5G in Horticulture

According to McKinsey, the invention and adoption of precision agriculture is shaped by big data and advanced analytic capabilities, AI and robotics. Aerial imagery, sensors, and sophisticated local weather forecasts enable optimised and predictive farming: for instance, the technologies can identify that a section of an orchard is suffering from poor soil health or that the temperature in one part of a polytunnel is at a level likely to cause produce damage. Armed with these insights, growers can deliver tailored treatments and interventions, such as applying fertiliser to only one part of a field rather than blanketing the whole field.

The Vodafone Foundation, in conjunction with Nokia, has launched Smart Agriculture-as-a-Service to improve the livelihood of 50,000 farmers across 10 districts in the states of Madhya Pradesh and Maharashtra in India. More than 400 sensors have been deployed over 100,000 hectares of farmland to collect data for analysis by the solution’s cloud-based and localised smart agriculture app. Sensors include soil probes and weather stations. Insights from the data will help farmers improve soy and cotton crop yields, as well as reduce their impact on the environment.

P. Balaji, Chief Regulatory & Corporate Affairs Officer at Vodafone Idea Limited, said: “Smart crop management using Smart IoT and AI-based solutions is transforming the prevalent agricultural practices into more ‘intelligent’ ones, enabling farmers with smart decision making and helping them improve production and crop quality through better utilisation of resources.”

Mäkelän Mansikka Oy, a mid-sized strawberry farm in Eastern Finland, has adopted technology to operate a smart farm that profitably grows strawberries - typically a highly capital-intensive business - efficiently and sustainably. The farm uses drip tape irrigation and automatic, real-time soil monitoring for the strawberry plants. Irrigation water and nutrients are fed directly to the plant's base, below the plastic cover. Irrigation and fertilisation are controlled automatically so that the plants are provided with the right amounts of nutrients according to their growth stage, weather conditions, and soil moisture.

Their soil sensors measure moisture, temperature and salinity, which is then analysed to help them plan irrigation and fertiliser timing and dimensioning. The data has also helped them to determine the optimal timing for planting and frost protection; sensors at the ground surface notify growers automatically when the temperature drops below two degrees centigrade. This reduces losses and labour costs because workers no longer need to manually collect temperatures.

This data-driven decision-making has empowered the team to make more efficient and effective interventions, reducing fertiliser and labour costs while extending their growing season.

A similar success story can be found in Algeria, where Nokia has helped a peach farmer increase his yields, reduce impact on the environment, and cut irrigation costs. Nokia created a first-of-its-kind Worldwide IoT Network Grid (WING) to support various IoT applications that mobile network operators can then access. For this trial, Nokia worked with Algeria's largest mobile network operator Djezzy. The farmer did not need to know about IoT; Nokia and Djezzy provided the technology and connectivity. Data was then collected around soil temperature, humidity, volumetric water content, water evaporation and salinity. These readings were analysed, which then allowed the farmer to accurately manage irrigation cycles and soil nutrition deployment.

After just one month, the farmer was able to reduce water consumption by 40% on a single irrigation line for one hectare, which increased his revenues by up to 5% per hectare. This delivered a return on investment in fewer than two years per irrigation line. 

It’s not only sensors that can provide insight. Agtelligence is a start-up combining LiDAR - a remote sensing technology that uses laser light, with hyperspectral data from a wide range of wavelengths - to accurately measure the distance to objects on the ground and create detailed images of the earth’s surface. This imagery enables Agtelligence to monitor and measure environmental metrics that capture soil health and biodiversity, as well as the anthropomorphic impact on the land. 

Providing insight across the agrifood value chain would not be possible without the recent development in space technologies. This is also true for other agri-tech applications. Cameras, for example, are now able to identify crop stress early, preventing the impact on yields by measuring and monitoring grassland to optimise grazing and soil carbon storage.

With advanced connectivity, computer vision can automatically identify weeds while robots deal with the leftover plant material. Ecorobotix has developed an automated precision weed spraying solution for various produce, enabling growers to reduce the use of protection products such as herbicides, fungicides, insecticides and fertilisers by up to 95%. In addition, surface spraying can be localised to a precision of 6x6 cm, ensuring only the target plant is sprayed.

