The future is here

Photo courtesy of Wageningen University | Photos courtesy of IUNU

A conversation with IUNU’s Allison Kopf

In June, IUNU released its new closed-loop AI irrigation system Intelligent Setpoint Control for LUNA AI. Upon its release, the company explained, “by pairing climate data with comprehensive plant performance data using computer vision, LUNA AI provides growers with ideal crop strategies and executes the selected strategy autonomously.”

IUNU also says it is the only closed-looped AI system for autonomous climate control in horticulture. Using computer vision, the crop strategies take into account how a crop is performing in real-time and makes adjustments to focus on driving the desired outcome in five-minute intervals. IUNU worked on the project with Wageningen University, as well as BrightFarms. Earlier this year, IUNU also released a new tomato imaging system that allows growers to collect more comprehensive data on their crops.

Allison Kopf, the chief growth officer at IUNU, discusses the role of artificial intelligence in horticulture, the launch of Intelligent Setpoint Control and why they focused on climate control in this product launch. This interview was edited for style and clarity.

Produce Grower: When you think about AI within this industry, where do you see AI? And why is it important for IUNU to be on the forefront of AI?

Allison Kopf: I think it's a really good question because it's so timely right now. In the world we are in, AI is getting a lot of attention because of the accessibility of tools like ChatGPT, so the question obviously becomes how do we use it in our space.

I think it's important to acknowledge that there's a full spectrum of things that fall into what AI is, as well. The major thing that we are focused on as a company is the computer vision side of this, which has tremendous application for the horticulture industry. Because computer vision, at the heart of it, is replicating what we do on a daily basis. This concept of having [AI] sense everything around you, applying it to understand what you see and then connecting that collected information to your brain to put that into analysis where you're located and what you're experiencing — we can do that in horticulture. And that process can actually be better than what we do as humans in a lot of ways.

Think of rote processes, things that are manual and repetitive and happen every day on the farm. It does not matter the crop — it could be lettuce, ornamentals, tomatoes, whatever — you have to count stuff all the time. Well, computers can do that better and more accurately than we can, and we can do it comprehensively for every plant if we use computers to do it. Scouting and looking for issues is a process that absolutely can be replaced by computers. So if we fit in where computer vision has a lot of value, do the data collection comprehensively and do it accurately across the farm, it has huge value applications for anyone in the horticulture industry.

Then you take AI and you have all of this data, so what can you do with it? And that's where you have to look at the return on investment for growers, because if you break that down into practical things, there can be a really good return on investment. Let's control crop strategies. The machines that execute crop strategies in a way that is autonomous and based on all of this data and close that loop by seeing how the plants actually perform to their environment are what can change the game for growers, if the price is right.

PG: How did you zero in on climate control as the function of Setpoint?

AK: There are a few areas for autonomous growing that make some sense — irrigation control, lighting control, climate control. Those are your mechanical systems that drive the crop yield. Part of this is that a lot of the systems already exist or don't have a reason to exist. Lighting control is a good example, where it doesn't need to be autonomous. You're turning them on and off. Growers, at the end of the day, don't see a need for that now because what works now works great. Similarly, irrigation control already exists — we have smart irrigation control. If you look at what Priva's offerings are or Ridder's offerings are, smart irrigation already works pretty well.

PG: What is it like working with growers and researchers on this? How applicable is their feedback?

AK: I think that's really important [that we work with growers]. Some of this is going to be brand new to folks. It's going to be kind of mind blowing. [A grower I was talking to] walks and measures the growth of plants on a daily basis and he asked, 'how many plants can you [automate] this for?' and I said, 'every single day and every single plant.' And he didn't believe it.

In computer vision, there's a common process called annotation where you are training something that's new. When it annoates, it learns.

We have an annotation team, so we do all of that work in-house. But you still want to work with the growers because they know what they are looking for. They see things on a daily basis and really what you're taking is this process of somebody that's been a skilled grower for 30, 40, 50 years and trying to replicate in technology what their eyes see. That's not an easy process, so having [growers] be a part of that process is something that's important for us and for them, as well.

