NatureInterface Co.,Ltd. FAX:+81-3-5222-3581
NatureInterface > No.07 > P026-030 [Japanese]

IT Agriculture Based on Measurement and Control Technology -- Jyun Kawauchi

IT for Agriculture Based on Measurement and Control Technologies


Yamatake Corporation, Product Business Management Board Member


Assistant Professor, Chuo University, Faculty of Sciences and Engineering, Department of Precision Machinery Engineering

Approach to IT for Agriculture

-- Your company°«s strength is in the field of control systems, ranging from control of individual instruments through to total-control systems, such as air-conditioning systems in buildings and control systems in plants. Why are you also engaged in the study of IT for agriculture? As a new field to specialize in, how is agriculture appealing to you?

Kawauchi (K)---When you try to control something, objects of which you know the mechanisms are easy to control, but it is quite difficult to control an object such as agriculture of which the mechanism cannot be specifically explained, hence most tasks have to rely on human expertise rather than technology. However if we had a technique to control agriculture, we could expect steady, high-quality harvests. We feel a strong potential in this aspect.

Exemplary farmers have excellent skills and knowledge in cultivation. They know the timing to supply the appropriate fertilizer and amount of water in order to grow higher quality and more abundant harvests. Unlike manufactured products, pursuing good taste is very important in the food industry. We believe that with our control technology, it would be possible to consistently produce high-quality crops. As eating is one of the basic human desires, there must be high expectations for realizing this application (Fig. 1).

-- That°«s true. When you eat something, it is much better to eat something delicious.

K---But the question arises whether the number of such farmers will increase in the future. They have acquired their knowledge through years of experience and also their talent. But if agriculture can somehow be automated, with control equipment replacing the farmers°« expertise, people will perhaps enjoy good food more often and at reasonable prices.

Object Modeling in Agriculture

---So the key lies in modeling the skills and knowledge of the farmers.

K---Yes. From the viewpoint of control, we would like to know the mechanism of the farmer°«s action. Organisms have the power to live on their own, and they grow under certain conditions but to a certain limit. Without any human intervention, crops will be of poor quality and low yield. That is the big difference between crops that grow in the wild and the ones grown by farmers. It is a matter of manipulating good conditions and environments for plants.

What do exemplary farmers observe and judge? Functions of the human senses are very complicated. A person senses a condition with his/her perception and hearing. Then he/she chooses the appropriate action according to each situation. But it is difficult to comprehend this whole process, unlike variables managed in a factory, which are quite explicit. Information obtained by the five senses cannot be easily transformed into numerical data. In fact, farmers themselves do not consciously recognize their special knowledge.

---So how would you actually receive such information detected by the five senses? Would a machine be used to take measurements?

K---No, that would be too difficult. In order to substitute a machine for the five senses, we would have to study alternative sensing methods first. Secondly, we would need to examine ways to match the information obtained by machines to the judgment of the farmers. This examination will be an indispensable step towards the automation of agriculture.

---I agree with you.

K---A concept like this could be applied to the idea of a soft sensor (Fig. 2). Our company calls it "TCBM" and we believe that by utilizing it we could contribute to the realization of automation in agriculture. By investigating how farmers make judgments and knowing what kind of sensor to use, we should be able to automate this process.

Control Technology Applicable to IT for Agriculture: TCBM

-- What is TCBM?

K---It stands for Topological Case-Based Modeling. In general PID method is used to control a simple object. But in order to achieve multiple purposes, for example, maintaining the most comfortable temperature for all workers in a plant or office while saving energy, using only PID is not sufficient. TCBM was developed for the purpose of controlling objects with complicated characteristics through a modeling process (Fig. 3).

---Is TCBM a commonly used term?

K---It is our original terminology. This technology uses modeling to predict output from inputs. For complicated objects, even if the purpose of prediction is clear (output=e.g. the growth rate of plants), influencing factors (input=e.g. factors in the growth rate of plants) are difficult to find. TCBM can lead us to some solution from measured data by taking into account the input factors and doing a modeling process. It also has a special evaluation index to determine a probable cause of unsuccessful modeling, such as problems due to the data or selected factors.

When using the developed model, collected data is sometimes not evenly distributed as it has gone through a number of channels, such as neutral networks or a multiple regression system. We can ensure a good degree of accuracy with plenty of data but not a small amount of data; therefore reliability is not always assured. The TCBM method of modeling was developed in order to overcome this problem.

In short, by using this method, we can attempt to predict the output value from corresponding input values by referring to a base of past cases. The fundamental idea is rather simple. If the input happens to coincide with the existing sample, then its output value is quoted immediately. If the input is different, the output might be expected as the average of the most similar samples. The reliability of the data is then calculated on the difference between the actual input and the similar one in the past.

