In 2013, Fujitsu took a clean room inside a factory, which had formerly been used to design semiconductor chips, and converted it into a farm. Now, it’s using a network of sensors and machine learning to optimize the care and feeding of the farm’s single crop: low-potassium lettuce.
Roughly 10 percent of Japan’s population suffers from chronic kidney disease, and some of those residents receive dialysis treatments and must therefore limit their intake of potassium. Yet, this leaves them with few raw-vegetable options for their diets. Thus, growing a low-potassium lettuce was a promising opportunity for the company, especially because researchers from the Aizufujikako and Akita Prefectural University were interested in participating by sharing cultivation techniques.
The pilot project was adopted by the Reconstruction Agency and Ministry of the Economy, Trade, and Industry (METI) for “a 2013 advanced agricultural production system field trial, as part of a program aimed at increasing the competitiveness of the agriculture, forestry, and fishery industries and to aid in the recovery of the region affected by the Tohoku Earthquake,” according to a 2013 Fujitsu press release describing the trial. A consortium of companies participated, including ones from within the Fujitsu Group, and Aizufujikako Co., an indoor farming company, and Fukushima Medical University.
The pilot project, which ran until March 2014, leveraged Fujitsu’s Intelligent Society Solution Akisai Food and Agriculture Cloud, a service that aggregates and analyzes data related to the management, production and sale of agricultural goods. Some of the data the service uses is collected from sensors deployed on farm or range land.
At the end of the field trial, the pilot moved into full production mode. But the system was not operating as well as it could, Kaori Matsushita, senior director of marketing for Fujitsu’s global strategic alliances division, told IOT Journal via email.
“There were various data we could collect from sensors in the plant, such as illumination level, humidity, temperature, pH level of liquid fertilizer, CO2 density, etc. However, we were not fully making use of such data. The production know-how was totally dependent on the experiences and intuition of each operator in the plant,” she explained.
So early this year, Fujitsu began working with Microsoft to improve how it collects and shares data—as part of an effort, through the Fujitsu machine-to-machine software platform called Eco Management Dashboard, to make production of the lettuce, which is grown hydroponically, more resource- and time-efficient.
Sensors tracking all environmental conditions inside the rows of hydroponic growing channels help workers keep the lettuce in optimal conditions, while the other sensors ensure they apply exactly as much fertilizer is needed. Because the plants are grown indoors, pesticides are not applied.
“We attempted to analyze those data with Microsoft’s technologies,” said Matsushita, “to develop a knowledge database to derive the optimum combination of elements for low-potassium lettuce and use it in our production.”
The system relies on Microsoft’s Power BI data analytics program, running on Fujitsu’s Cloud A5 cloud platform, which it developed using Microsoft’s Azure cloud computing platform.
“Machine learning was used to analyze the correlation among various elements of the collected data, stream analytics provided the real-time selection of the critical and useful data for operators. The knowledge database from those analyses was made visual by Power BI so that people in the plant can use them on their Windows tablets,” she explained.