If you guessed sensors, pat yourself on the back. The minute technology is starting to have a tangible impact around the world when it comes to labour, sustainability and environmental stewardship. With sensor technology growing ever-more sophisticated, farm managers will soon be able to do more with less, saving time and money while they wager it will also boost their bottom line. Sensor technology is not new, but it is constantly improving. What once used to be commercially non-viable, sensors continue to drop in price and have also become available for purchase commercially. Paired with cutting-edge robotics, the adaptation to agriculture is moving along at a blistering pace. New AG International has learned about three unique sensor applications for environment, ripeness and pests, designed to make a farm manager’s life easier. ROOT AI: Ripe for the picking Horticulture labourers have a job that is intensive, demanding and repetitive. In other words, not everyone is clamouring to apply. A lack of labour options has forced necessity and it begins with sensor technology. With the small yet mighty technological component becoming cheaper and more commercially available, it dawned on Root AI CEO Josh Lessing that the potential was there to help morph the future of agricultural labour. Motivated by the realization that there is an “uncanny need of people to interact with plants, bring in harvest and feed people,” he began to explore ideas.
“How do you move food objects, that are incredibly delicate, with speed?” poses Lessing, who studied robotics. “The market has needed a solution for that fact that they can’t find people to pick their crops. This problem is an existential threat to our ability to feed people around the world – a lack of workers who want to pick produce is a universal challenge were seeing in every single market. This needs to be solved.” As recently as 2015, the sensors used to build robots were expensive and for many applications they needed to be made from scratch. However, a much wider selection of low cost and capable sensors are now available for purchase right off store shelves, making this foundational element of a robotics platform widely available. “The advancements in AI, the resources to build, computer vision models, available hardware, sensors off the shelf that gives robots awareness; the convergence of it all has allowed us to say, ‘now is the time,’” says Lessing. Byte sized The company has built fruit and vegetable picking robots to do the laborious work that has become extremely difficult to staff. But how can a robot do something as exact as plucking a tomato with the corresponding level of gentleness? It all starts with a sensor. The sensor itself, which can fit into the palm of a hand, works like this: thousands of tiny pixels on a sensor’s surface instantly build a 3-D image of the environment it is looking at, in this case, a vine tomato. One of many layers of information being received onto the sensor relates to colour. This allows the sensor to communicate to the robot’s decision-making centre, which then makes an instantaneous choice to pluck or not pluck the fruit. It utilizes what’s called stereovision which is a technique, inspired by how humans see, that takes two 2-D images of the same object from different angles and converts this data into a 3-D image of the environment, a vital offering of cutting-edge sensors. By understanding depth, a sensor can provide a robot with the information it needs to figure out how to get around an object, whether an unripe piece of fruit or vine. “[We] create computer vision models that allow the system to understand when a product is ripe and ready to harvest,” says Lessing. “Certainly, the metrics for each crop are different. Nevertheless, you can always train new crop models by simply annotating imagery of that crop in its natural environment.”
As sophisticated as human eyes and just as fast, the sensors and computers that are all part of the robot work as long as needed, with the same delicacy and accuracy from the first fruit to last. “What robots deliver is resilience,” he explains. “Robotic workflows built with sensors and software allow your farming operation to get extraordinarily nimble very quickly.” Root AI’s tech works because the sensor works. With extremely accurate sensors, a critical first step in changing the face of agricultural labour, it is now possible to communicate unbelievably complex data instantly to robots able to carry out actions. Lessing believes this will change food production worldwide. “There’s no better place to use this technology than in food because everyone needs to eat,” he says.
With the robotic system developed, Lessing’s company sells their technology as part of a harvesting service to agricultural companies and farms eager to solve the ongoing issue of human labour and improve efficiencies. Root AI is not just involved in tomatoes, but also strawberries and cucumbers. With time, Lessing’s hope is that his startup will have algorithms designed for all fruit and vegetable crops that need to be picked, disrupting an analog job with a digital solution. “It’s the digitization of food,” he says. “The digitization of food comes from when you make all these atoms across the supply chain into bits, and that’s where the sensor comes in.”
