The Earth now needs to feed 8 billion people… every single day. The only way to meet this almost unimaginable demand for food is to continuously improve the effectiveness, efficiency, and cost basis of growing and delivering that food. That is the agriculture industry’s biggest challenge in the 21st century. Increasingly, agriculture is relying on artificial intelligence (AI) to deliver those gains.
John Deere is one of the leaders in applying AI to agriculture. The company has been supplying farmers with a hands-free driving option for its tractors for 20 years. Building on that technology, John Deere now offers tractors, such as its 8R series, that can work in the fields autonomously. Farmers that own older John Deere tractors with hands-free driving capability can retrofit their tractors with AI-enabled autonomous operation.
Autonomous operation allows the tractors to operate 24/7 in the field without an operator. After setting up the field boundaries by driving the boundary with a vehicle equipped with the company’s GPS-based StarFire receiver, the farmer can mark do-not-enter zones for obstacles within the field such as waterways, pumps, and windmills. Once the field has been mapped, the farmer creates a work plan for the tractor, based on the equipment it will be towing, such as a tiller or a spraying boom. The work plan includes settings for work speed, the rate of turns at the end of each row, and the settings for the towed equipment. In the case of a tiller or plow, for example, the setting might include the soil work depth of the tool.
Then, the farmer drives the autonomous tractor to the field and activates the autonomous mode, and the tractor proceeds to execute the work plan for that field, without an operator in the cab. The tractor is equipped with cameras so that if it encounters an unexpected obstacle, such as a dropped hay bale, the tractor will stop and alert the farmer to the situation using a cellular connection, which can include live video from the tractor’s cameras. The company demonstrated these capabilities live, on actual farm acreage, earlier this year at the John Deere Tech Summit held in Austin, Texas.
John Deere’s AI implementation allows multiple pieces of autonomous equipment to coordinate with each other. This capability allows multiple autonomous farm vehicles to work in the same field. One example where this capability is quite useful is when a combine grain harvester needs to coordinate with a tractor pulling a grain cart so that it can unload the harvested grain. It takes about two and a half grain carts to unload a harvester. With autonomous operation, the tractor pulling the grain card can match speeds with and synchronize with the combine so that the combine can continue harvesting while it’s unloading harvested grain to the cart. One combine can synchronize with multiple carts so that harvesting and unloading become one continuous process.
Another critically important factor is the need to improve agriculture’s efficiency by reducing the use of fuel, fertilizer, insecticide, and pesticide. Reducing the use of these increasingly expensive consumables reduces the operating costs of farming and is better for the planet because it reduces the amount of pollution and soil runoff created by farming operations. Efficient fieldwork, as illustrated by the above examples, reduces fuel costs.
Further use of AI can reduce the amount of fertilizer, insecticide, and pesticide used in the fields. One company focused on this aspect of farming is Blue River Technology, which has developed a 120-foot spray boom equipped with 36 cameras and AI video processing. As a tractor pulls the spray boom over a field at 15 miles/hour, the cameras capture video of oncoming plants. AI processing of the video, which is entering the AI system at 15 gigapixels/second, identifies weeds in real time and activates sprayers at just the right time to apply insecticide to the weed with 2-centimeter accuracy. The result of this precision spraying is to reduce the amount of insecticide used by two thirds. Blue River Technology’s founder and CEO Jorge Heraud discussed the capabilities of his company’s products at NXP Connect, held earlier this year in Santa Clara, California. John Deere purchased Blue River Technology in 2017.
InsightTrac takes an entirely different approach to AI-enabled pest control with its autonomous Rover agricultural robot. This robot is designed to fight insect pests that plague almond orchards such as the navel orangeworm. These pests take delight in finding nuts that remain on the almond tree after harvesting. These nuts turn rotten over time, develop a fuzzy green outer hull and serve as enticing food sources for insect pests. The insects burrow into the rotten nut, called a “mummy nut,” and deposit their eggs. The eggs mature into larvae, feed on the inside of the mummy nut and make the almond tree susceptible to fungal infections. Losses due to navel orangeworm damage can be as much as $300 per acre. California alone currently has about 1.3 million acres of land planted with almond trees, so the potential crop losses in California can amount to roughly a quarter of a billion dollars annually.
The conventional method for clearing mummy nuts from an almond tree is to shake the tree vigorously by grabbing the tree’s trunk with a large vibrating machine. The vibrations cause the mummy nuts to fall off in a process called “sanitation.” However, mummy nuts can be stubborn and some of them inevitably refuse to leave the tree when shaken.
InsightTrac’s autonomous Rover robot takes an entirely different approach to mummy nut sanitation: it shoots the mummy nut off the tree with pneumatic pellet guns mounted on swivel turrets. The robot is equipped with two such guns, and each gun is equipped with an AI-powered, binocular, vision-based system to detect and target the mummy nuts. The guns fire biodegradable pellets that resemble paintball ammo and are said to be accurate to a range of 30 feet. A Rover can fire as many as 3600 pellets per hour and, with a hopper capacity of nearly 130,000 pellets, a Rover can autonomously roam an orchard for up to three days before it needs a pellet refill.
Rovers roam the almond orchards at 2 miles/hour using a combination of GPS and LIDAR, so they can navigate in all sorts of weather, including rain or fog. The Rover robot’s image-recognition system employs a ResNet convolutional neural network that has been trained with thousands of images of mummy nuts, leaves, tree branches and birds. (See “Autonomous Robot Tank Eradicates Almond Pests with One Silent Shot.”)
The pieces of agricultural equipment from John Deere, Blue River Technology, and InsightTrac described above are just a few examples of how agricultural equipment makers are starting to use on-device AI to improve efficiency on the farm and out in the fields. Robotics is at the pinnacle of the IOT, and, as the above examples demonstrate, AI technologies advance IOT by greatly expanding sensing, cognition and autonomous capabilities. Different forms of robots are currently being integrated into our daily lives and are poised to transform numerous industries, including agriculture.
