Comprehensive coverage

The robotic farmers of the future

Researchers from the Technion are developing an unusual solution to the problem of the lack of picking hands in agriculture: robots that use small drones, which create a computerized image of the trees, to carry out agricultural operations in fruit orchards

Smart agriculture. Image: depositphotos.com
Smart agriculture. Image: depositphotos.com

This period of early summer has always been harvest time. If in ancient times it was necessary for the whole community to mobilize in order to succeed in fully harvesting the grain at the right time, in 2022 there are sophisticated combines that travel in the fields and make it possible to process many acres in varied tasks throughout the year, quickly, efficiently and with the need for only a few human operators.

However, even nowadays, quite a few human hands and eyes are still needed during the picking of many fruits, which the crude combines, designed to handle field crops, do not know how to do. The shortage of Katif workers is reported almost everywhere - in the United States, Bאוסטרליה, in Great Britain, Vietnam and Brazil. When the produce remains on the trees instead of being picked, this results in the loss of quality food and causes enormous economic losses - a fact that will become more and more problematic with the constant increase in the world's population.

In a series of new studies recently presented at the "Water and Environmental Engineering in the Face of Climate Change" conference of the Environmental Engineering, Water and Agriculture Unit at the Technion's Faculty of Civil and Environmental Engineering, a mobile ground robot is being developed that can bring the work of harvesting and fruit farming into the 21st century. The robot, whose development was led by Associate Professor Amir Daghani from the Environmental, Water and Agricultural Engineering Unit in the Faculty of Civil and Environmental Engineering at the Technion, will know how to use a drone or several small drones to perform all the operations required in the plantations - with significantly greater precision and economy than is possible for farmers today.

The existing developments are not suitable for fruit picking

The gap between the number of harvest workers and the scope of work is expected to become more significant as the world's population continues to grow. By 2050, there are expected to be more than 9 billion people in the world, and it will be necessary to feed them all Increase the scope of food production by 60-35 percent (Unless the whole world switches to a plant-based diet). One would perhaps expect that with such a populated world there would be no shortage of working hands, but this is not the case. "For decades, people have been moving from the village to the city - and less want to engage in manual labor," explains Dagani. "You see it in construction and agriculture, and it happens everywhere - including in countries with a very large population, such as India and China. In India, for example, picking coconuts is a very important task - but fewer and fewer people know and want to work in it." According to him, the problem also exists in Israel. "As in many Western countries, in Israel too there are quite a few years when apples reach the ground because there is no one to pick them in time."

Degani believes that the solution to these problems lies in robots that know how to harvest. "Just as automation solved many of the problems that characterized field crops, with the help of machines such as combines, once we can adapt it to plantations, it will be able to help farmers become more efficient and reduce the uncertainty that currently lies in obtaining skilled personnel for specific periods of the year," he says.

It is important to note that picking automation should be different from that used to treat field crops - which includes robots that are too large and too expensive. "In field crops, massive picking and harvesting is carried out over the entire area, usually without distinction," says Dagani. "Picking fruit for food should be gentle and selective, the fruits should be picked one by one and placed carefully."

The researchers found that when using a low-flying drone's perspective, differences in the configuration of the tree canopies next to the robot could be used to identify its location. Photo: The Laboratory for Civil Environmental and Agricultural Robotics at the Faculty of Civil and Environmental Engineering, Technion
The researchers found that when using a low-flying drone's perspective, differences in the configuration of the tree canopies next to the robot could be used to identify its location. Photo: The Laboratory for Civil Environmental and Agricultural Robotics at the Faculty of Civil and Environmental Engineering, Technion

According to Dagani, even robotic arms that are currently used in factories, which have a large range of motion and an ability of precision that humans can only strive to imitate, are not suitable for the task. "It is true that the robotic arms know how to perform a pre-planned action - but their sensing and decision-making capabilities are limited, and not suitable for agriculture," he says. "Agriculture is a more difficult world, it is necessary to work in an environment with great uncertainty and changing lighting and working conditions, so the robot must have sensing capabilities and complex decision-making." According to him, to all this is added the price constraint - it is necessary that the robot not be too expensive, so that farmers can buy it.

