John Reid, Center for Digital Agriculture
This is a special edition of the Closing Market Report. I'm University of Illinois Extension's Todd Gleason out of the office for the afternoon, so no update of the commodity markets. The following presentation was made by John Reed, the director for the Center of Digital Agriculture at the University of Illinois during the March 4 All Day I Got Look at the Beef House in Covington, Indiana. Reed spent nineteen years at John Deere and has recently returned for his second stint at the U of I.
John Reid: 00:38So thank you so much and thanks for the kind intro. And I just want to kind of start out is that how many people use automatic guidance today on your farming systems? Just raise your hand. So, your hands up. How many people were using it fifteen years ago?
John Reid: 00:58Okay. Quite a bit fewer. How many people use a farm management information system to manage your machine operations on your farm today? A few. Okay.
John Reid: 01:11So think about that for this presentation because a lot of this is an evolution of technology and guidance. I used to when I first was working in automatic guidance, I was driving up and down the road and you don't see very many of the GPS receivers on vehicles. You kind of wonder if the technology that's being developed is actually maturing very quickly. But then you start seeing over time that there is more broader adoption as there's value for this. I think the thing I want to kind of leave you with on automatic with autonomous vehicles is many of the examples we're seeing today are extensions of all the things in guidance and all the other types of technologies, including the farm management information system, and then using those together with some opportunities to, in some situations, to remove the operator from the machine.
John Reid: 02:05So, just to get started with this, what are we seeing in terms of autonomous vehicles? There's of press release and we live in an age of social media, so you see CES, for example, John Deere announces their new extension. For several years they've been talking about this. Last year they talked about doing one production step, tillage, and showing how it could be done autonomously, whereas this year they're much broader vision. They're talking about how these production steps could be more of them.
John Reid: 02:39So that's heading in the right direction if we're going to have autonomous agriculture on a farm. Can't just be used for one niche thing and then all the rest of the time you don't get utility out of it. So that's good. Actually, really interesting with John Deere, they're also seeing that this farm management information system, which they call Ops Center, links to other spaces even. So, it's used for agriculture today, if they're going to do autonomy in construction or in golf courses, these other places where there's labor and lots of machinery challenges, then you need to have the same kind of thing.
John Reid: 03:12So, what is it in an autonomous tractor? First of all, it's a highly automated tractor to begin with. It doesn't have to be. Actually, you'll see a lot of companies that will retrofit a vehicle by putting autonomy kit on an existing tractor, but if it's highly automated, it's got more electronic surfaces that can be controlled through software, makes it easier to integrate with the autonomous vehicle. Actually, we're talking about the example in Champaign.
John Reid: 03:39That was the first Magnum tractor from Case IH that had a CAN bus on it. And for a researcher, it was really a nice way of doing autonomy because usually I had to give the tractors back every year to the company. I'd spend six months making it into a robot tractor and in three months doing work, well, with the changes of having electronics backbone and networks, it became a one week job of converting a tractor into an autonomous vehicle. So, you've got this highly automated tractor. It's got an ability to sense what's happening in this environment that's called situational awareness.
John Reid: 04:19Most commonly we think about GPS telling where the vehicle is and how you could use that to plant straight or to do operations more efficiently through automatic guidance, as an example. But it has other situational awareness that's needed too, because the most important part of a tractor operating in the field is the operator sitting in the seat and sensing what's going on around the vehicle and the implement system, as well as understanding what's happening in the field, where do you want to go. So it's just not an issue that nobody does automatic tractoring or just driving the tractor around the field. You have an implement, you're doing a work task, you're trying to get a job done, so there's extra sensors needed. Today, we're starting to see cameras and other types of sensing like that that are looking out ahead of the vehicle to look for people, to sense the rows, to and then also those cameras are looking around the vehicle and seeing the implement and trying to understand is the implement doing its job.
John Reid: 05:22Then, as I mentioned, you see this with some OEMs coming out. Nearly every OEM is talking about autonomy today and how they're exploring it, getting involved in it. You don't see very many price tags for it yet because a lot of this is innovation that is still emerging and is still evolving. I think it's almost like working with individual customers to learn is part of what many companies are doing. However, there are companies that are selling retrofit autonomy.