SEE AND SPRAY: New tech identifies weeds, reduces herbicide usage Overuse of herbicides continues to increase resistance across the United State and many other nations with commercial scale row crops. Not only are herbicides a costly input for farmers, but with greater resistance comes less chemistries available to control weeds in the field. However, new technology is nearly ready to come to market and, when it does, manufacturers are hoping it will be a game changer for farmers. Originally created by Blue River Technology and later acquired by John Deere, the See and Spray technology consists of high-tech sensors and cameras located in pairs that span the length of a spray boom. While the technology started out with lettuce and proved itself and its deep machine learning, the hope through John Deere is now to scale up and bring this solution to global commodity crops such as corn, cotton, soy and more.
"We can use deep learning to solve a problem, [which is the indentification of] a weed from a plant, and that way we unlock this opportunity to only spray the weeds," says Julian Sanchez, John Deere's director of emerging technologies. The unit is not unlike a typical sprayer, the chief difference being sensors integrated into high resolution cameras are plentiful. As a sprayer moves through a field, the sensors are rapidly analyzing any green material that shows up. When it does, the sensors, which are trained through deep machine learning algorithms with the similar technology utilized in facial recognition software, send a message of whether it is a crop or a weed to the central processor, which then will trigger a direct application of herbicide onto the weed. If it is a crop, the sprayer simply passes over it with no action taken. A second sensor-equipped camera take a hindsight image of what was just sprayed and allows the sensor and machine, as well as the farm manager, to understand afterwards if the technology was correct.
“[The] second camera, after a pass, takes another image to validate ‘did we get that right or wrong?’” notes Sanchez. “The nice thing, it provides a feedback loop of ‘did I do it right or wrong?’” According to Sanchez, the sensors become more accurate each time they are put into fields and the end game is, invariably, perfect detection and application rates. “The big breakthrough with machine learning is you can show the system examples of what to find and it can interpolate between those examples,” says Sanchez. “This saves you from having to teach the system every single example of what a weed can look like, which of course in nature is impossible.” So, each time a farmer sprays a field while sensors scour for weeds, the machine becomes smarter every
single time. When testing began, the sensors and machine were unable to correctly identify pigweed. However, after analyzing more than 10,000 images of pigweed in different fields and at different angles and stages of maturity, the sensors are now able to identify the weed just as effectively as any farmer or agronomist. Overall, the numbers speak for themselves. The technology is said to eliminate excess herbicide volume sprayed by up to 70 percent. A reduction of that scale means greater returns to the farm gate without sacrificing efficacy. Currently, the technology is being utilized in what Sanchez calls “green on brown” scenarios, namely pre-emergent fields, and this is what will be available to farmers in 2021. As the company’s machine learning continues to evolve, See and Spray will move into the true test: green on green. Identifying weeds on emergent plants means the sensors must have a greater depth of understanding. Sanchez says that, “green on green is a whole other magnitude to solve.” The release date for that technology is currently to be decided. As field-scale trials continue, John Deere’s AI learning bank continues to deepen as the machines learn more and more examples of what a weed is since they come in many shapes and sizes. SMART PIVOT: Irrigated farm tech continues to push boundaries With water being one of the most precious and expensive resources a farmer can get their hands on, the fight continues day and night to best utilize the precious liquid. What better way than through smart technology, which includes sensors? Created by Lindsay Corp., which manufactures Zimmatic irrigation pivots, its FieldNET remote irrigation management system covers off most wants a farmer would have when it comes to water management while remaining profitable. It focuses on equipping farmers with data and insights to best run their equipment, and it all begins with sensors. First, pivots were affixed with GPS sensors so a farmer would be able to pinpoint exactly where the pivot was in a field. As the adoption of sensor technology increased, it helped farmers know