AeroFarms is a US vertical farming company that operates fully connected smart vertical farms that produce crops all year round. The company uses drones and AI algorithms to monitor crops in real-time. Drones inspect vertical farms and capture high-res photographs of each plant; the data.is then sent for cloud processing, before computer vision and machine learning algorithms analyse each plant leaf, stem length, curvature, spotting and tearing. Detecting poor growth and diagnosing the causes. Alongside increasing yields, this saves labour and input costs.

“When I watch the drones autonomously imaging our plants, I’m blown away by how this truly represents the power to grow the best plants possible” - David Rosenberg, CEO, AeroFarms

EarthSense has produced a ground robot, the TerraSentia, which uses a mix of sensors—including visual cameras, light detecting and ranging (LIDAR) tools and GPS devices—to collect data on the health, physiology and stress responses of a variety of produces. Collecting under-canopy data and insights into trunk displacement, maximum daily shrinkage, daily gain, and tree water deficit​​, it can scan 10 plants a second. Its cloud-based platform also enables crop scientists to teach the robot to automatically measure a range of key traits like height, condition and leaf-area index. 

Gardin similarly uses technology to provide an early indication of plant stress by measuring the photosynthetic performance of a plant growing in either a vertical farm or greenhouse. The system uses a robotic sensor to monitor the crop in real-time: for instance, automating watering strategies for leafy crops informed by photosynthetic patterns, with these crops outperforming those grown with static watering patterns.

Gardin has also been exploring the impact of lighting through an 18-month research programme funded by Innovate UK, Project SysSen. They tested hundreds of light conditions and found interesting relationships between the plant and light quality – specifically the ratio between certain wavelengths. Light was found to not only influence yield but also nutritional content, flavour and appearance.

Image

Clover is a hydroponic farm based in Bangalore, India that uses wireless sensors to monitor plant growth in their greenhouses. Aggregating data from different locations has allowed them to improve growth protocols, yielding healthier plants and a plentiful harvest.

Insects and rodents can cause untold damage to produce. Insects, in particular, are often hard to detect with the human eye until they’ve created significant damage. Blanket application pesticides and insecticides can play a major role in preventing infestations but pose a variety of unintended consequences including intoxicating plants with harmful chemicals. Additionally, continuous exposure allows insects and bugs to develop resistance, forcing growers to rely on heavier pesticides. Alternative approaches are required to protect the environment, reduce grower costs and align with increasing consumer demand for more sustainable growing practices.

Low-cost image-capturing sensors are an option but can only identify insects that are visible to the naked eye. The Bayer Innovation Lab has developed the MagicTrap -- a smart insect trap equipped with a camera and minicomputer that photographs the inside of the tray at regular intervals and automatically evaluates its contents. The results are visible to farmers via the MagicScout app, saving countless hours of unnecessary trips and inspections.

More sophisticated methods can, however, offer earlier identification. Fluorescence image sensing captures changes in a leaf’s chlorophyll while acoustic or even gas sensors can identify specific chemical compounds which plants produce when stressed. This allows growers to apply insecticides only where and when needed. 

But it’s not just during the growth phase where produce is at risk from infestation. ContraMoth is a real-time sensor that detects various insects at all stages of their development. The technology senses the pheromones of the adult insect and the semiochemicals of the larvae, detecting at a level not previously achieved.  ​​

Harvesting accounts for 20% of all agricultural work. Autonomous robot crop pickers can therefore reduce the pressure on labour shortages, making abandoned harvest-ready fields a thing of the past. What’s more, with robots conducting time-critical work such as harvesting, operations can continue 24/7, regardless of the weather conditions.

Fieldwork Robotics is trialling a robot that can pick more than 25,000 raspberries a day, outperforming human workers who can pick roughly 15,000 in an eight-hour shift. Using machine learning, with guidance from sensors and 3D cameras, Fieldwork Robotics is trialling a robot that can pick more than 25,000 raspberries a day, outperforming human workers who can pick roughly 15,000 in an eight-hour shift.

Various robotic solutions are also available for apple orchards. California-based robotics company Advanced Farm is one of a handful of robotics companies conducting field tests in Central Washington. The company’s flagship robotic apple harvester is powered by a main computer that independently controls the motion of six surprisingly nimble arms.