John Reid: 05:53Sabanto is one out of the Chicago area that will take any color vehicle and basically add an autonomy kit to it. Then they have a management system that helps run that vehicle autonomously and it can be remotely monitored and managed. What's really interesting in this space, though, is really not autonomy of the driving, but autonomy of the implement itself. If you've heard of precision spraying technology like ExactApply from John Deere or other See and Spray types of technologies, These technologies are putting the AI and the intelligence on the implement and allowing the implement to work at a level that's beyond what has been possible in the past, being able to very precisely spray individual weeds and save chemical and get those benefits, but then still have an operator in his seat. And I think this is really kind of an important point about this is that if we are going to be doing these operations, we need implements that can be able to sense things that are working and not working that today depend on the operator or the farmer in the seat to detect.
John Reid: 07:03So think about a debris on a plow shank or some types of effects, something breaking. Right now, lot of the implements are not instrumented to be able to tell that something's wrong. So, you really are going to have effective autonomy, don't want to kind of come back to the field where a tractor just finished operating and seeing that it hadn't completed its operations. So anyhow, this intelligent implement side is not really autonomy, but it is an element that also has to emerge if autonomous agriculture is going to take place. So again, we have the tractor, we have guidance systems, we have situational awareness or perception on the vehicle.
John Reid: 07:47And then usually these are tied into some kind of farm management information system for autonomy. That system is used to essentially understand the fields where you're going to operate, understand what kind of path plans and AI is used to generate those plans, to cover a field. For some operations like tillage, that's all you need. You basically can designate that, the operator or the farmer moves the vehicle to the field and then you can get out and execute the task. I want to come back to this because we're seeing a lot of the visual videos of demonstrating the task working, but when you really think about this, a lot of operations, like your planting or anything that has spring, anything that has materials and inputs that you have to manage, there's a lot of work besides just the execution of the planting or spraying task.
John Reid: 08:46Have to consider how do you resupply and actually in some cases, there's a robotics company that's looking at automating the material logistics and moving materials to and fro, which seems like also a good idea and be really important for a fully autonomous system. So you have these operations taking place, and I want to talk about the jobs to be done. People are showing autonomy in terms of, Okay, I'm going to get out of the tractor, hit go, and this vehicle is going to perform its tillage operation over a field. But, some of the big challenges are getting to that point of being able to hit go in running that application. So, for example, if you start at your farm site and your field is some distance away, you can't autonomously go on the highways today.
John Reid: 09:42It's not possible. In fact, I think it's a really hard challenge. I don't expect to see that. Even in automotive, we're seeing that autonomy isn't where it needs to be. And actually, in agriculture, I think it's got more potential.
John Reid: 09:55But, okay, have to move the equipment to the field. And maybe you have materials and things to set up the job that are done. Then actually when you're running the job, you're monitoring the task and these are things that the farm management information systems can do. They can essentially load data that understands what operation took place, how accurate was it done. And as as long as you don't have any materials to supply to the vehicle, like seed to a planter or or fertilizer, then then, you know, it's probably okay for those types of operations.
John Reid: 10:30But if you need to logistically do those things, then somebody needs to be involved or additional autonomy is needed to make those things happen. And then after the job is done, tractor finishes, stops, shuts down, well, somebody has to go fetch it and bring it back and clean up the operation. So all the things that you have to do when you're doing the operation, really we're only talking about one piece of it where the vehicle is driving autonomously and doing these operations. So it's really very, very possible to do it today. The benefits of it are still somewhat emerging.
John Reid: 11:06And there are places where there are benefits, but I just want to kind of point out that this is still kind of the first inning of a long game and we're at the very beginning of seeing what's possible. Like guidance at the very beginning, only a few people, lead adopters were adopting it. I'd expect this is even more challenging because of the accumulation of technologies that are needed to make autonomy happen and then some of these other operations to get the job done, which is part of the complexity of that task. So, some good news, though, are there are probably many cases which could be compelling examples of autonomy. One of them, for example, you see with combine harvesting operations where you have a tractor and a green cart that's moving between the combine and the edge of the field to transport materials.