their water pressures and flows were working at optimum levels. Lindsay’s FieldNET Pivot Watch device is smaller than an American gridiron football and weighs just three pounds. The solar-powered device straps onto a farmer’s pivot with a band clamp, like a giant wristwatch. The curved unit is designed so water flows right off, preventing water from pooling and potentially damaging the sensor or other components. While it may look fairly simple, it performs important communicative tasks, offering farmers real-time reporting on the presence of water and current pivot position, status, direction and speed. With alerts from FieldNET going directly to the farmer via computer, phone or tablet, their ability to make a timely decision has never been faster. The USD$400 product could quickly save farmers that much money, and more, in a relatively short amount of time. “That little device is cool because with the remote control and GPS, you are not electrically wired into the pivot, but you know the pivot is moving and running forward or reverse,” says Reece Andrews, the company’s product manager, adding that sensors can also detect if the pivot has stopped, which is especially important during critical times such as the heat of the day. “It gives them total monitoring, which helps them feel in control. Farmers save time by not running around checking pivots, and also save by not putting wear and tear on vehicles, not to mention fuel costs. Another advantage is peace of mind, knowing their irrigation system is running.” To make irrigation decisions even easier, the sensor-driven device can be paired with additional company products that can even link together irrigation schedules, greater insights based off field data, and machine learning. Eventually the tech will grow sophisticated enough to make recommendations of when, where and how much to irrigate. Thanks to a sensor giving specific data, farmers will theoretically be able to make the most-informed decisions related to their irrigable farmland than ever before. Knowing exact application rates means farmers have a greater understanding of when to turn the pivot on or off, which increases the return on investment while also conserving resources, says Andrews. “With data insights from FieldNET and FieldNET Advisor, farmers can often apply less water and actually get as good, if not better, yield,” he explains. The sensors are also responsible for helping drive sustainability. With timely alerts triggered by the sensor, a farmer will be better able to control run-off and leaching that may occur due to overwatering, according to Andrews. The sensors, where everything begins, have become increasingly sophisticated in recent years and Andrews believe the technology will continue to increase for farmers’ overall benefit. ●
Root AI sensors build a 3-D image of the environment it is looking at, in this case, a vine tomato. One of many layers of information being received onto the sensor relates to colour. This allows the sensor to communicate to the robot’s decision-making centre, which then makes an instantaneous choice to pluck or not pluck the fruit. Photo: Root AI
The See and Spray technology consists of high-tech sensors and cameras located in pairs that span the length of a spray boom. As a sprayer moves through a field, the sensors are rapidly analyzing any green material that shows up. Photo: John Deere
Agricultural journalist Trevor Bacque, currently serving as president of the Canadian Farm Writers' Federation, investigates for New AG International
Julian Sanchez, John Deere’s director of emerging technologies. Photo: John Deere
Lindsay’s FieldNET Pivot Watch allows growers to get real-time reporting on the presence of water, current position, status, direction and speed of centre pivots.
Photo: Lindsay Corp.
Researchers at Linköping University in Sweden have developed biosensors that make it possible to monitor sugar levels in real time deep in the plant tissues – something that has previously been impossible. The information from the sensors may help agriculture to adapt production as the world faces climate change. The primary source of nutrition for most of the Earth’s population is mainly plants, which are also the foundation of the complete ecosystem on which we all depend. Global population is rising, and rapid climate change is at the same time changing the conditions for crop cultivation and agriculture. “We will have to secure our food supply in the coming decades. And we must do this using the same, or even fewer, resources as today. This is why it is important to understand how plants react to changes in the environment and how they adapt”, says Eleni Stavrinidou, associate professor in the laboratory of organic electronics, department of science and technology at Linköping University.
