news    views    talk    learn    |    about    contribute     republish     crowdfunding     archives     events
podcast archive     

Robots Podcast #186: Towards Automating Fieldwork, with Hans-Peter Grothaus

CLAAS         


interview by
July 10, 2015

CLAAS_Field_Automation

Transcript included.

In this episode, Per Sjöborg talks to Hans-Peter Grothaus, from CLAASabout automation in agriculture.

There is a small window of time when crops are ready to harvest. If, during this window, not all of the crops are harvested, the farmer can experience large financial losses. This means that systems that automate fieldwork must be reliable. And reliability, or robustness, is a major challenge because field conditions change during the day (it could start raining) and it is tough to know soil conditions in advance. In this interview, these challenges are discussed, as well as how field data can help farmers make decisions and the market for agricultural robotics.

 

Hans-Peter Grothaus

Hans-Peter_Grothaus

Dr. Hans-Peter Grothaus studied Agriculture Sciences at the Universität Göttingen in Germany, where he wrote his dissertation. Since 2008, Grothaus has been the head of development for system-based services at CLAAS in Harsewinkel, Germany.

 

 

Links:

Transcript

Per Sjöborg:  Welcome to the podcast. I’m here with Hans-Peter Grothaus from CLAAS, and we’re going to talk about automation and fleet integration within the agricultural community, a reasonably well established field in advanced automation. But it’s still growing and adding features as we go along. Could you tell our listeners, who may never have been on a farm, what kind of equipment you’re working with and how it is used by your customers?

Hans-Peter Grothaus:  What we’re using is harvesting machines and also machines for tractors, for overloading the harvested material, for infield processes and also for transportation, to bring the harvested goods to the farm.

Per Sjöborg:  Now we’re talking about corn, potatoes, and what else?

Hans-Peter Grothaus:  So, we work more with grain and grapeseed and all the fruits that are above the earth, nothing under the earth.

Per Sjöborg:  Yeah, because the harvester is in the middle of a network; it can’t do its job on its own; it has to have a fleet of tractors with carrying capacity and then trucks to get to a storage facility, and all of these integrate and work together. Could you tell us a bit more about the processes you use and how these are optimized?

Hans-Peter Grothaus:  So, traditionally we have single harvesting machines, with an overloading facility at the end of the field. But in bigger farms in larger surroundings, we also have multiple harvesting machines working together with overloading facilities, which bring the harvested goods to a truck waiting at the end of the field.

Per Sjöborg: So this means that there are actually two instances of offloading and unloading: from the harvester to the tractor, that is able to operate in the field, and then from the tractor to the truck that is only able  to operate on regular roads.

Hans-Peter Grothaus:  Yes, that’s true, because the trucks are normally not used in the fields, therefore we use tractors and special overloading facilities; overloading wagons which have low tire pressure so that the ground is not damaged by the tires.

Per Sjöborg:  And the vehicles don’t get stuck and so forth.

Hans-Peter Grothaus:  Yes.

Per Sjöborg:  And all this has to be optimized, of course, so that we use as little fuel as possible and so that we can harvest as much as possible in as short a time as possible. I guess that’s very hard to optimize properly.

Hans-Peter Grothaus:  Yes, you have different goals. For example, when you harvest depends on weather conditions and on the available equipment you have, and therefore you have to plan. We are developing infield planning systems. Just imagine you have a navigation system and you can say, “Okay, I want to be ready early with my harvest because bad weather is coming,” or you have very stable weather conditions and you say, “Okay, I have another target, I want to have the highest quality, with low fuel consumption,” or you want to have the same process but not drive over the whole field to avoid soil compaction.

Per Sjöborg:  Yes, because, as you’ve said, you have no roof over your operation and so you really feel the robotics operating out there under rapidly changing conditions. I mean, you can have a prediction of good weather but it can change immediately, or in a very short period of time, and you have to re-optimize everything.

Hans-Peter Grothaus:  Yes, you know that you have to harvest once a year, you know that weeks and years beforehand, but what you don’t know are the actual conditions. And you don’t know exactly what’s grown on a certain plot – that can vary a lot, even within a field. You don’t know the condition of the soil, how well you can drive on it or how weather conditions will develop throughout the day. So, when you start harvesting everything it could be optimal, and after two hours it’s not so optimal, so you have to change the settings of the fleet and the single machines.

Per Sjöborg:  Then, of course, this dynamically redefines itself all the time towards the overall goals of the operation.

Hans-Peter Grothaus:  Yes.

Per Sjöborg:  I understand that you use navigation systems, usually GPS. Do you also use any visual verification of where you are in the field, where other machines are in the field?

Hans-Peter Grothaus:  For steering purposes, we have different solutions on the market. We have camera based solutions and we have GPS based solutions, with correction signals that are accurate to two centimeters. So that works well. We also have laser-based systems that are for steering the machines. All these planning tools are in development. The first prototypes work quite well. The first step is to show the other drivers the actual location of a machine. When you imagine a field of wheat or barley, you can see the machine on the horizon, but when you have a maize field – maize is four meters high – you might only hear the machine, but not know exactly where it is. When you can see that the other machine is at a certain spot and you can see how full the overloading bin already is – this information helps you.

