Science - March 29, 2018

Wanted: computer with green fingers

Albert Sikkema

A smart greenhouse that controls itself without any human intervention? It’s possible. In fact, it will exist in 10 years’ time, says horticultural researcher Silke Hemming. She is coordinating a challenge in which international teams will design such autonomous greenhouses.

illustration Pascal Tieman

At present, horticultural farmers determine what happens in their greenhouses. They might have a climate computer that records and adjusts the temperature, humidity, CO2 levels and incident light. They may even have a ‘fertigation’ computer that controls the supply of water and nutrients. But ultimately it is they who press the buttons. Based on their own knowledge and experience, they decide what to do in order to maximize production while using as little fertilizer, water, energy and pesticide as possible. Dutch growers are among the best in the world in that regard, says researcher Silke Hemming of the Greenhouse Horticulture business unit of Wageningen Plant Research in Bleiswijk.

Artificial intelligence
Even so, Hemming ‑ who leads the Greenhouse Technology research team at WUR ‑ expects to see autonomous greenhouses controlled by artificial intelligence in the near future. There is growing global demand for high-quality food grown with the minimum impact on the environment. Greenhouses can produce that food, and increasing numbers of businesses are therefore investing in greenhouse horticulture. ‘These newcomers have much less knowledge and experience than our growers. Of course you need to train these people, but perhaps you could also use IT and artificial intelligence to replace that knowledge and experience.’

Some of the technology is already available in the current high-tech greenhouses. They have sensors that measure the temperature, humidity, CO2 concentration and incident light, using that data to steer crop growth. Hemming: ‘These sensors generate data. The question is what do you do with it. That takes you from data to management advice. For example, it might be 25 degrees, which is good for the crop. So the grower or computer wouldn’t need to do anything. But the data from the other sensors also has to be assessed and combined to give a recommendation and action; that may result in a lower temperature giving the same yield but with lower energy consumption. The computer can also use data from the past — what worked then will work now too.’ This gives you the simplest control system for a horticultural greenhouse, says Hemming.

A more advanced system can also turn that data into a prediction or a recommendation for the future. ‘A crop growth model could help here, an algorithm that calculates various scenarios for the future and then gives the grower a recommendation, or directly takes over control of the greenhouse. Then you have an autonomous greenhouse. The ultimate situation is when you give the computer an instruction — “I want to produce one kilo of tomatoes in six weeks” — and it then adjusts the entire system to achieve that goal.’

We want to work towards having a robot or sensor check plants for pathogens
Silke Hemming - Wageningen Plant Research

Replacing the practiced eye
But we are not there yet, as only some of the information about the cultivation of fruit and vegetables is available digitally. The biggest omission currently is what the grower sees with their own eyes while walking around the greenhouse. They may notice the leaves of the young plants drooping slightly. Can a camera record that and turn it into a cultivation recommendation? Later, the grower sees how much fruit there is on the plants and whether it looks ripe. Can a camera register that with equal precision? And what about spotting pests such as whitefly? Growers regularly check the underside of leaves for pests. Is there a sensor that can take over this task? Hemming: ‘We want to work towards having a robot or sensor check plants for pathogens, followed by a little cart with pesticides driving up to the infected plants and spraying them where needed.’

This example makes clear that new sensors, robots and artificial intelligence need to be developed first to replace the grower’s practiced eye and green fingers. But the sensors also need to communicate with one another properly so that the digital grower initiates the right actions.

The experience with picking robots shows how difficult it is to develop an autonomous greenhouse. WUR has developed several such robots, for example for cucumbers and roses, but there are still not any commercially available picking robots. Hemming: ‘The growers expect the robot to pick all the flowers or vegetables but that is still really difficult.’ So they still prefer casual labour, mainly from Eastern Europe. Even so, you will need picking robots too in the ultimate autonomous greenhouse, says Hemming.

Which means there is work to do for the participants in Wageningen Plant Research’s Autonomous Greenhouse Challenge. The institute is making a modern greenhouse complex available in Bleiswijk to five international teams, each of which will get 96 square metres in which to design and operate an autonomous greenhouse. WUR will be actively inviting teams but others can also apply on their own initiative via the special website. In early June, an international jury will select the five best teams from among the applicants. Those teams will start building their computer with green fingers in early September.


If you are interested in the Autonomous Greenhouse Challenge, see


Tencent investing in autonomous greenhouses

Innovative food production is a new market for internet companies
David Wallerstein - Tencent ©Koen Verheijden

The Chinese internet company Tencent is the main financer of WUR’s Autonomous Greenhouse Challenge, with a contribution of 150,000 euros. What is more, Chief Exploration Officer David Wallerstein is donating 50,000 euros of his own money to the competition. The American joined Tencent in 2001. With a slogan of ‘how to improve the world with technology’, he is looking for new investment opportunities. He was in Wageningen this month.

Innovative food production is a new market for internet companies, says Wallerstein. Tencent previously invested in a company that uses radar images for precision agriculture and a company that measures whether plants are suffering from heat or water stress using sensors. Now the Chinese Google wants to see whether computers could be better at managing a greenhouse than people. ‘We are looking for new models for indoor farming because we think we need this to feed the growing world population sustainably.’ Wallerstein expects computers to generate new management options that increase yields or save on costs.