Sugar sensors The research group at Linköping University, led by Stavrinidou, together with Totte Niittylä and his group from Umeå Plant Science Centre, has developed sugar sensors based on organic electrochemical transistors that can be implanted in plants. The biosensors can monitor the sugar levels of trees in real time, continuously for up to two days. The information from the sensors can be related to growth and other biological processes. Plants use sugars for energy, and sugars are also important signal substances that influence the development of the plant and its response to changes in the surrounding environment. While biosensors for monitoring sugar levels in humans are widely available, in particular the glucometer used by people who have diabetes, this technology has not previously been applied to plants. “The sensors now are used for basic plant science research, but in the future they can be used in agriculture to optimize the conditions for growth or to monitor the quality of the product, for example. In the long term, the sensors can also be used to guide the production of new types of plant that can grow in non-optimal conditions”, says Stavrinidou. Plant metabolism The mechanisms by which plant metabolism is regulated and how changes in sugar levels affect growth are still relatively unknown. Previous experiments have typically used methods that rely on detaching parts of the plant. However, the sensor developed by the research group gives information without damaging the plant and may provide further pieces of the puzzle of how plant metabolism works. “We found a variation in sugar levels in the trees that had not been previously observed. Future studies will focus on understanding how plants sugar levels change when plants are under stress”, says Stavrinidou. The research is mainly funded by the European Union’s Horizon 2020 research and innovation program. Additional funding comes from: the Wallenberg Wood Science Center, the Swedish Foundation for Strategic Research, the Knut and Alice Wallenberg Foundation, the Swedish Research Council, and the Swedish Strategic Research Area in New Functional Materials (AFM) at Linköping University. ●
Every backyard gardener knows how hard it can be to tell when to water the plants. Multiply that by tens or hundreds of acres and it’s easy to see the challenges growers face keeping their crops healthy while managing water resources wisely. To determine water needs accurately, growers hand-pluck individual leaves from plants, put them in pressure chambers, and apply air pressure to see when water begins to leak from the leaf stems. That kind of testing is time consuming and means growers can only reach so many areas of a field each day and cannot test as frequently as needed to accurately determine optimal irrigation scheduling patterns. A group of researchers from UC Riverside and UC Merced have received a grant for more than USD$1 million from the U.S. Department of Agriculture through the National Science Foundation’s National Robotics Initiative to address these challenges. As part of the project, the group is developing a robotic pressure chamber that can autonomously sample leaves and immediately test them on site to provide the freshest data. The system will work to gather data even in large fields, and over a period of time, rather than just providing a snapshot. Frequently updated data can help growers better plan irrigation schedules to conserve water, optimize the time and effort spent by crop specialists tasked with determining and analyzing lead
water potential, and help decrease some of the costs in the food-production chain. Current measuring techniques involve collecting leaf samples and transporting them to an off-site location, where testers can use very accurate, expensive pressure chambers; or sampling and analyzing leaf samples in the field using hand-held pressure chambers. “In the first category, leaf samples can get mixed up, making it impossible to track them back to the specific areas of the field they came from,” says Konstantinos Karydis, assistant professor in the department of electrical and computer engineering at UC Riverside. “In addition, the properties of the leaf might vary given the time elapsed between being sampled and being analyzed, which in turn may yield misleading results.”
Hand-held instruments in the field can be less accurate, but testing can be done multiple times with different leaves from the same plants. This method is time and labour intensive, and must be undertaken by specially trained personnel. Stefano Carpin, professor of computer science at UC Merced, has already worked with colleagues at UC Davis and UC Berkeley to create the Robot-Assisted Precision Irrigation Delivery, or RAPID, system, which travels along rows of crops adjusting irrigation flows according to sensor data that tells the robot precisely what’s needed for each plant. The project will use the same mobile base robot as in RAPID but equip it with a custom-made robotic leaf sampler and pressure chamber being designed by the researchers at UC Riverside, and pair it with drones that can survey the fields and direct the robot to areas of interest. “Using this process, growers could survey plants all day long, even in large fields,” says Carpin. The four-year project will support graduate students as well as summer research opportunities for undergraduates. The project has four phases: development of the chamber; developing machine vision so the robot can “see” the water coming from the leaf stems; coordinating multiple robots — in the air and on the ground; and evaluation. The researchers, which includes Amit K. Roy-Chowdhury, professor in the department of electrical and computer engineering at UC Riverside, and Joshua Viers, professor of environmental engineering at US Merced, plan to have the first set of automated pressure chamber prototypes fabricated by spring 2021, and to evaluate their performance and refine designs in controlled settings over spring and summer 2021. They expect to have a completed setup by winter 2022, so they can begin controlled field testing. “We have to be quick about it because if we miss a peak growing season, we have to wait another nine months for the next one,” says Carpin. “We’d like to be able to start testing next summer and test every summer, and we need to be able to maximize the tests.” When all of the components have been designed, the designs and code will be made open source, and all the data collected during the project will be made available to the scientific community, the researchers wrote in their proposal. The project came about after Carpin and Viers, director of the Center for Information Technology Research in the Interest of Society, or CITRIS, at UC Merced, had been talking with area farmers about the challenges of growing almonds and grapes. Karydis and Roy-Chowdhury had been hearing the same challenges from citrus and avocado growers in the Riverside area, so the four partnered up. “California agriculture presents a challenge in terms of scalability,” says Carpin. “But this an exciting collaboration because we’ll get to develop a system that will work on different kinds of crops.” ●
Agricultural platform Solinftec and IBM are collaborating to provide real-time decision-making solutions for agriculture. Through this partnership, Solinftec utilizes local weather data and forecasting from The Weather Company, an IBM business, to deliver digital agricultural solutions that leverage operational efficiency, use of inputs and agriculture compliance, ultimately reducing environmental impacts.
