Agricultural scientists who grow vegetables with AI are trying to start an industrial revolution

On Chongming Island in Shanghai, summer sunlight pours through the glass roof of a huge “plant factory”. Cucumbers, eggplants, melons, peppers, and other crops in the factory are exposed to the sunlight, and the interior of the factory becomes blazing hot as a result. At 9 a.m., the sweaty workers decide to call it a day. Dr. Ding Xiaotao, one of the plant’s researchers and a member of the Plant Factory team from the Shanghai Academy of Agricultural Sciences (SAAS), is standing outside looking for a moment of coolness.
This is the Chongming base for facility-based agriculture under the Shanghai Academy of Agricultural Sciences, where the futuristic “plant factory” is located – the latter uses a set of intelligent control systems to create an environment that meets all the nutritional elements (light, temperature, water, fertilizer, gas, etc.) required for plant growth. It makes agricultural cultivation completely free from the limitations of nature. Here, human beings can independently regulate the growth environment of plants to get the best growth conditions. Starting from each nutrient element, little by little, the yield and quality of plants will be pushed to new heights. This outlines the significance of the development of “plant factories”: standing on the scientific and technological imagination, it allows human beings to get rid of the restrictions of natural conditions, in Mars, Antarctica, oceans, and other extreme environments to realize the “freedom” of vegetables; standing on the reality of the people’s livelihood, it’s intensive, all-green planting methods, but also to help China cope with the arable land and other resource constraints so that the people can eat all year round to a super-high quality of vegetables.
However, the advantage of the “plant factory” has always been its fatal disadvantage – the cost of building such a complete set of intelligent planting systems is too high, making it difficult to commercialize the landing, and can only stop at aerospace, polar research, and other non-civilian application scenarios.
In the past two years, the civilization of a “plant factory” has been put on the agenda by a group of scientists. Dozens of agricultural researchers and scholars from top institutions across the country have decided to work together to find a solution for the industry. What they are thinking is: how to control the plant growth environment, with the lowest production cost, to obtain the highest growth yield (yield + quality), so as to break the status quo of “plant factories” can not make ends meet. Their past experience in agricultural farming and cutting-edge digital technologies, including AI, will help them tackle this challenge.

1. What is the state-of-the-art in agriculture?
In a “plant factory”, the key elements of crop growth can be artificially controlled. Light comes not only from nature but also from artificial LEDs on top; temperature is controlled by dozens of air conditioners and fans; water is brought in from artificial reservoirs and automatically irrigated to the plants along with nutrient solutions (fertilizers); and carbon dioxide rechargers provide the gases needed by the plants. The coupling of these nutritional elements together ultimately determines the growth of the plant. This is exactly what researchers like Ding Xiaotao are doing – allowing plants to achieve optimal growth through the precise regulation of each nutrient element.
“Whether it’s in the nutrient solution or the light formula (and other nutrients), I’m very happy to have a little research breakthrough that can improve the yield of plants and make them grow better,” he said. He said. Thanks to their efforts, the crops in the “plant factory” grow efficiently. At the site, Geek Park saw that the vines of cucumbers, eggplants, melons, and other crops were stacked one on top of the other, which means that they were planted more than ten times before pulling the rice seedlings (end of growth). In contrast, traditional planting methods can only plant one or two crops. In terms of intuitive yield figures, the annual output of plants in the Chongming plant factory is 10-20 times higher than that of traditional planting.
Not only the yield but also the quality of plants is better. Under the traditional planting method, once the light, temperature, and other growth conditions are not suitable, the plant will easily get sick. In the end, it can only be solved by pesticides. However, in the “plant factory”, due to the controlled conditions, the growth of plants is very healthy. Geek Park saw that the staff just put a kind of yellow paper around the crops to stick the bugs. “We didn’t do any medication, you can just pick it and eat it,” he said. After saying that, Ding Xiaotao picked a cucumber and ate it. Quality also means nutritional value, and experts can also customize vegetables with trace elements, such as potassium, by regulating the nutrient solution.