Air assisted robot

The difficulties did not deter Degani and his team, who went to work. The first challenge they faced was the need for the robot to find its way around the plantation - a more complex task than it sounds. "In order for the robot to patrol and go through all the trees - for example, in order to identify which tree and where in the tree it detects pests or ripe apples - it is important that it knows exactly where it is," explains Dagani. The plantation environment is relatively homogeneous from a ground point of view, with most of the trees looking roughly the same, the ground surface is relatively disturbed and the GPS reception is not very reliable.

The drone in action. The photograph of the laboratory for civil, environmental and agricultural robotics at the Faculty of Civil and Environmental Engineering, at the Technion
The researchers found that when using a low-flying drone's perspective, differences in the configuration of the tree canopies next to the robot could be used to identify its location. Photo: The Laboratory for Civil Environmental and Agricultural Robotics at the Faculty of Civil and Environmental Engineering, Technion

This difficulty gave rise to the idea of ​​cooperation between a ground robot and a drone. The researchers found that when using a low-flying drone's point of view, it was possible to use the difference in the configuration of the tree canopies (tree tops that form a canopy-like canopy) next to the robot to identify its location. The first study on the subject Published in IEEE Magazine, which deals with robotics and automation.

Today, the researchers are working on additional ways in which the robot-farmer can use small drones to perform the operations required in the plantation. First, they demonstrated that a drone can circle a tree, and create a detailed XNUMXD image of each of the trees in the field. These are required in order to optimize the agricultural work and move to a modern model of "precision agriculture": "The intention is that instead of making decisions on issues such as fertilization, irrigation, thinning, pest control or harvesting at the level of the entire field, we will look at the agricultural plot with a higher resolution and make such decisions until The level of the solitary tree", explains Dagani. This will make it possible to increase the amount of produce, by providing the best conditions for each individual tree - and beyond that, save the use of resources such as water and fertilizer, and potentially dangerous pesticides.

"The solution I believe in is that of a mobile robot that is on the ground and knows how to feel the tree - and at some point actually perform the required mechanical operations," explains Degani. According to the model promoted in his research, the small and cheap drones that the ground robot will control will be able to assist it in operations such as navigation, sensing or actual contact with the tree - such as pollinating the flowers, another project that is currently in the development process in the laboratory.

Apples only for the rich?

Today, Degani's research is in the prototype stage, demonstrating possibilities for future development. Be that as it may, in the fruit industry there are already several automation attempts - such as The development of FFRobotics The Israeli one, of a traveling robot equipped with arms that perform velvet, or Tevel-Tech's robot, which uses harvesting drones and should enter a pilot phase in the coming year.

Beyond that, it turns out that not only the identity of the pickers is expected to change in the future, from human to robotic - but also the structure of the plantation itself. "The way we engineer and grow trees will change, and they will be designed in a way that will be suitable for robotic picking," explains Degani. "Already today you can see apple orchards in the world that look almost like a two-dimensional wall on which fruits grow. This is not genetic engineering but mechanical engineering operations designed to make the plantation as efficient as possible." The new orchard structure allows denser planting with a larger crop, and is designed to allow easier picking - for people and robots. are currently in progress Studies which are intended to determine what is the most efficient configuration, in preparation for the era when robots will enter the fields.

In the end, according to Degani, everything is directed towards efficiency - simply because we have no other choice. "Even in modern agriculture, the farmer will be very important, but he will need much less working hands," he says. "Like quite a few things, the data will be at the center, to help him make informed decisions - and the robots will perform the tasks in the field. This is what will adjust the efficiency, so that we can reach the goal of a sufficient harvest, which will make it possible to feed all the people who are increasing in number", he says.

"Since over time there is less land, less resources and less manpower, there is a necessity to find a solution - otherwise, fruits like apples will only be accessible to the very rich," Daghani concludes.