Pinduoduo Strawberry Planting Competition: Current Scene Map Of Data And Fusion AI Agriculture

"In our ideal state, farmers will become similar to industrial workers or professional growers. They have a certain ability of data analysis, and their decision-making is not entirely dependent on subjective experience." This is min Qian Xixi's imagination of digital agriculture from Yunnan. She is also a PhD student in greenhouse horticulture at Wageningen University in the Netherlands.

On July 22, min Qian Xixi participated in the strawberry planting competition jointly sponsored by pinduoduo and China Agricultural University. The AICU team, which she and her classmates formed together, played the role of AI team in the competition. Four artificial planting teams, four AI teams, 120 days, waiting for nearly 1000 square meters of strawberry harvest.

This is the first time in China that artificial intelligence in the field of agriculture competes with top farmers, and it is also a signal that pinduoduo is going deep into the agricultural field.

Pinduoduo's eyes have never left agriculture. Different from the Standardized Agriculture in the United States, China's agriculture as a whole presents scattered small-scale farming. In the whole agricultural cycle, farmers are in the information blind area, and the judgment of their subjective will is facing a market full of uncertainty. In addition, the distribution and circulation structure between farmers and consumers is complex, farmers are often forced to harvest early, and consumers can not get the same value of products.

The data information of pinduoduo platform is centralized, which makes the mode of collage gather consumer demand and further transmit to the upstream farmers. The essence of this model is to enable manufacturers' data and give them more time to plan production and match consumer demand.

In 2019, pinduoduo's total orders reached 19.7 billion, and the number of active agricultural (by-product) businesses on the platform reached 586000, up 142% year-on-year. This strawberry planting competition is a further extension of the platform to the agricultural field layout.

"The purpose of the competition is not simply to win or lose, but to combine the new generation of artificial intelligence technology with agricultural production through the same platform of the old generation of farmers and the new generation of technology, so as to provide a low-cost and easy-to-use digital agricultural production management method for China's numerous agricultural product industrial belts and small farmers." Gong Ze, senior director of pinduoduo new agricultural and Rural Research Institute, told 21st century economic reporter. This is another achievement driven by artificial intelligence.

Cross disciplinary teams

It's not the first time Minqian Xixi has participated in a similar competition. Previously, Tencent and Wageningen University jointly held the "international artificial intelligence greenhouse planting competition". She participated in two sessions, one for Cucumber Planting and the other for tomato planting. AICU itself has practical experience. This time, AICU made it to the final of 17 AI teams.

For AICU, who grew up in the background of Dutch agriculture, the team has to face two difficulties from the formation to the competition. The first is the interdisciplinary challenge of team members.

Before the formal formation of the team, members from non-agricultural background had been very interested in modern agriculture, and even some members had put their ideas into family gardening, "building a small greenhouse with environmental sensors at home, such as building a small growth space with fresh-keeping boxes". These small practical experience, for the members to participate in large-scale competition to build confidence. Team members with agricultural background have already made achievements in the field of agricultural digitization, and they will not be unfamiliar with AI algorithm.

"The difficulty of interdisciplinary communication is that we must be able to learn and understand some basic knowledge and knowledge in each other's field, and then integrate different disciplines." Min Qian Xixi told reporters.

As far as plant science is concerned, data-driven has been applied at different levels in the Netherlands. Several members majoring in plant science also have some basic knowledge and experience in data analysis and programming. For AI algorithm, it is difficult to understand plants and their cultivation patterns. "There are many default knowledge points in the field of plant science and crop cultivation. Sometimes we will ignore them in communication. When we find this situation, we will communicate in time." At the same time, members also participated in the opening day held by the Dutch greenhouse industry every year to learn about the greenhouse cultivation mode.

The second challenge for teams is to compete across countries. Min Qian Xixi said, "the cultivation of crops is essentially the same, but the experience we are familiar with in Holland needs to be adjusted in China according to local conditions, including greenhouse equipment, sensors and planting ideas. At the same time, we need to accumulate a lot of data. "

The greenhouse horticulture in Holland has developed for nearly half a century, and the whole industrial chain is in a relatively mature state. From the daily cultivation management, to the application of sensors and control equipment, to the management of energy and labor, as well as the supply chain docking of supermarkets or markets, each link has developed a relatively complete mode. Minqian Xixi said that modern greenhouse horticulture and digital production in China almost go hand in hand. "We don't need to evolve completely according to the Dutch model, maybe we can achieve leapfrog development."

In addition to this, it can also form the advantages of digital production of greenhouse manufacturers.

The test of AI agriculture

AI agriculture has great potential, but its practice is also facing a test. According to min Qian Xixi's first two entries, the application of AI algorithm in greenhouse horticulture needs to be implemented in stages. In the two sessions of "international artificial intelligence greenhouse planting competition", the scores of Ai Group and artificial control group have changed. In the last year, there was only one artificial planting team. In the second competition, AI teams all exceeded the artificial growers in profit.

"It's hard to compare the two competitions directly because of the different crops planted. In the second session, the organizing committee made a detailed analysis on why AI exceeded the manual. The output and cost (including energy, water, labor, losses, etc.) of each group are calculated by subdivision. At the same time, there is also a dynamic comparison according to the time schedule. In terms of energy and resource sustainability, we have a great advantage, which also lays the foundation for profit winning. " Minqian Xixi said.

In the current scenario, the data collection of AI agriculture is still focused on the perception of the environment. "We don't know what the state of plants is now." Modern greenhouse needs experts and labors to collect samples regularly to measure plant indexes. On the one hand, it is harmful and the data is not timely. On the other hand, the workload is very large and the loss of human resources is high.

In addition, the lack of cultivation management data is also the bottleneck of AI agricultural development. Min Qian Xixi told reporters, "at present, what decisions you make on the state of plants, and what kind of reactions plants give after that are very important information. What we call human knowledge and experience is the part that needs to be learned by AI."

The purpose of AI intelligent agriculture is to assist the growers to copy the successful planting techniques and methods, and to control the environmental factors that affect plant growth, so as to achieve the unified mode of production, which is of great significance to improve the production efficiency. It's just like sharpening a knife and cutting firewood. The study of knives is far more meaningful than that of chopping.

At present, the development of AI agriculture is still in the primary stage, and learning human experience and knowledge by using data is still the core of its development, but the automatic layout of agricultural field will come eventually.

A preview of digital scene in AI agriculture

?