Per Sjöborg:  Especially, as you mentioned, in a maize field, where it might be very scary to see a big combine harvester only meters away!

Hans-Peter Grothaus:  Or you only hear it.

Per Sjöborg:  Yeah, but where is it? Because you’re in this high maize field, in a smaller vehicle.

Hans-Peter Grothaus:  Yes, but that is not the main danger, because you can react. For these inferior processes, you need the right overloading position. With a maize harvest, there is a lot of bulky material which you have to transport. Just for finding the right rendezvous point – where the transportation vehicles should meet, to avoid waiting time and additional driving – it helps a lot and takes the stress out of the process.

Per Sjöborg:  You use a lot of sensors to collect data about what is being harvested in real time. So, you know that, for instance, next year this area of the field can give X amount of return. What kind of sensors are you using to detect the quality of the grain etc.? They have to be very reliable and advanced, if the farmer only has a window of a couple of weeks or a month.

Hans-Peter Grothaus:  We use different cameras. For example, still cameras to detect grain quality, and this was hard to develop because there are difficult surroundings in which we use them and therefore they have to be very robust. They’re very reliable because, if they fail, then it takes a long time to fix. The farmer has to do his harvest right now, not in two weeks time right? Farmers want the machines to work without interruption during the harvesting period.

Per Sjöborg:  It’s a stressful period for the farmer.

Hans-Peter Grothaus:  Yes, so they have to be very reliable during the harvesting period. That could be part of the premium our customers pay for the products, because they’re very reliable. We do everything we can to ensure that the machine does not fail in the harvesting period. So there are lots of sensors and condition monitoring. We can make yield maps, so we can map exactly how much grain or other goods were harvested at a certain plot, and maps where you make overlays over a number of years and you can see how much you have harvested in a certain area. With this information, you can decide how much fertilizer to use on a plot to manage it. So, you see, there is an inter-linkage between the machines, processes over the years and the management systems. You have connected managing systems, with all the data from the whole farm and the machines input data but also use the data from this management system.

Per Sjöborg:  This also goes into the optimization process, of course. Say, if we have dry weather now but we are expecting wet weather, we can optimize to harvest the part of the field where the soil is usually the wettest, because it might be un-drivable if it rains further on in the day or the week. So you can optimize the data for that?

Hans-Peter Grothaus:  Yes, when you know that, you can start harvesting a certain area of a field earlier. You can also use your machine for a longer period, but when you are using a contractor for the first time, where should he get the information from, to know where to start? So, normally he drives around and then, after some time, he knows what to do, but you can save time if you give him additional information.

Per Sjöborg:  And in this reasonably small window of harvesting time, every hour, every half a day counts a lot. It’s very important to do it on time.

Hans-Peter Grothaus:  Yes, that’s true, and systems which help farmers to make better quality decisions are economical because an hour of harvesting time is very expensive, especially if you are not able to use it.

Per Sjöborg:  If weather catches up with you.

Hans-Peter Grothaus:  If the weather catches up and you have no additional capacity, that would be the worst case scenario.

Per Sjöborg:  Harvesting is one of your major businesses but you do other things in agriculture too. We’ve already mentioned pesticides and fertilizing, but there’s also tilling. Can you give us a short overview of your work in non-harvesting related activities?

Hans-Peter Grothaus:  We also have tractors and these act with implements on the farms, so we don’t have to produce sprayers or fertilizers. We produce machines for grass harvesting, for example. There is also a lot of intelligence, such as systems where the tractor is steered by the implement; the implement in these processes is much more intelligent than the tractor.

Per Sjöborg:  Because it knows what it’s doing and what it needs from the tractor to do it the best way.

Hans-Peter Grothaus:  Yes, the implement knows its own process, and the implement has more and more intelligence, and processes and computers. The idea is that the tractor gets an image of the implement and you can steer it, but the intelligence is in the implement.

Per Sjöborg:  Because it’s unique to a particular task. This is very interesting. As I understand it, the process of using automated and assisted driving is more or less standard today, so could you describe how these things have been selling in the farming industry, because I think that many people out there have ideas for robotics and need information on how to sell this to a community. How did you convince farmers that this reasonably advanced level of technology would work, and that it would be beneficial and they could handle it? How did you convince them that this was a good idea?

Hans-Peter Grothaus:  Most professional farmers and most customers can decide very rationally, if they see that a system helps them work for a longer period without exhaustion, and helps them to save money because the steering is more accurate. Just imagine you’re mowing your lawn and you always overlap a little, because you don’t want grass to be left. In the harvesting process, when you have to cut a bar with twelve meters, nobody is able to do it with such great accuracy and at such a distance, so we always overlap, maybe half a meter.

Per Sjöborg:  Which is a lot if you’re doing many fields.

Hans-Peter Grothaus:  Which is a lot if you’re doing kilometers of driving a day.