Each arm is equipped with a suction cup on the end, which eliminates bruising during the picking process. Built-in cameras locate each apple and assess whether it’s ripe enough to pick. Once the apples in a given area have been harvested, the robot moves forward and repeats the process.

“This robot can work 24 hours a day. That can help us get the job done.” Kent Karstetter, Orchardist

However, not all fruit-picking robots are land-based. Israeli robotics company, Tevel Aerobotics, has developed a robot that consists of eight autonomous flying drones. Mounted to each is a small rod with an attached suction cup. Cables tether the drones to a long, flat conveyor and the harvester is outfitted with sensing technology that determines the location and ripeness of each apple. When the fruit is ready for picking, it uses the suction cup to pluck apples off the tree.

In the UK, the University of Plymouth is working with Agri-Tech Cornwall to develop robotic systems for automating manual picking operations for cauliflower, broccoli, kale and cabbage, all of which are extensively grown in Cornwall. A two-handed robot test rig is currently being designed, built, and tested under field conditions.

The project uses soft robot arm technology, in which the robot arm joints can vary their stiffness in real-time, softening to withstand an impact during fast ballistic phases of movement, and then stiffening to ensure accuracy during the approach and picking phase. 

Companies such as Small Robot Company, Muddy Machines and Antobot are developing commercial services to assist growers. Muddy Machines has focused on produce with high labour-intensive harvesting methods, such as asparagus, while the Small Robot Company - which successfully demonstrated their capabilities in the 5G RuralDorset project - claims their treatment maps and integrated sprayers can reduce herbicide usage by 90% and fertiliser by 24%.

Panasonic has developed a tomato-picking robot. Suspended from a rail, the robot is equipped with a camera that has an image recognition function that helps it to identify tomatoes and whether they are ready to be harvested or not. While the robot collects tomatoes slightly slower than the average human picker, it can work for 10 consecutive hours or more, meaning it is more efficient overall.

Image

"The ability of robots to lighten the harvesting workload is a major advantage. The time spent on harvesting is at least 20% of the entire agricultural workload. About 160,000 man-hours are spent each year working inside our greenhouses, and 35,000-36,000 hours of this is spent in harvesting. We can automate this by using harvesting robots," said Masataka Nakamura, manager of a farm in Japan, who uses the harvesting robots.

Robots are also being deployed for pruning and weeding: for example, consider volunteer plants, which typically are removed through spraying pesticides or human labour, both of which can be costly. With advanced connectivity, computer vision can identify weeds automatically while robots deal with the leftover plant material. Ecorobotix has developed an automated precision weed spraying solution for various produce, enabling growers to reduce the use of protection products such as herbicides, fungicides, insecticides and fertilisers by up to 95%. In addition, surface spray can be localised to a precision of 6x6 cm, ensuring only the target plant is sprayed.

Image

Agrointelli is a Danish company that develops autonomous robots for farms. Among other things, they are currently being used by growers to fight volunteer potatoes that spoil sugar beet crops. Potato leftovers grow more quickly and block the sunlight from reaching the sugar beets, preventing them from absorbing sufficient nutrients and growing normally.

The 5G-connected robot is equipped with cameras and precision sprayers. It takes photos of plants and sends them to a cloud-based server. The machine learning algorithm compares these photos with over 6,000 images of weeds and potato plants. After classifying each image, the server sends them back to the robot. If the plant is a potato, the robot sprays it with glyphosate. This full cycle takes approximately 250 milliseconds. Manually spraying volunteer potato tubers takes, on average, 20 hours per hectare and costs around €400. It takes a robot around three hours to process one hectare, with up to 95% of volunteer weeds identified. 

Other automated options for weeding are also being developed. Naio Technologies has specifically created a robot, named Ted, that can be used on vineyards. Ted provides a precise mechanised alternative to herbicides. This reduces the use of chemicals, something which is particularly valuable amid a climate of consumers increasingly concerned about their usage.

Image

WeedBot has developed a tool that can distinguish between produce such as carrots and weeds, before treating the weeds with high-power laser light. This method is fast, requires no chemicals and offers unparalleled levels of precision, spraying weeds just millimetres away from crops. In a similar vein, Saga Robotics has developed a robot equipped with ultraviolet lights that can kill mildew on plants, reducing fungicide use by up to 90%.