China is a country with an extreme lack of resources per capita, and facility-based agriculture – a modernized way of planting that regulates the environment through industrialized means to achieve efficient, high-quality production of plants – has become an important initiative to deal with the country’s shortage of agricultural resources. As the highest form of facility-based agriculture, plant factories are the most important part of agricultural development. At present, China is the world’s largest country in facility-based agriculture, with 88% of the world’s total area under facility-based cultivation. However, more than 95% of them are solar greenhouses and plastic greenhouses, and the development of “plant factories” is still insufficient. “Our goal is to promote the development of facility agriculture through the development of plant factories,” Ding Xiaotao said. Ding Xiaotao said.
In the last two years, from academia to industry, research on plant factories has quietly become popular. Last year, the Shanghai Academy of Agricultural Sciences (SAAS) set up this “plant factory” team. Members Ding Xiaotao and He Lizhong have worked in the Chongming base for nearly ten years, other team members are also proficient in the deployment of plant nutrient elements, they are the first explorers of “plant factories”. In March of this year, they received a new challenge. That is, in a more advanced plant factory with a completely controlled environment (no natural light, all artificial light), to grow a more difficult, very sensitive to the environment, a new vegetable variety “Cuitian lettuce” (the slightest change in nutrients will affect the shape of its direction). This means that the relationship between environmental regulation and plant growth will be completely linked. As a result, the team’s plant factory ‘growing techniques’ will be exposed.
There are three other teams participating in the competition. They are a team of students from China Agricultural University, who have been studying “plant factories” for many years, a team from Shanghai Jiaotong University, who are studying “plant factories” from an engineering perspective, and a team from a company that has mainly studied greenhouse tomato cultivation in the past. (For more information about the competition, please refer to Geek Park’s previous report: “Build a Plant Factory and Practice the Ultimate Form of Wandering Earth in Advance”).
“More is better in the field.” In the opinion of Zhao Chunjiang, chairman of the tournament jury, academician of the Chinese Academy of Engineering, and director of the National Agricultural Informatization Engineering and Technology Research Center, the collision of the top teams from various fields of industry, agriculture, and enterprise “will help us to simulate the most ideal growing environment for crops, to understand more thoroughly the smart agriculture represented by plant factories, as well as to better commercialize and commercialize it in China.” Perhaps the significance of this exchange is that through the collision of the top research teams of domestic “plant factories”, we will see how far human beings can push the growth of plants and how far the performance of plant factories can be utilized.
2. Large-scale commercialization of “plant factories” is approaching an inflection point.
At the Chongming base, there is another special “plant factory” – it is modeled after the Antarctic research station “plant factory” 1:1 construction. At the bottom of the house are 15 sturdy “support pillars” that are firmly rooted in the Antarctic tundra to protect the plant from strong winds and heavy snow. This small house helped the researchers grow more than 20 kinds of vegetables, such as lettuce, Shanghai green, leeks, etc., in the extreme cold, so that they could eat vegetables two or three times a week. This is exactly the most important application scenario of the “plant factory” in the early years – to serve national strategic and military purposes, to get rid of the natural conditions for human beings to grow vegetables in extreme environments.
In the last century, the birth of “plant factories” originated from NASA’s need to grow vegetables in space, and in 2014, Chongming Base built this “plant factory” for the Antarctic research station. In the past two years, the application of “plant factories” has begun to expand to more scenarios. During the epidemic period, some industrial parties began to envision that if ‘plant factories’ could be built around megacities, downstairs in neighborhoods, and in residents’ homes, then people would be supplied with a more stable, fresher, and higher-quality supply of vegetables. This is also in line with the ultimate consumer trend of plant factories. The development of “plant factories” has been accelerated. But the final hurdle in the journey between plant factories and people’s tables is cost.
As the highest level of agricultural technology, “plant factories” have an Achilles’ heel – that is, the cost of building a factory with a complete industrialized planting system is too high – take lettuce as an example, at present, every 1 kg of lettuce produced in a “plant factory” requires about 10 degrees of consumption. For every 1 kilogram of lettuce produced in a “plant factory”, about 10 kilowatts of electricity (about 6 yuan) are consumed. Together with the costs of seeds, nutrient solution, labor, etc., the price of lettuce in “plant factories” is about 10 times the price of ordinary lettuce. This results in high prices for the products, making it difficult for them to reach the tables of ordinary consumers. Taking the “plant factory” in Chongming as an example, Ding Xiaotao said that the company does not rely on selling vegetables to consumers to make money, but on selling “plant factory” equipment to customers to make money. This is the common pain point of “plant factories” at home and abroad.