Per Sjöborg:  So if you can get that down to two decimeters…

Hans-Peter Grothaus:  With ten centimeters, you have gained a lot on fuel consumption. Less time means less soil compaction. So the steering systems pay off in one year now. Other systems? That is when our customers look at all the new features and say, “I don’t know if I need it.” But if they do buy it, they get used to it because the work gets easier; it’s more relaxing, less stressful and the farmer can, for example, do management tasks on the machine. You don’t need to do stupid work, like steering the machine. It does it on its own.

Per Sjöborg:  Yeah, and it does it better, but I guess he is better at the management though.

Hans-Peter Grothaus:  Yes, but he has to supervise the processes, to see if there are obstacles that the machine may not have seen.

Per Sjöborg:  This leads us to the future of these systems. Are we going to see these fleets become fully autonomous? You’ll have someone overseeing a fleet of these vehicles – a few harvesters and the truck distribution network we talked about – to the edge of the field. I don’t think that the actual truck taking it to the storage facility is a problem. Will we see fully automated fields in the near future?

Hans-Peter Grothaus:  I can imagine that we’ll see more “master-slave” based systems in the near future. The labor cost is not the limiting factor at the moment, because the machines are not very cheap but, on the other hand, you have to bring all the machines into the field and you have to get them back. Therefore you have to change the logistics when you have several machines with only one driver. So you have new challenges. There are ideas in academia about swarm robots, which could harvest. But at the moment, especially in the harvesting process, we have lots of very messy goods. We have to transport many tons and that is not suited to tiny robots, so they should be big robots.

Per Sjöborg:  Then they become dangerous robots.

Hans-Peter Grothaus:  They will become more dangerous for man and then you have new challenges.

Per Sjöborg:  As you’ve pointed out, even if they were autonomous in the field, you’d still have to bring them into the field.

Hans-Peter Grothaus:  You have to bring them and maintain them – you have to recharge them when they’re electrically powered. So, if you’re monitoring; if you have a small robot which monitors the health of plants, for example, that could drive around and give us this information.

Per Sjöborg:  “Do I need to add more fertilizer or water or pesticides?” It could do those kind of things.

Hans-Peter Grothaus:  Yes it could also do fertilizer and pesticides. But, as a first step, it would just give information.

Per Sjöborg:  Yes, to determine if we need to.

Hans-Peter Grothaus:  If we need to, and then give this information to the process.

Per Sjöborg:  So, for instance, it could go over a field and detect where there are a lot of weeds, to see if more pesticides are needed or if the grain isn’t developing properly and needs more fertilizer. It could also give you advanced information about where the crop hasn’t grown.

Hans-Peter Grothaus:  Yes, and bear in mind that these vehicles would travel very slowly, and they have a lot of time. They’d make maps of the field. Then, the sprayer or the fertilizer machine, with bags of material or fertilizer, would go in at high speed, and it would have an exact map of where to spray.

Per Sjöborg:  That would take the optimization we talked about in the beginning a step further; it wouldn’t only rely on the data you had last year, it would actually look at the field and base the optimization on how it appears now.

Hans-Peter Grothaus:  There are different systems just to decouple the processes.

Per Sjöborg:  Very interesting. So, where do you see the future? What are you working on in your secret labs right now? Where is the cutting edge of infield agricultural operation?

Hans-Peter Grothaus:  The first step is more autonomy in the infield process, in the logistics. We have installed high capacity in the machines but sometimes, in the logistic processes, there is a reason for the gaps so the farmer does not get the installed capacity. Aside from the infield logistics, you also have outfield logistics. So, where there are no roads, the normal navigation system doesn’t always have the right information for our surroundings.

Per Sjöborg:  No, it’s not on a paved road, you’re out there in the field and the maps aren’t very accurate.

Hans-Peter Grothaus:  Yes, and it doesn’t know where the right spot is and where there are obstacles like electricity wires, telephone cables and trees. This information is not collected today and it could help to make it available, to make plans for the outfield logistics.

Per Sjöborg:  Yes, and entrance and exit points for a field. The system could also tell you that a point is placed incorrectly and we could optimize much better.

Hans-Peter Grothaus:  Yes.

Per Sjöborg:  Then you could take the information you gathered and optimize and even adapt to that. Yeah, very interesting.

Hans-Peter Grothaus:  You can also simulate processes and you can use your simulation to educate the driver. Because there are only six to eight weeks in the harvesting period, and, for the rest of the year, the farmers aren’t harvesting and the drivers are not trained, so they need a couple of days to get into the process again.

Per Sjöborg:  They’re better at the end of the harvest than they are in the beginning. It’s like going on a skiing holiday, when it was a couple of years ago that you last skied, and you’re a much better skier the last day of your holiday than the first. But if we could train someone properly ahead of time, we could optimize this critical window of opportunity we have at harvesting time.

Hans-Peter Grothaus:  We have a certain period where we can do the work but we could use historical data and simulation to train the drivers.

Per Sjöborg:  Perfect. Thank you very much for taking the time to do the interview.

Hans-Peter Grothaus:  Thank you, it was a pleasure for me.

All audio interviews are transcribed and edited for clarity with great care, however, we cannot assume responsibility for their accuracy.


comments powered by Disqus


follow Robots Podcast:

Recent episodes: