But before we cross the border, we first want to make sure that we have our solution really mastered here in the Netherlands. Only when that is the case will it be time to scale up."

More information?

Visit the MOOS website. Or ask your questions viaemail.

Interested in the MIT scheme?

SMEs, including many startups, have an important economic and innovation role. Research shows that SMEs face many barriers in applying AI. The AiNed MIT AI call to bridge the gap between the knowledge base and application of AI innovations is supported from the National Growth Fund Programme AiNed. This scheme will be repeated annually through 2026. Our website will keep you informed.

Published: 30/06/2023
Focus areas:
Published: 30/06/2023