Precision agriculture requires a set of technologies — sensors, high-speed connectivity, artificial intelligence, and automation — to work together. It is driving the autonomous farming revolution forward with the objective of maximizing crop yields and profitability while minimizing the environmental impact.
In a keynote session at the recent Green Engineering Summit, Yulin Wang, technology analyst at IDTechEx, explained how precision agriculture makes farming more sustainable and discussed opportunities and challenges in agricultural robotics.
Sensing the environment
Not so long ago, farmers had to rely on satellite and aerial imagery or other mapping systems to track conditions in their growing areas. In agriculture, however, it is critical to make timely informed decisions. If farmers miss the perfect planting or nurturing window in their geographic area, the result is a lower crop yield.
Wireless sensor networks (WSNs) are now used to collect a large variety of environmental information, including light intensity and spectrum of incident light. Sensors can also detect climate variabilities (e.g., air temperature, humidity levels, carbon dioxide levels, and air speed), plant phenotyping (e.g., leaf color, plant mass, and plant size), and nutrient supply (e.g., pH level and nutrient concentration).
Once sufficient data is collected and transferred, “we utilize software systems such as machine learning, image processing, and AI techniques to conduct analysis of the input data to determine the optimal temperature, light intensity, nutrient supply, and whether the crops need weed management, along with many others,” said Wang.
Gaining value from AI
There is a tremendous amount of variability and unpredictability in farming. Weather patterns are a moving target, and there might be three or four soil types within a single field. AI is beginning to deliver on its promise to provide real value, driven by recent advances in pattern-recognition algorithms and higher computational resources.
“AI algorithms learn to make decisions by being trained on data, which can come from any number of sources,” Wang noted. “This data should carry sufficient information about the expected output. Better data leads to better outputs, which means data needs to be accurate, available at required frequencies and levels of granularity, and diverse. AI algorithms’ performance is always as good as the quality of the data used to train them.”
AI can be used in agriculture to improve crop management and productivity by rapidly classifying species, identifying plant diseases, efficiently detecting weeds, predicting yields, and monitoring crop quality.
Weed killing
One of the main applications of AI in agricultural robotics is smart weeding, in which the robot can distinguish between weeds and crops and selectively destroy the weeds. This enables intra-row weeding or spot spraying, which cannot be accomplished with conventional agricultural technology.
AI for weed recognition typically uses deep-learning algorithms trained on large datasets alongside RGB cameras used for imaging. The large datasets required can make data sharing from farms necessary to improve the algorithms, Wang said.
Livestock management
A drone can scan a field in just a few minutes, but it’s a different story with livestock. Animals are constantly on the move, which makes monitoring them much more challenging.
The role of AI in livestock management includes animal welfare and production such as milk yield. As Wang explained, sensors collect useful information such as electrical conductivity, fat measurement, facial temperature, sweat signals, eyeball movement, and ear position. This sensory data is then processed to guide dietary and medication intake.
Automating farming tasks
According to IDTechEx, the agricultural-robot market is set to triple by 2025, with autonomous tractors experiencing the fastest growth and weeding and seeding robots experiencing the second-largest increase. Harvesting robots, in contrast, are seeing slow adoption.
Recent advances in AI, computer vision, and positioning technologies have unlocked the value of agricultural robots and accelerated their development and adoption. Major equipment providers such as John Deere, AGCO, and Kubota have launched autonomous tractors, while startups have developed task-oriented robots. For example, French startup Naïo Technologies has commercialized Ted, a vineyard robot that weeds vine plots without assistance, while U.S.-based TerraClear aims to improve farmers’ productivity through end-to-end rock clearance automation.
Among all agricultural robotic applications, robotic milking is “by far the most popular” and is used by a significant percentage of dairy farms in Europe, said Wang. Agricultural drones are also beginning to find widespread application for imaging and spraying, although regulations continue to limit their use in some parts of the world and task autonomy remains somewhat limited.
Fully autonomous farms are just over the horizon. The U.S. and Europe are leading the way in robot development, but many challenges remain.
Technical challenges
Most agricultural robotic applications have reached a technology readiness level of 8 or 9 and are being commercialized. The only application that has not reached this readiness level is robotic harvesting of fresh fruits and vegetables.
Wang outlined three technical constraints. First, fruits and vegetables are delicate, and the force of picking or harvesting must be precisely controlled. Second, it is extremely difficult for machine vision to identify whether the fruits are sufficiently ripe. Third, tasks such as fruit identification, path planning, integrated conveyor belt operation, and force control of the grippers must work together harmoniously, compounding the challenges for the control systems.
3 ways autonomous farming is driving a new era of agriculture
The second is efficiency. Some harvesting robots simply cannot operate as quickly and efficiently as human operators.
Connectivity inequality
The adoption of autonomous farm equipment is largely dependent on the broad availability of secure and reliable wireless connectivity. For short-range field farming, vertical farming, and digital greenhouses, typical connectivity options include Wi-Fi, Bluetooth, Zigbee, and Z-Wave. For long-range field-farming, cellular options such as 3G, 4G, and 5G are commonly used. Wang noted, however, that connectivity “has been a pain point for agricultural digitization because field farming often takes place in rural areas, where the telecommunication infrastructure is limited.”
Data ownership
Farmers need to have their data at their fingertips so that they can coordinate their plan of attack based on the history of their farm and their ability to interpret it. Today, data flows to the cloud-based system, and farmers can analyze it on the go via a mobile app on their phone.
One of the largest uncertainties surrounding agricultural robotics, however, is the ownership of the data and whether it belongs to the farmers, the data collectors, the technology providers, or the landowners, said Wang.
“In the absence of specific regulations on who owns agricultural data, ownership is generally defined through a series of contracts. This can create a complex situation for farmers, particularly when different robotics platforms are used on the same farm.”
The raw data cannot be owned, but once it has been processed or organized into a database, it may be subject to copyright protection. This can give exclusive rights to the owners of the data and prevent farmers from taking advantage of the data they have generated or force them to work with specific vendors. The concern is that this possibility will discourage the adoption of agricultural robotics, as well as other approaches to precision agriculture, Wang said.
Cost barrier
Projected profit margins in agriculture are getting tighter and tighter, and investments are tied to a rationalized strategy. But robots don’t have to cost an arm and a leg.
While the costs of transitioning to automation have long been out of reach for many small businesses, technological advances have lowered the barrier to entry. The typical price for a small field robot is now between €12,000 and €36,000, said Wang, specifying that robots lower labor costs and solve the growing labor shortages in agriculture.
Robot-as-a-service (RaaS), meanwhile, is gaining popularity because it lets farmers access the benefits of robotics without a significant upfront capital investment. They pay only for the service and can rely on a trained operator to handle technical issues and minimize downtime. This can be of particular benefit in the case of expensive robotic equipment with multiple sensors and complex algorithms.
