When will AI come to the commercial greenhouse industry?

Watching news reports on the COVID-19 pandemic one quickly realizes the importance accurate data plays in our everyday lives. Most industries are data-driven, whether this data relates to business management or specific production-related operations.

For the horticulture industry, data is an integral part of ensuring greenhouse facilities operate at maximum capacity. Unfortunately, growers have limited access to the data being collected in their greenhouses and are unable to utilize this data in a way that could help them increase operation efficiency and yields.

"The data being collected by greenhouse growers is being siloed, meaning the data is stored in different closed systems," said Ken Tran, founder of Koidra LLC. "These closed systems don't communicate with each other and growers do not have a way to unify the data for whatever purpose or whatever analysis they might want to do. This can be greenhouse environmental data, biological data or business management data.

"For climate control data, it is not uncommon to have this type of data living in different systems as well. For example, growers can have climate control data such as temperature and relative humidity in one system. The data for lighting supplied by another company may be in a different system. There are many lighting companies that provide their own controls. Most companies that growers are familiar with don't want to expose the data that is being collected in a way that the systems can talk to each other."

Limiting data analysis

Another critical problem with data being siloed is even if the growers' data is in one system growers may not be able to do data analysis. In most cases, the only way the data is available is to export it to Excel files, which is very limiting.

"Climate control data is collected automatically and put in a system," Tran said. "Depending on the type of climate control system that is used, data is collected in a database that is hidden under the interface of the climate control company. Growers are limited by what the climate control company interface will provide.

"If growers want to use the data, the systems can only provide limited capability in terms of data analysis. Growers may be only able to look at the data from one season's crop. But the climate control software will not allow growers to build predictive models from the data. The only way growers can build predictive models is to be able to access the database. Growers should be able to use their data however they want."

Giving growers access to their data

Tran said most growers are dealing with multiple databases depending on the type of data that is being collected.

"If a company's expertise is in climate control management it makes sense that the company doesn't focus on biological management data or business management data," he said. "The best way to move forward is for these companies to open their data interfaces to the growers so that growers truly own their data. This would allow growers to access the databases so that they can hire third party companies to do data integration.

"Even though this is the best way for growers to access their data, it's not the only way. Koidra offers data integration service as part of its umbrella autonomous greenhouse product to overcome this problem. It doesn't necessarily require the companies maintaining the data to open their data interfaces."

Currently in the horticulture industry the only growers looking at having more control over the data they are collecting are a few well-funded indoor vertical farms.
Photo courtesy of Ken Tran

According to Tran, this situation is not unique to horticulture and is common in industries that have fallen behind in the technology curve. Some industries are more advanced when it comes to being tech savvy. Agriculture and some older manufacturing industries may have issues with the digital transformation curve.

Tran said many climate control companies see the trend toward artificial intelligence (AI) and they want to be able to expand their capabilities to the growers.

"The notion of data management and leveraging data analytics and machine learning are new," he said. "A few years ago these topics weren't even being considered by these companies. I haven't yet seen the need for data management. There hasn't been a demand from the growers to have access to this data. Even if they had access to this data what would they do with it? Most growers don't have the capabilities to build their own predictive models.

"Some growers would like to work with companies that can do the analytics. Only a very few well-funded indoor vertical farm companies have chosen to develop complete in-house systems so that they can have more control over their data. Many companies want to have more control over their data and would like to do more with their data."

Tran said growers can only truly own their data when:

1. They can store and transfer their data however they want.

2. They can query their data to get better insights however they want.

3. They can use whatever tools on their data as they want.

All of these require a programmatic interface to the data storage systems, which is currently lacking.

Building an autonomous greenhouse

The internet of things (IoT) is a network of interconnected devices that is embedded in sensor software that enables them to collect and exchange data making them responsive.

"IoT can be thought of as a system that enables automated, real-time and high-frequency data collection," Tran said. "One type of device is a temperature sensor. Using this sensor there wouldn't be a need to have humans collecting and inputting data. The sensor is connected to a network and it can transfer the data to the growers' database automatically. It can communicate temperature data to growers or to their systems. IoT can be thought of as systems that enable automated, real-time and high-frequency data collection.

"Every business is connected to the internet. With the right data management infrastructure, growers should be able to get the right information at the right time from anywhere and on any device. Once full situational awareness of the business occurs, the business can effectively be managed remotely."

Ken Tran is developing a commercial artificial intelligence program that will be adaptable to a variety of crops grown in vertical farms and greenhouses.
Photo courtesy of Mike Evans, Virginia Tech

Manual data collection or no data collection at all is the opposite of IoT. Manual data collection is not done in real time, is done infrequently and is expensive to do.

"IoT is an enabler for high-speed, high-volume and low-cost data collection," Tran said. "This would allow growers to develop AI applications that leverage big data. AI capabilities can only be realized after the right information infrastructure (IA) is created. As the AI community tends to say, "There is no AI without IA. The fact that IoT is being adopted heavily in the greenhouse industry makes AI even more attractive."

Tran said what will drive the development of autonomous greenhouses is what greenhouse owners and operators want.

"They want higher profits and yields and lower operational costs," he said. "During the first International Autonomous Greenhouse Challenge in the Netherlands it was shown that an autonomous greenhouse program can produce higher yields and higher resource usage than expert growers."

During the competition the winning Project Sonoma team, led by Tran, outperformed a team of expert Dutch growers. The Sonoma team produced more than 55 kilograms of cucumbers per square meter. The net profit on the cucumbers for the Sonoma team was 17 percent higher than for the team of Dutch growers.

But not every autonomous greenhouse is efficient.

"An autonomous greenhouse can be less efficient than a good grower," Tran said. "This was shown by the results of the Autonomous Greenhouse Challenge. The Sonoma team was the only one that outperformed the expert growers. All the other teams did worse than the growers.

"All companies want their businesses to be more automated, more scalable and more efficient. This is where AI, built upon rich IoT and crop management data, can help. A good AI program not only provides the value of automation, but higher efficiency as well."

Is the commercial greenhouse industry ready for AI? Tran thinks so.

"It's already happening, demonstrated by the Autonomous Greenhouse Challenge," he said. "Innovative companies that offer both data integration and AI services can help make that reality faster for greenhouse growers."

For more: Ken Tran, Koidra LLC, (512) 436-3250; ken@koidra.ai.

This article is property of Urban Ag News and was written by David Kuack, a freelance technical writer in Fort Worth, Texas.

Photo: During the first International Autonomous Greenhouse Challenge the Project Sonoma team, led by Ken Tran (second from left), showed that an autonomous greenhouse could produce higher yields and profits than a team of expert growers.
Photo courtesy of Ken Tran

Source: Urban Ag News