John Reid: 12:00Can that be made autonomous? Because you're kind of in an open field, there's kind of a fixed location, except for the combine, which is always moving, and autonomy could be a way of helping a single operator be able to harvest and the logistics management, at least to the edge of the field, can be handled. And that's going to be an example where these accumulation of technologies can come together to provide that type of opportunity. In spring, chemicals in orchards, and I know that probably most of you aren't dealing with specialty crops, but in those kinds of operations, you may not want to be on the vehicle when you're spraying. Some of these chemicals have a long period of time before you can reenter the orchard, so why not get the operator off the machine entirely?
John Reid: 12:47So where do these examples pay off? I'll give a couple of examples for you. It tends to be on, for many of the applications, where there's a lot of already labor involved and autonomy is a way of getting better performance output out of labor that could be low skilled and displace some of that. So one example, there's a study by the Western growers about a system called carbon robotics. It's not an autonomous system, but it's a very intelligent laser weed control system.
John Reid: 13:26Very expensive. It's about a million dollars just for the implement. But in those operations where you have three crews of 25 people weeding in the field, this kind of system has a payback and can provide value in that particular scenario. So, again, that's one scenario. Another example that I worked on in the past was working with citrus operations in Florida, where they have lots of acreage, they're spraying chemicals or they're mowing, and they have just fleets of machines that are doing this all the time.
John Reid: 14:04And we were able to show that we could take the drivers out of three to five machines and train a new kind of skill level, somebody that was a mission manager that could sit almost like a security guard, either in a truck or in a remote air conditioned office. And they could monitor these systems working. And from our data working with these operators for over eighteen months, they achieved 30% more productivity than people that were driving the machines. Part of that was because, through planning, the orchards are somewhat confusing and they're not all just perfectly straight rows, so they could avoid passing through places twice. That was one improvement in productivity.
John Reid: 14:52The other thing was that a human operator in these machines with trees being close would only drive a certain speed, but it was safer to drive faster with the autonomy system. And it wasn't more like in a cab, you would be nervous perhaps driving that speed, but with autonomy it was visibly and actually performed safely in that way. So, what's coming with these technologies is that, as I said, I think it's a long journey. We're seeing a lot of the initial things like we asked in the guidance questions. There are a few early adopters that are trying these technologies.
John Reid: 15:31Think we're going to see moving through the hype as industry gets more experienced. We're going to see more and more examples of this. It is not going to be for everyone. In fact, I would say if you are interested in autonomy, really the journey is still start with automation and get the value of the productivity of automatic guidance and some of these other type of things. And integrating into farm management information systems is kind of a next step that leads to this.
John Reid: 16:01As I said, twenty five years ago when I was looking at guidance, actually similar to what your answers were here, fairly small adoption rate that grew over time to higher numbers as these systems matured and the costs came down. Autonomy will be more complex because it's like a suite of technologies that integrate together. And yes, the technology readiness is there. We're still working on the viability and understanding the business model, and there is some customer value on certain types of applications in terms of especially around labor, productivity and performance. So I'll just stop there and use the time for questions if, anyone has any.
John Reid: 16:53Yeah. I it's it's it's certainly putting sensors in the ground in the soil would be something that many many have looked at. And so I I don't know that it's commercially available yet, but, you know, I know there is things like a John Deere has something called exact shot that's spraying a little bit of chemicals on the seed at planting time with their with their high precision exact emerge. But, yeah, I think that's types of devices that we use today for connectivity, sentencing and things like that. Can they be integrated into the ground engaging parts?
John Reid: 17:29It's certainly a possibility.
Todd Gleason: 17:31So the question was whether or not you could soil test as you were planting or doing some other operation simultaneously. Other questions about autonomous agriculture that might be on your farm. So I've got a series of questions for you. The university to this point, and we started with autonomous ag, with large tractors, have moved to smaller vehicles, usually robots. What's the in between or what do you see?
