Artificial intelligence and robotics are becoming more ubiquitous in every facet of our lives, including in how farmers run their operations. The future of semi- and fully autonomous farm equipment is closer than you think. Understanding present and future technology and considerations for when to adopt it is an important financial decision.
How technology is evolving
A survey from the University of Nebraska-Lincoln found that 80 percent of farmers use GPS navigation systems. However, technology for semi-autonomous and autonomous is still in a conceptual phase. Yet Matt Nielsen, corporate communications manager for Autonomous Solutions, Inc. (ASI) pointed out, it is very functional. Nielsen explained how the company’s software and hardware technology is integrated into the CNH Industrial (Case IH, New Holland, Steyr, etc.) platform.
“It incorporates a variety of sensors, cameras, GPS, radio, communication … to bring these sensors all together. Then Mobius platform, (ASI’s command and control platform for robotic vehicle operations), is what gives the operator the control of that vehicle or vehicles,” Nielsen said. “Speaking in concept, in a scenario, an operator would be interfacing with those vehicles through a tablet or through a computer at that farmer’s home or job or anywhere to control those vehicles (and) to monitor what they’re doing while freeing him up to do other things as well. Those sensors provide the capabilities to till a field, seed a field, plant a field, and then also to work in tandem.”
Nielsen explained its technology, in conjunction with equipment manufacturers, can resolve two common challenges for autonomous farming equipment. First, the technology can correct coordination between multiple vehicles when they’re using different types of implements or separately sized implements, which has been a recent and common issue. Second, it can identify and avoid obstacles, plus provide the operator with a path around the obstacle and opportunity to clear the obstruction.
Researchers see a progression that is moving into a human-machine collaboration environment. “What we prefer to do is program the machine so when it sees something that it’s not expecting, it (will) ping the farm supervisor,” explained Herman Herman, director of the National Robotics Engineering Center at Carnegie Mellon University in Pittsburgh. “You can imagine you have the path on your smartphone and the tractor shows you the issue on your smartphone. Then you can decide whether you want to remove the tree or to tell the tractor to go around it or do other things. Coupling human and machine, we believe, is the right thing to do.”
With new technology comes the unpredictability of the law. Todd J. Janzen of Janzen Agricultural Law LLC in Indianapolis explained there is still a bit of gray matter to autonomous vehicles.
“It depends what the machine is doing,” he said. “For most farm activities, (including tillage and harvesting), I think that there’s no law that would prevent an autonomous machine from performing its duties out on the field,” Janzen said.
However, there are exceptions: noting some states have requirements for nutrient application or spraying pesticides that are done by certified operators.
“It’s an open-ended question whether or not a farmer watching the machine from a computer back at the farm would qualify as being done under a licensed or certified applicator,” he said.
When using semi- and fully autonomous farming equipment, it often depends on the crop being harvested, said Scott A. Shearer, Ph.D., P.E., professor and chair of The Ohio State University’s Department of Food, Agricultural and Biological Engineering Department.
Shearer used apple growers in Washington state as an analogy to illustrate the difference in approach to a more mechanized and potentially fully autonomous harvesting.
“They’re coming up with tree canopy structures that lend themselves to automated harvest,” Shearer said. “If you spread all your fingers out on your hand and you keep your hand flat, that’s pretty much what the tree looks like. Then you have two rows of those and you lean them over at about a 30-degree angle so the branches of the top of the trees touch. If you put your hands at 30-degree angles, underneath the canopy, the apples hang down so you can go through there with a robotic harvester and harvest the apples really easily.”
Shearer highlighted that while harvesting apples with robotics has real potential, it’s not as simple as creating a machine to harvest the fruit. Compared with crops like corn, wheat and soybeans where the tractors are being equipped with more features traditionally handled by the operator, orchard owners and citrus growers must create a growing environment, which requires a 25- to 30-year commitment when they plant to make their orchard or grove conducive for automated harvesting, according to Shearer.
GPS functioning and technical considerations
Most systems are exclusively reliant on GPS; however, Kraig Schulz, president and CEO of Autonomous Tractor Corporation (ATC), noted there are still lingering issues.
“GPS is good, not great. It’s a very weak signal,” he said. “You get fades and sunspots, reflections and ionospheric interference, and all sorts of challenges that make GPS good most of the time, and then not great sometimes.”
Schulz explained the real-world potential for GPS errors based on the how the system works. Satellites orbiting the earth transmit a signal through the atmosphere, notably the ionosphere. Refraction occurs due to the ionosphere’s varying thickness. When the signal is refracted, errors emerge in the GPS signal.
Base or real-time kinematic (RTK, a satellite navigation technique used to enhance the precision of position data) stations are put in place to correct the errors as best they can. However, the further the RTK station is from a tractor, coupled with processing and communication time and a tractor moving 10 to 20 inches per second, there’s a delay in processing time. Schulz explained that GPS is good at determining where the tractor has been, but not as good at indicating where it is in real time.
What Schulz recommends to improve GPS data is closer RTK stations within the field. The closer to the tractor the RTK correction station is, Schulz explained, the less latency occurs. The shorter distance provided by the RTK correction station to the tractor, compared with the distance between the satellite and tractor as Schulz detailed, the closer the GPS data is in real time to the tractor’s real-time location.
Herman, of the National Robotics Engineering Center at Carnegie Mellon, explained that while GPS systems are improving because the number of base stations is improving, there are environmental factors that can’t be as easily improved. GPS interference can still occur due to towering trees or buildings or even “high faulted transmission lines.” As with Washington state apple growers, step hills can shadow the GPS satellites. Similarly, the satellite’s position – fewer satellites available for signal transmission – can also make it make sensitive to disturbances.
GPS’ role in autonomy
Herman said that if you know where you are, a GPS system would suffice. But in many applications it’s also important to know where you are in respect to things around you, and that’s not captured by GPS.
“In many farms, many people move the collection box, [the] cart or other implements in many places in the field,” he explained. “To make it safe, the robot needs to be able to detect where the humans are located – whether they are on a ladder … sitting on the ground or just walking around, or if somebody parked their pickup truck at the end of the road.”
“We do use GPS so that farmers can use the field work patterns they have already established with their GPS systems,” Schulz said. “We just supplement the GPS data so that we are not reliant on GPS, which has many well-documented issues.” One way to increase accuracy further can be accomplished via an on-farm system.
Schulz’s approach, a product known as Autodrive, provides farmers with on-farm data. Using a laser-radio navigation system, this technology has sensors on the tractor and the field to create hyper-local data for its current location. It provides three safety zones when operating: a 5-foot verification zone, a 15-foot breaking zone and a 30-foot safety zone.
Herman explained how automation research on agricultural vehicles is progressing from playback to nonplayback technology. Using the example of small- and medium-Pennsylvania apple orchards, he explained that one method is to travel through the orchard to train the tractor through all rows and it can “play back that part.” However, if there’s an unintended object, such as a flooded row or a downed tree, it can’t adjust to that unforeseen event.
“The problem with teaching by this playback system is that it’s very inflexible. So if you just use this playback system, the robot can just replicate what you show them, but it cannot come up with a new path or sequence,” Herman said.
For the foreseeable future, instead of using a playback system, Herman said autonomous software will contain a layout of the farm and will be able to navigate from one point to the next by real-time information from its laser scanner and video camera system. While Herman explained a tractor can be programmed to avoid and re-route itself around a downed tree, for example, the next phase is a man and machine working hand-in-hand when an unexpected scenario occurs to maximize safety with autonomy.