Breaking through this industrial problem has been the direction of researchers like Ding Xiaotao in recent years. Still starting from the production elements (light, temperature, water, gas, fertilizer), agronomists can explore the optimal control of the production elements – in the case of the lowest cost of production, so that the plant grows the most and the best, so as to further reduce the marginal cost of production. In this way, agronomists are expected to level the input-output ratio model of ‘plant factories’.
This is the core purpose of the competition and the expectation of the agronomists themselves. “The commercialization of plant factories is currently a challenge for the industry, and as researchers in this industry, we hope to solve this challenge,” said Zheng Jianfeng, the team leader of the team from China Agricultural University, whose goal of the competition was to “explore a set of solutions for the industrialization and profitability of plant factories.” However, there are two difficulties in exploring such a set of industrialized solutions: First, how to precisely regulate the production elements to optimize the performance of the plant factory. Second, how to promote this solution to the whole industry without threshold.
It has been proved that digitalization + AI is a common solution to these two problems.
3. Digitization, AI + agronomists: speeding up the efficiency of plant factories
If we compare the operation of a plant factory to a system: one end inputs nutritional parameters such as light, temperature, water, gas, and fertilizer (cost), and the other end outputs the state of plant growth (yield and quality). Agronomists’ job is to understand the plant’s growth needs at every moment and to give it the most suitable nutritional parameters so that the plant can realize the best growth state. This enables the plant factory to operate efficiently.
And all this work of agronomists can be digitized – more accurately, artificial intelligence (AI) – to improve efficiency. Take understanding plant growth needs as an example. Plants don’t talk, and agronomists judge their status and needs by watching how they grow and look. Now, without having to go to the field, with the help of image sensors, agronomists can observe the growth of the plant anytime, anywhere. Without even having to observe, with the help of carbon dioxide sensors, agronomists can accurately calculate how much a plant has grown from how much it has “eaten”. Knowing the plant’s current growth status and needs, combined with the plant’s own growth demand curve, experts can quickly adjust to determine the nutritional parameters needed by the plant. At present, this parameter is first made by the AI algorithm, and then through the agronomist’s feedback and determination, and finally deposited in the AI algorithm. “Half by experts, half by AI,” said team member Wang Hong. In the Shanghai Academy of Agricultural Sciences, half of the “plant factory” team are traditional agronomists like Wang Hong, and the other half are agronomists who are proficient in artificial intelligence and modeling algorithms. Together, the agronomists and AI drive the growth of the plants.

Taking carbon dioxide as an example, the algorithm identifies the rapid growth nodes of lettuce by capturing its growth needs. At this point, the lettuce’s consumption of carbon dioxide becomes greater, “just like a small child, when he enters a rapid growth period, his need for nutrition is very strong. If you don’t replenish it properly, it will delay his growth.” Wang Hong said. In the end, the team replenished the optimal concentration of carbon dioxide at the right time. Take humidity as an example, lettuce in the seedling stage has low humidity, so the algorithm sets a humidity parameter of no less than 60 grams per cubic meter for it. As the transpiration of lettuce rises during the growing period, the algorithm controls the humidity parameter to 80 g/m3. Taking temperature as an example, temperature changes tend to make lettuce “heartburn” – i.e., the inner leaves are “burned”, which ultimately affects yield quality. Therefore, during the pre-growing process, the algorithm keeps the temperature constant at 22 degrees. One week before harvest, the algorithm adjusted the temperature accordingly.
In the end, the Shanghai Academy of Agricultural Sciences team grew the highest-yielding lettuce in the competition. They also won the championship. “That’s because we understood better what the lettuce needed and gave it the optimal growing conditions,” Wang Hong said. As for the control of planting costs, it can still be optimized with algorithms. To save on growing costs, the team led by Zheng Jianfeng designed an algorithm-driven automatic control system. In short, it is to “save as much as possible” on the input of each production factor. Taking temperature as an example, this system prevents the air conditioner from adding heat when the growing environment switches from a dark period (less light) to a bright period (more light). On the contrary, the system allows the air conditioner to “judge whether to lower the temperature”. In the case of nutrient solution, for example, the system allows the nutrient solution pump to run “at an appropriate frequency” and “intermittently”. In the early stage of planting, the system does not cool down the nutrient solution, but only in the “critical period” of planting. In the end, this student team from China Agricultural University took the ‘Best Energy Saving Award’ and was the runner-up.