---It figures out the level of uncertainty, doesn't it?

K---Exactly. We have been developing this idea for several years. At the moment, we use TCBM for predicting the load on air conditioners as well as the flow of sewage to a treatment plant, to prevent sewage overflow and urban disasters. We hope to apply TCBM to agriculture, which involves complicated factors. In agriculture, expressing its mechanism using numerical models is difficult. So we gather samples of what will occur due to various phenomena, such as the effects of time and likely period of sunshine, and water supply on the plant°«s growth. Accumulating data of these variables helps us to predict what happens next and estimate the probability of occurrences although the mechanism remains unknown. The system therefore succeeds in acquiring the knowledge and imitating the actions of an experienced farmer.

Case Study: Harvest Yield Prediction for Komatsuna (Japanese Leaf Vegetable)

---Is there a practical example of agricultural use of IT?

K---We have attempted to predict the harvest yield of komatsuna, a leaf vegetable, jointly with the Agricultural Research Institute of Kanagawa Prefecture. When the harvest of komatsuna fluctuates due to weather changes, that sometimes produces a bumper crop and the price of komatsuna falls significantly. So we investigated the possibility of predicting its season and harvest yield for the year. We also had an idea of controlling crop yield by controlling the environment of the greenhouse. Here is a relatively simple example. We input such variables as temperature, amount of sunshine, and any daily differences we found into TCBM, which went smoothly. We also input factors as environmental conditions in a greenhouse and weather forecasts made by the Japan Meteorological Agency. We also ran experimental simulations to investigate how temperature affects the growth of plants in a greenhouse.

---Is this system ready for practical use?

K---It°«s still a little difficult at this point. It is a very effective way to know the exact harvest season of komatsuna, but is yet to be proven profitable for the farmer who adopts this method. For instance, if the farmer grows other vegetables in the same greenhouse, the practicality of this system becomes questionable.

Another joint research is being carried out with Professor Takehiko Hoshi of Tokai University on the growth of cherry tomatoes. Although the harvest prediction has shown relatively good results, there is an important point to be mentioned. The fact is that harvest methods largely influence the harvest in the following year. In other words, the human factor seems to be involved to a huge extent. In order to make this system more practical, we will have to learn to manage this factor. In this case, the number of working hours is also input as numerical data.

Agriculture involves various factors, such as nature, environment in a greenhouse, and human work. For this system to be successful, it is very important to select the necessary factors to utilize as data. The tricky part is that just inputting data is not enough to produce an output.

Therefore, it will be essential to set up a system in which specialists, such as experienced farmers and academics researching the growth mechanism of plants, cooperate in estimating the growth mechanism and processing related data.

Deployment into Business

K---When we consider the practical use of the system, the most important point is its stability. Same as in industry, when a method leads to failure, we can°«t afford to lose everything. The same thing applies to developing an agriculture business. It is acceptable to have poor quality crops so long as we have a harvest, but it would be a problem if all the crops died in a bad year, or all crops grew well in a successful year.

We are still at the stage of developing future applications for this system. But as business is business, preparations must be made carefully. There is an obvious gap between the content of IT for agriculture and one that is practical in business.

---How is the present application for business going?

K---We are presently at the stage of modifying both measurement and control equipment for agriculture and gathering data. Another issue is the control of the environment in the greenhouse as explained earlier, but it is a kind of extension of air-conditioning in buildings. In fact, we are just trying to control of the environment of the greenhouse rather than the growth of crop.

-- What kind of sensors do you use for IT for agriculture?

K---There are sensors for temperature, humidity, CO2, and pH value of soil. In agriculture the demand for CO2 sensor is high as the growth of plant depends on the amount of CO2 for photosynthesis.

Future IT for Agriculture

K---In order to avoid misunderstanding I would like to emphasize that we are not specialists in agriculture. As we do not have expert knowledge of agriculture, the issue is how to contribute to agriculture through our measurement and control technologies.

In the era of information technology, computers have become so powerful and any kind of information can be collected through various networks. If we could utilize TCBM in analyzing and modeling collected data, so much more can be done. In particular, the mechanism of agriculture is so complex, and this system would be useful for predicting future events from data. This is the point that totally differs from conventional control technology. It works out estimated values from data and creates outputs, even though mechanism is not completely clear. To do so, just obtaining reliable data and utilizing sophisticated processing system are not enough. We need people to provide the data, or else the information processing ability would be wasted. Therefore, it is important to determine ways to collaborate with the specialists in the field, the use of TCBM, and utilization of a soft sensor. We hope to play a crucial role in these aspects in the future.

---Thank you very much.

NatureInterface Co.,Ltd. FAX:+81-3-5222-3581
HTML by i16 2019/11/15