78 Smarter Farming in Japan

Overview. Unlike the first two case studies, this is an actual case rather than a hypothetical or composite one. It shares with the first two cases a focus on preserving valuable resources but in the radically different domain of farming.

This case concerns an initiative by Fujitsu, a Japanese technology firm, to apply “smart computing” and lean manufacturing techniques to the agricultural sector, which lags in technology use. Fujitsu is testing a “farm work management system” at six Japanese farms. In this case study we will focus on the farm highlighted in a 2011 Wall Street Journal story.^[1]

This test farm is located in southern Japan. It has 60 different crops spread over 100 hectares (about 250 acres), an area slightly larger than the central campus of the University of California at Berkeley.

What is being organized? Sensors are deployed in each of 300 different farm plots to collect readings on temperature, soil, and moisture levels. Video cameras also monitor each plot.

The 72 relatively unskilled workers on the farm are also managed resources. Each of them carries a mobile phone for communication, transmission of pictures, and GPS tracking of their location.

Why is it being organized? The highest-level goal for Fujitsu is to expand its reach as a technology firm by applying the concepts of lean manufacturing, statistical process control, and continual improvement to new domains. Farming is an obvious choice in Japan because it is a relatively unproductive sector where the average age is over sixty. It is essential that farms use more computing capability to increase efficiency and to capture and reuse the scarce knowledge possessed by aging workers.

The Fujitsu farm work management system supports numerous types of interactions to achieve these goals. For example, workers can send pictures of infected crops for diagnosis by an expert farmer in the farm’s office, who can then investigate further by studying recorded video from the affected plot.

As more farms deploy the Fujitsu system, the aggregated knowledge and sensor information can be analyzed to enable economies of scale that will allow separate and widely distributed farms to function as if they were all part of a single large firm with centralized management.^[2]

How much is it being organized? The current design of the system treats farm workers as relatively passive resources that are managed very closely. The system generates a daily schedule of planting, maintenance, harvesting, and other activities for each worker. At a daily wrap-up meeting the farm manager reviews each worker’s performance based on GPS and sensor readings.

The sensor data is analyzed and organized extensively by Fujitsu computers to make recommendations, both agricultural ones (e.g., what crop grows best in each plot and the work schedule that optimizes quality and yield) and business ones (the profitability of growing this crop on this plot of land).

When is it being organized? The farm work management system is continually organizing and reorganizing what it knows about the farm as it analyzes sensor and production information. In contrast, the information created by the workers is captured but its analysis is deferred to an expert.

It is conceivable that as the farm workers become more expert as a result of the guidance and instruction they receive from the system that they can be more autonomous and do more analysis and interpretation on their own. It is also likely that the inexorable forces of Moore’s law will enable more data collection and more processing of the sensor data at its time of collection, which might result in increased real-time information exchange with the workers.

How or by whom is it being organized? The physical organization of the farm, with 300 small plots of land with 60 different fruits and vegetables, is the legacy arrangement of the farm before the Fujitsu trial began. Because of the sizable investment that Fujitsu has made in the farm to deploy the system, it is likely that the farm manager defers to recommendations made by the system to change crop arrangements. So it is reasonable to conclude that most of the decisions about the organizing system are made by computational processes rather than by people.

Other considerations. Fujitsu built this system for managing farms, but there are several other resource domains with similar challenges about capturing and reusing operational knowledge: vineyards, forests, and fish farms come to mind.^[3] It will be interesting to see if the farm work management system can be made more abstract and configurable so that the same system can be used in all of these domains.

Farm crops, vineyards, trees, and fish pens do not move around, so a more challenging application of sensor technologies arises with cattle herd management. Nevertheless, sensors inserted in the genitals of a female dairy cow can trigger a text message to a herd manager’s cell phone when the cow is in heat, preventing the economic loss of missing a reproductive cycle.^[4]

Somewhat more remote domains for potential application of systems that combine sensor networks with workforce management include sales, field support, and logistics.