“Through computerized processing, we simplified the original process of operating the farm, and ultimately increased both energy control and productivity (yield and quality) by 30-50%,” said Wang Jinhua, director of the Agriculture Department of the Bright Home Port and a judge of the competition. Bright Food Group, the main food supplier to the mega-city of Shanghai, and Pinduoduo, the country’s largest platform for uploading agricultural products, are the organizers of the competition. Their expectations for the tournament may have been met. Wang Jinhua said that the success of this competition is that it has tested a set of “more efficient” and “simpler” “plant factory” production models for the industry through the addition of intelligence and algorithms – a set of models that, after improvement, will have the opportunity to be truly promoted and landed in the industry. Wang Jian, Senior Vice President of Pinduoduo, believes that Pinduoduo has been actively participating in front-end agricultural science and technology innovation and that only by strengthening agricultural science and technology innovation can we promote the high-quality development of agriculture, improve the quality and added-value of agricultural products, and satisfy consumers’ demand for high-quality, safe and healthy agricultural products.
4. Special takeaway: thinking about commercialization
One of the unforgettable takeaways that these agronomists mentioned during their exchanges was ‘thinking commercially’. Yang Ruimei, a PhD student from China Agricultural University, said that in the past, when she did research in the lab, she only had to focus on her niche, such as light parameters. But today, their research has taken on the demands of industry and started to introduce an economic perspective. This is also the first time that Ding Xiaotao’s research team has written a commercialization proposal. On the 54-page PPT, they submitted to the organizers, they not only designed a packaging logo for the lettuce they grew but also wrote down their thoughts on the target audience, product advantages, marketing strategy, sales model, and other aspects – although these commercialization ideas only accounted for 1/20 of the final score.
“We not only need to know how to grow well but also need to think about how to sell well, which is a very meaningful experience,” Ding Xiaotao said. It was a very rewarding experience,” says Ding Xiaotao. Currently, they are preparing to apply for patents and software copyrights, and hope to form a corporate standard for this lettuce cultivation technology and promote it to the industry. Zheng Jianfeng, who already wanted to explore industrialized solutions for “plant factories,” took the first step. He found that if the planting model they explored this time is followed by the large-scale production of “plant factories”, the cost of lettuce will be reduced to RMB 2.8 per plant. If the target group is Shanghai’s high net worth families (he estimates 300,000), and the lettuce is sold at RMB 4 per plant, it could be profitable within five years. He hopes to start with some small-scale pilots to attract the attention of the target group and even industrial capital. At present, some industrial parties have begun to contact the participating teams.
This is exactly the significance of the agricultural research competition organized by Poundland, which allows the research results of scientific research to be transformed into social value more quickly. Zheng Jianfeng has participated in the competition twice, in the first agricultural research competition, he led the team to win the AI group. At that time, they explored the dynamic adjustment technology of nutrient solution formula, which has been written into a scientific paper and embedded into the algorithm model, and is being applied in Beijing Xiaotangshan base. The runner-up team of the same session has already set up the company “Wisdom Berry” to promote their intelligent planting system to the whole country. In Laowo Village, Nujiang Prefecture, Yunnan Province, the planting system has reduced the cost of common labor in the local strawberry industry by more than 30%, increased production by 30%, and promoted the development of the local strawberry industry.

This may be a win-win situation for all parties. For the researchers, they have increased their industrial perspective and can more clearly see the value of their research results. For the industry, they can unite the power of scientific research institutions to find solutions for their own industrialization problems. From the perspective of society, the innovation of platform companies such as Pinduoduo, which penetrates into the upper reaches of agriculture and penetrates into the cutting-edge scenarios of segmentation, can not only better drive the experience revolution of the consumer side but also help agricultural innovation to continue to break the circle.
This is a meaningful attempt in this era, worth doing for a long time.


Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top