More precise GPS controls, paired with computer vision and sensors and advanced connectivity solutions such as 5G, can further advance the deployment of smart and autonomous farm machinery. Soon farmers will be able to operate a variety of equipment on their field simultaneously without human intervention, freeing up time and resources. 

Autonomous machines are also more efficient and precise at working a field than human-operated ones, which could generate fuel savings and higher yields. 

Drones can also play a key role in automating key processes in the growing process, whether using machine vision and AI to check individual plants or monitoring indoor and outdoor produce to identify the optimal time to harvest.

Dendra, based in the UK, is using drones to plant trees. Focusing on ecosystem restoration, their solution is capable of planting 120 seedpods per minute with the potential to also be used for planting produce in horticulture.

AeroFarms is a vertical farming company in the US. They operate fully connected smart farms that produce crops all year round. Working with Nokia Bella Lab, they use drones and AI algorithms to monitor crops in real time. Due to the vertical nature of the farm, this is extremely hard to do manually. Drones inspect and capture detailed photographs of each plant for crop inspections that go far beyond the human eye. The data is then sent for cloud processing using a 5G connection and, with AI processing capabilities, allows appropriate action to be taken, from watering to harvesting. The drones are managed by an orchestration system that enables them to work in unison, fully autonomously, with each drone assigned specific monitoring tasks to avoid overlap. AeroFarms credits this technology with helping them achieve up to 390 times greater productivity per square foot than traditional farms while using up to 95% less water and zero pesticides.

David Rosenberg, CEO at AeroFarms, said: “With Nokia Bell Labs, we have developed the next-generation system that can image every plant every day in a cost-effective way at scale. This level of detailed imaging and insights helps us be better farmers by monitoring our plant biology dynamically and allowing us to course correct as needed to ensure the highest level of quality all year round. When I watch the drones autonomously imaging our plants, I am blown away by how this truly represents the power of harnessing leading-edge technologies and bringing brilliant problem solvers together from diverse groups to grow the best plants possible.”

Drones can also play an important role in helping to protect produce in the face of pest infestation. DJI Agriculture was brought to protect farms in China’s Yunnan Province when news arrived of an incoming swarm of yellow-spined bamboo locusts that had already decimated 6,667 hectares of fields. Drones were used for aerial application of anti-locust pesticides and successfully covered 2,000 hectares of land, protecting produce far quicker than traditional spraying methods.

AeroFarms is an indoor vertical farm in New Jersey that uses drones and AI to optimise the management of its crops, rocket, bok choy, kale, watercress and microgreens. Equipped with AI-trained computer vision, the drones move through the indoor environment and scan plants, capturing data down to the pixel level. This allows the growers to detect areas of poor growth, but also diagnose the causes, for instance, malfunctioning irrigation systems or suboptimal lighting). The challenge is that drones can capture tens of thousands of images each day, far more than humans could process in the appropriate window needed for correction action to be taken.  

Using AI, Nokia Bell Labs has developed machine vision technology that captures the botanical data of individual plants. Size segmentation, quantification and pixel-based scanning identify the consistency and variation in plant growth and yield prediction. A cloud-based processing platform has the power to run the required advanced algorithms; this enables AeroFarms plant scientists to use the new machine learning training tools that Nokia Bell Labs has created to enhance and adapt the system over time.

Image

“So much information can come from an image,”  said David Rosenberg, AeroFarms CEO. “Size in terms of width, length, stem length, petiole length ratio of those, colour, discolouration, spotting, tearing, curvature – and those are examples of what the eye can process. Certain images can reveal much more. All of this information not only helps us understand plant biology at new levels but also helps us be better farmers by course correcting how we commercially grow plants dynamically as needed.”

Closer to home, UK-based agritech company Xihelm has created an AI-powered robot designed to harvest fruits and vegetables using AI to identify only the ripest fruit.

Deep Planet uses AI to help vineyard growers better manage yield, maturity and irrigation. Its technology helps to predict grape maturity and optimal harvesting dates, forecast yields for picking and vintage planning. This optimises irrigation scheduling, and reduces variability across the vineyard, identifying problem areas.