According to Daniel Padrão, director of operations at Solinftec, accurate weather data is crucial on the farm, not just for operational excellence, but to better manage inputs and to increase yields. “We are leveraging IBM’s unique insights to provide farmers and retailers with the climate-based information they need,” he said. Fernanda Borges, IBM business development leader for weather in Latin America adds, “It’s not only a matter of weather data, it’s what the weather data can bring to farmers and to their businesses.” Solinftec combines real-time agronomy, machine and weather data to feed algorithms and drive operational excellence and compliance. “As an example, Solinftec users in Brazil have seen an increase in the efficiency of their operations by up to 40 percent,” said Padrão. Since 2018, Solinftec has been using IBM weather data to leverage solutions that benefit its consumers and society, promoting innovation and addressing critical problems related to agriculture, food production, sustainability and waste reduction. “This is a key partnership for IBM and we are working together to explore new technologies that can expand Solinftec’s capabilities, delivering higher value data to their customers and more precise business insights,” said Jose Luis Spagnuolo, distinguished engineer and director WW Tech Sales for IBM AI Applications. ●
VOA has an online platform where farmers in Brazil can book a drone to perform an application. Please provide a description of your services. VOA’s solution seeks to sustainably protect crops from pests and diseases by deploying a fully integrated drone service through our platform VOAline.com, that allows for precise monitoring and application of biological agents and rational application of chemical pesticides and fertilizers. Our platform, and competitive advantage to other drone ag-tech companies, is a “marketplace” for farmers (demand) and pilots with drones (supply), creating connections based on geographic location, availability and technical capacity. It also allows for traceable and transparent services, as the services are pre-programmed by the farmer himself on the platform, and once the order is complete, the farmer can evaluate the service report and rate the quality of the service.
VOA does not manufacture drones, but it does modify them depending on what is being applied. Can you elaborate on this, with reference to your “family” of drones? Indeed, selling drones is not our business model, but we do have them available in case a pilot needs one (we provide them to the pilots for free), and we sell them to franchises as part of the contract with them. Actually, we would love another company (like DJI) to flood the market with the drones we need, but at the moment no company has all the application mechanisms that are required in the Brazilian market – that is why we buy the parts and assemble them ourselves, in the case of VOA-001, and 3D print them, in the case of VOA X1. We have two different types of drones with several application mechanisms. VOA-001 application mechanisms include spraying of micro-organisms or liquids, application of macro-organisms in the form of living eggs, and application of macro-organisms in the form of biodegradable capsules.