John Reid: 18:03Yeah, actually, I'm If you work for a large OEM, and I did for twenty years, you kind of see the productivity we get out of these machines that last one hundred years wider, bigger, faster, high performance. Adding autonomy to those machines is an element of trying to achieve increased productivity with do more with less, like what John Deere says. Now, on the other end of the spectrum are small robots that are going down between the rows. And, you know, at my center, you know, I came into this center a couple years ago, but they've talked about this a lot. And they're showing things like cover cropping.
John Reid: 18:43But, you know, think about the energy and power requirements of agriculture. I think we all go hungry if we just depended on small small robots to to do some of those types of things. So they do unique things, like if you wanted to plant seed in August between corn so that you start a cover crop just before harvest, Definitely doable. Can it connect collect data? Yes.
John Reid: 19:06Drones also can collect collect data.
Todd Gleason: 19:07Was my next question. Yeah. What what what are your thoughts about drone technology?
John Reid: 19:11Some some of it is, one of my my greatest experiences is for a period of time I had to work with advanced marketing people and put technology and marketing together. And every technologist has a hammer. So if you're doing the small robot or you're doing the big robot, you know, you want to do that. And that's what you know. But I think the reality is you have to understand what's the effectiveness of the solution, what does it cost, how much work does it take to get it done.
John Reid: 19:41Think there's just a variety of options that we have today, including remote sensing. I've used remote sensing services that are available today that give me a prediction of yield on a field that are are better than the yield monitor on a combine. So, you know, that's actually easy. You can pay a subscription fee to give them your farm. I mean, you just buy it for your own area of land and you have a layer of information that can help manage how the combine operates more efficiently.
John Reid: 20:12Again, we have to be a little agnostic to the specific version of technology. Now back to this point, I think there are going to be some operations. So Lymphtech is a company that shows a solar powered sprayer. And they've even I talked about this idea of tendering. They even show that they're building a unit that it can park and tender.
John Reid: 20:33And this this just runs really slow because it's solar powered, but it runs all the time. And so the idea of being able to do, you know, perhaps some levels of weed management in certain operations and not have to touch it, if you really don't have to touch it, then that's potentially valuable for the farmer to consider. On On the other hand, if you have to spend a lot of time and resources to get everything set up and use it, and then you're going to stand there in the field and watch it while it's operating, well, that doesn't feel like it improved your productivity very much. It just is kind of a novelty that's pretty cool to see. So we have to kind of get past those early learning phases to where you can trust, have trust in automation, and you don't have to sit there and babysit it.
Todd Gleason: 21:18One of your colleagues, when you were still at Deere, was here a couple of years ago talking about, and he headed up the combine section, was talking about machine learning versus AI. That's in your wheelhouse. What do you think about the difference between the two, if there is one?
John Reid: 21:34Well, learning is a form of AI, but one of the reasons I'm really enjoying being back at university is the power of AI is getting cheaper and it's more capable today and can do some really interesting things. My team is working on things like embodied AI that can go inside the machine and collaborate with the farmer in terms of advice and ask questions, document information, so you don't have to hit the display, but you can just have information automatically ingested. We actually have an AI tool that's a large language model called Crop Wizard, and there'll be a big announcement on this this week. Working with a number of companies where it's kind of an expert agronomist. It's been trained on USDA data agronomy.
John Reid: 22:21You can all go find it at uiuc.chat, and it's one of the programs that's listed there. It's also in the early phases of development, but it's based by factual information that scientists have collected, and it ends up being a quick way of answering questions and getting a perspective. Now, when you get the answer, you still have to use your brain. You have to understand like everything else, is it telling me something that's factually true? Look at the references.
John Reid: 22:49But it's kind of a nice aid that could be integrated in the machinery as well as used in the farm, site.
Todd Gleason: 22:54John Reed is with the University of Illinois. You can find the chatbot at uiuc.chat/cropwizard. You've been listening to the closing market report from Illinois Public Media. I'm Extensions, Todd Gleason.