Using AI and satellite imagery, Deep Planet’s VineSignal can reduce variability and disease by tracking daily, weekly and monthly changes in vegetation (NDVI). The unique two-week prediction determines the impact of irrigation and fertiliser application, delivered at an accuracy of 93%, and can forecast grape tonnage by block and variety with 90% accuracy.

Sushma Shankar, Co-Founder and COO of Deep Planet says: “Our goal is to help agricultural businesses, supply chain companies and growers to gain new and unbiased insights from satellite data while supporting sustainability in their operations.”

Following the UK’s departure from the EU, Defra set out its agricultural transition plan The Sustainable Farming Incentive supports sustainable approaches to farm husbandry. This includes actions to improve soil health and water quality, enhance hedgerows and promote integrated pest management. Similar agendas and policies are being ruled out across other nations and regions: for instance, the European Green Deal calls for a 50% reduction in pesticide use compared to 2020 levels. 

But not only policymakers are pushing for change: consumers are becoming increasingly aware of environmental impacts, choosing to buy products that demonstrate sustainable practices. Smart monitoring through sensors, connected cameras and automated equipment can digitally record rich data (including growing time and knowing what has been applied to produce), enabling growers to produce more resilient crops and provide field-to-fork traceability to meet demands for reduced chemical application, more efficient irrigation and optimised harvest conditions. This also makes it easier to achieve organic certification.

Growers are also grappling with a scarcity of resources: the World Bank reports that farming accounts for 70% of the world’s water withdrawal via irrigation. Conventional growing methods have been associated with soil degradation, water pollution and loss of biodiversity but IoT and automation can enable precision farming, allowing for more efficient use of water and dramatically reducing the use of pesticides and fertilisers. CropX demonstrated the potential to reduce water usage via a controlled trial at a Californian citrus orchard. Fifty trees within the orchard were chosen to test how the grower’s existing practices compared to the CropX-connected irrigation advice system. Control trees were irrigated using a sprinkler system with Hunter irrigation controllers; a separate, automated irrigation system was installed for trees irrigated using the CropX irrigation advice. The results speak for themselves: CropX irrigated trees experienced a 14% yield increase and a 57% water saving. 

Banana growers in Israel use CropX systems to manage irrigated and rain-fed banana plantations. In addition to using the CropX system to manage irrigation, users track nitrogen movement in the soil for more precise fertiliser applications, avoiding runoff and leaching. Growers in Israel have reported a 27% yield increase and 15% savings on fertilisers. 

In Algeria, Nokia has helped a peach farmer increase his yields, reduce impact on the environment, and cut irrigation costs. Nokia created a first-of-its-kind Worldwide IoT Network Grid (WING) to support various IoT applications that mobile network operators can access. For this trial, Nokia worked with Algeria's largest mobile network operator Djezzy. The farmer did not need to know about IoT; Nokia and Djezzy provided the technology and connectivity. Data was then collected around soil temperature, humidity, volumetric water content, water evaporation and salinity. These readings were analysed, which then allowed the farmer to accurately manage irrigation cycles and soil nutrition deployment.

After just one month, the farmer was able to reduce water consumption by 40% on a single irrigation line for one hectare, which increased his revenues by up to 5% per hectare. This delivered a return on investment in fewer than two years per irrigation line. 

Closer to home, companies such as Small Robot Company, Muddy Machines and Antobot are developing commercial services and working with farmers in the field. The Small Robot Company - which successfully demonstrated their capabilities in the 5G RuralDorset project - claims their treatment maps and integrated sprayers can reduce herbicide usage by 90% and fertiliser by 24%.

The Ted robot produced by Naio Technologies provides a precise mechanised weeding alternative to using herbicides. Automation can play a key role in reducing wastage of produce: harvesting robots can work 24/7, ensuring produce is picked at the optimal time and reducing the chances of produce being left unpicked. With current labour shortages, this provides a tangible solution to a very real problem. Alongside AI, the optimal time to harvest and expected yields can be identified, enabling more effective planning and reduced wastage of produce.

Fieldwork Robotics is trialling a robot that can pick more than 25,000 raspberries a day, outperforming human workers who can pick roughly 15,000 in an eight-hour shift. With guidance from sensors and 3D cameras, using machine learning, its gripper identifies and picks ripe fruit.