Automated lettuce production at Wageningen University

Wageningen University, a leading Dutch university specializing in agriculture, has been on the forefront of autonomous growing.

Through their Autonomous Greenhouse Challenges, the university has stepped up what encompasses a fully automated growing cycle. In their most recent challenge, teams were tasked with growing lettuces in two crop cycles fully autonomously with an AI algorithm. No human interaction or interference was allowed in the Wageningen University & Research greenhouses in Bleiswijk, The Netherlands.

Each of the five teams took on different strategies. Each team had the same sized growing space to start with. Every growing space was equipped with standard, industry-grade climate sensors and equipment. But beyond that, everything was different. The different growers spaced out their plants differently, deployed different climate strategies, different temperature ranges and different LED lighting.

The end result was telling. Autonomously grown lettuces reached the desired weight for a crop that could be sold in stores. But, at the same time, each growing team had three-times higher electricity costs than a rough industry average. Most of the teams also had higher heating costs. So, no team achieved the goal of a net profit.

That, though, doesn’t mean it’s not possible — it’s going to take time and refinement of this process to make it possible and for the cost to make sense for a company’s bottom line.

Elevate greenhouse manufacturing

Editor’s note: A version of this article ran in the July issue of Greenhouse Management.

At the 2023 NGMA annual meeting, Wil Lammers, Ridder's managing director in the Americas, gave a presentation titled "Shaping the Future of the Industry: Insights on Emerging Trends and Technologies" to attendees. On one of the slides, which detailed different components growers can consider for their greenhouse, "crop management" was positioned at the middle.

The reason is simple: Any and all growing businesses depend on how well crops can be managed.

"Crop management is always at the center," Lammers says. "If you look at a greenhouse operation, how well you run this determines your financial performance, your ability to have positive cash flow. That's for any greenhouse, whether it's orchids or bedding plants or tomatoes or peppers. Whatever you grow, wherever you grow, it's about creating the most optimal for the 12 months of the year you are growing. That's how it all comes together in the end."

That said, growers should also consider a cost-benefit analysis. Typically, the more automation and control growers build in, the higher the cost in the short term. So, growers have to consider a) how much up-front costs they are willing to take on and b) what level of automation and control makes sense as a long term investment.

Choosing the right components

Lammers says there are a number of ways to figure out which components are right for a given business.

Even in a controlled environment, the needs of a grower in Arizona are different than a grower in Florida. Both are typically warm environments, but Arizona is typically arid while Florida is more humid.

"Take a greenhouse with a pad wall for cooling, for example," he says. "That will work very well in a dry climate. But if you take the same cookie cutter greenhouse and put it in Florida with 90% humidity, the ecosystem is completely different because you don't get the same cooling effect because of the humidity outside compared to Arizona or Colorado where everything is dry. So you need to pick the components that allow you to grow on the inside optimally based on what is going on outside."

Potential labor savings

Labor, Lammers says, remains one of the biggest concerns for growers, both from a cost standpoint and from an availability standpoint.

For some growers, labor costs are going up as minimum wage increases or they need to raise wages to be competitive with other jobs. For some growers, it's become harder to find employees who want to work in a greenhouse setting and do manual labor. Some growers have turned to H-2A programs to fill labor needs, but there are costs associated with that as well.

That is where automation can come in, Lammers says. There will never be a world where growers don't need some form of human labor to operate day-to-day. But with the right equipment, they can cut down on the need for labor for some tasks. Then, labor growers do have can be more efficiently applied to other tasks.

An example Lammers cites is environmental controls. Even at scale — in facilities over 100 acres — the technology exits that allows growers to not have to manually check and adjust the environment.

Instead, it can be managed automatically with some check-ins and adjustments needed based on how the plants are growing. But with less time spent on those tasks, growers can then focus on other tasks.

"What we can do with AI and autonomous greenhouse control," Lammers says, "is give a top-notch greenhouse grower the capability to cover more acres with less people."