The drone is taken to the location by the drone pilot. Apparently, it won’t become activated and take off unless it is in the right start location specified by the farmer. What’s the main problem that this functionality overcomes? Actually, the wider problem VOA is trying to solve, specific to drone ag-techs, is that other drone ag-techs out there that carry out these application services are not connected to a platform that runs the entire fleet. From planning the operation, executing the operation and evaluating the operation – it is not possible to do this smoothly without a platform. Current Brazilian companies book the applications via WhatsApp – quite literally texting the farmer asking for a good day. This way of working is not scalable, and so VOA’s vision was to create a scalable operation that can only be done with a platform like ours. The technical fact that the drones do not take off if they are not in the right place at the right time just speaks to the fact that they are connected to a platform that is able to manage the drones. If we do not know where the drone is, we are not able to tell where it could go after (for another service in another location), nor whether it has actually done the service it needs to do, nor if the quality of the service was up to the quality that it needs to be. Additionally, of course, it creates a level of security that our drones will not just disappear, and pilots will go off to do their own applications and receive payment outside the platform. Your largest drone can carry 15 kilograms of liquid fertilizer. What kind of grower in Brazil will use this service? Many, among them vegetable and fruit growers. So usually high value crops, with smaller distances than large sugar cane plantations for example. VOA refers to itself as the Uber of drone services in Brazil. Could you explain how this relates to your business model? VOA is like the Uber of drone services because it acts as a “marketplace” where supply (pilots with drones) can find access to demand (farmers that need services). Same as Uber, which acts as a marketplace for drivers and riders. VOAline.com matches pilots and farmers through an alogarithm that optimizes efficiency based on geographic location, availability and technical capacity of the pilot. Moreover, VOA is different from other drone application companies that don’t have a platform; it allows for a traceable, manageable and transparent service. Your largest sector by area is sugar cane. Do you know roughly how many hectares of sugar cane VOA has serviced? Would this be Trichogramma application or other? Yes, our platform logs everything. Trichogramma (living eggs) 15,000 hectares and Cotésia (capsules - living wasps) 710,000 hectares. We will be doing more by end of the year. How has VOA been financed? Will you be doing any further funding rounds? We have angel investors who financed USD$800,000, and we are just closing a second round with an international accelerator. You’ve mentioned the U.S. market. When do you think that might be, and are there any specific modifications (equipment or otherwise) that you will need for that market? We are still doing the due diligence of what is the best way to enter the U.S. market. But we are not changing our business model; the market is quite concentrated on five main crops with very similar needs in terms of protection and nutrition. ●
Ag-Analytics and Davide Cammarano, Purdue University associate professor of agronomy, have announced a recently established research partnership. Using precision agriculture data, Cammarano’s research team will develop farm management strategies that optimize economic outcomes for businesses and individuals. Ag-Analytics is a farm management platform that specializes in precision agriculture data analytics, remote sensing data analytics and risk management. Through innovative tools and services for farmers, crop consultants and research partnerships, Ag-Analytics helps users mitigate risks, decrease environmental impacts, improve yields and increase profits by making better-informed farm management decisions. This partnership will allow growers to opt in to work confidentially with university researchers with field-level precision agriculture data. All data will be anonymized and confidential, and researchers will be working only with growers who opt into the agreement with their university. “This venture will be mutually beneficial, advancing my research while helping Ag-Analytics develop tools for farmers and optimize their production,” Cammarano said. “The hope is that this partnership will create a living research hub for integrating digital tools and models that support farmers.” According to Joshua Woodard, CEO and founder of Ag-Analytics, partnering with Purdue enables precision agriculture research, which drives the industry forward toward data-driven decisions and deeper analysis of field operations. “We are excited to establish a partnership with a premier agricultural research institution and a close neighbour to our West Lafayette, Indiana, headquarters.” Farmers can easily sign up to participate at https://profit.ag/Purdue. Signup is free, and farmers who sign-up will receive a free Ag-Analytics account, which offers access to high-resolution satellite NDVI imagery, historical and current weather conditions, insurance estimates and other key elements. ●
The base robot for the new plant-moisture-measuring system researchers are developing will navigate rows of crops to reach individual leaves and stems.
Chiara Diacci, PhD student at the laboratory of organic electronics and first author of the paper, inserts the sensor in the young tree.
The biosensor can give information about sugar levels without damaging the plant.
Photo: Thor Balkhed
Photo: UC Riverside/UC Merced
Brazil-based VOA is a specialist in drone application of biocontrols and fertilizers in Latin America. New AG International Chief Editor Luke Hutson spoke with Nicole Engels, Head of Growth with VOA, to find out more.