The University of Plymouth is working with Agri-Tech Cornwall to develop robotic systems for automating manual picking operations for cauliflower, broccoli, kale and cabbage, all of which are extensively grown in Cornwall. A two-handed robot test rig is being designed, built, and tested under field conditions.

The project uses soft robot arm technology, in which the robot arm joints can vary their stiffness in real-time, softening to withstand an impact during fast ballistic phases of movement, and then stiffening to ensure accuracy during the approach and picking phase. 

Panasonic has also developed a tomato-picking robot. Suspended from a rail, the robot is equipped with a camera that has an image recognition function that helps it to identify tomatoes and whether they are ready to be harvested or not. While the robot collects tomatoes slightly slower than the average human picker, it can work for 10 consecutive hours or more, meaning it is more efficient overall.

It’s not just monitoring and managing produce where there are opportunities to operate more sustainably. Monitoring conditions and usage of buildings and equipment also has the potential to reduce energy consumption.

As a part of the Worcestershire 5G project, manufacturer Mazak successfully deployed automated remote predictive maintenance, taking advantage of 5G’s ability to process large amounts of data. The factory is now able to provide real-time analysis of machine status and feed this information to a cloud system. The company’s spindles are usually only removed for corrective maintenance after an issue or failure occurs but with the arrival of 5G, early warning signs of damage are available; this reduces repair costs, as well as downtime.

The Port of Felixstowe used 5G IoT devices and predictive data analytics to reduce unscheduled downtime of its 31 quay-side and 82-yard cranes. AI optimises the crane’s predictive maintenance cycle, which improves performance and the productivity of ship-to-shore operations. 

These examples, though from other sectors,  clearly demonstrate the important role connectivity can play in digitising the maintenance process, facilitating early detection of faults and minimising downtime. This can have a significant impact on the bottom line, which is particularly relevant for the farming industry where costs are spiralling.

There are several organisations already offering such IoT systems to the horticulture sector, including allMETEO, Smart Elements, and Pycno.

Developed by AMA Instruments, InField measures soil humidity and texture, air temperature, wind speed, and sun exposure. Deployed in remote fields or orchards, weather stations can successfully utilise lower-power wide access network (LPWAN) connectivity but will also benefit from 5G, which enables more data-dense observation and edge computing.

Worryingly, the UK is now experiencing wildfires due to increased temperatures, which can have devastating impacts on grazing land and risk to life for livestock. A fire can spread up to 200 metres every minute: by the time a wildfire has been spotted by a farm worker, the damage can be significant. Ericsson SmartForest is a solution that uses sensors and AI to detect and monitor wildfires, enabling faster response times.  

"We're now treating wildfires as business as usual. And the conditions are going to get more extreme as the next two decades move on," said Surrey fire investigation officer Matt Oakley.

Robots in horticulture are being increasingly used for conducting monotonous tasks such as picking and packing fruits and vegetables, from strawberries in polytunnels to apples in orchards. These tasks are increasingly difficult to find human workers due to the repetitive, strenuous nature of the work and low wages. Bringing in robots frees up limited labour to focus on work they find more satisfying.

“The younger generation doesn’t aspire to be an apple picker. Right now … we’re treading water, trying to figure out how we can keep enough employees to get our crops picked until this technology evolves,” said Kent Karstetter, Orchardist.

Robot vehicle units and robot palletisers can also work alongside human pickers, packing and palletising harvested products. This can minimise the need for humans to carry heavy loads and reduce interaction around hazardous equipment such as forklifts, helping to create safer work environments and reducing injuries from workplace accidents or repetitive strain. 

Tim Price, Rural Affairs Specialist at NFU Mutual, said: “By combining modern technology with physical fortifications, farmers are trying to keep one step ahead of the thieves”.

Inadequate post-harvest storage can result in insect and rodent infestations, microbial infections and harmful changes in moisture content – all of which can lead to losses of produce and livelihoods. Connected silo sensors, such as those provided by companies like Blue Level Technologies, can be used to improve the shelf life of inputs and reduce post-harvest losses by monitoring and automatically optimising storage conditions. 

Sensors can also be used to monitor the conditions of goods while in transit throughout the supply chain, helping to reduce damage and loss of food products in their journey from producer to distributor.