Latest / Talking Dairy / Precision Dairy Farming Series: What can dairy learn from horticulture robotics? | Ep. 7

Transcript

Precision Dairy Farming Series: What can dairy learn from horticulture robotics? | Ep. 7 · Talking Dairy

• 0:01
  Speaker 1: Welcome to Talking Dairy.
• 0:03
  Speaker 1: I'm your host Jac McGowan from DairyNZ, and this episode is part of a special series recorded at the 2025 Precision Dairy Farming Conference Ōtautahi Christchurch.
• 0:14
  Speaker 1: In this episode, I'm with Professor Mike Duke from the University of Waikato.
• 0:19
  Speaker 1: He's one of our keynote speakers.
• 0:21
  Speaker 1: Mike's team at the University of Waikato work mostly in horticulture, developing things like robots that harvest asparagus, prune grapevines, and survey kiwifruit orchards.
• 0:32
  Speaker 1: He has come to do a talk about how horticultural robotics might be able to transfer to dairy.
• 0:39
  Speaker 1: Kia ora Mike, thank you so much for joining us.
• 0:42
  Speaker 1: Now, as I said, you work mostly in horticulture.
• 0:45
  Speaker 1: How did you end up as keynote speaker at a dairy conference?
• 0:49
  Speaker: That is a very good question.
• 0:51
  Speaker: I asked that myself because uh when they phoned me up, they went, Oh, would you come and give a keynote?
• 0:55
  Speaker: Because I do keynotes in horticultural robotics mainly.
• 0:58
  Speaker: I said, It's dairy.
• 0:59
  Speaker: I said, Why you don't want to?
• 1:00
  Speaker: What do you want me there for?
• 1:02
  Speaker: And anyway, we had a chat.
• 1:03
  Speaker: We talked about what we do in horticulture and obviously what's happening in dairy, even though it's not my area.
• 1:09
  Speaker: And we said, actually, yeah, it could work.
• 1:12
  Speaker: So we decided that we would go ahead with it.
• 1:15
  Speaker: I feel I'm going to the lion's den here, but you know, what's the worst they can do to me?
• 1:20
  Speaker 1: Fair, fair.
• 1:21
  Speaker 1: We're friendly people.
• 1:22
  Speaker 1: Now, forgive me for saying this, but robotics has been a thing in dairy for more than 20 years.
• 1:29
  Speaker 1: I worked on a robotic dairy farm when I was back as a technician a long time ago.
• 1:35
  Speaker 1: Has horticulture been slower to find applications for robotics?
• 1:39
  Speaker 1: Has dairy led the way?
• 1:42
  Speaker: That is a good question.
• 1:43
  Speaker: You can answer that two ways.
• 1:44
  Speaker: If you looked in the horticultural sector and you went to the apple orchards, put them to one side, when all the apples are harvested, if you went to the pack house, you are looking at super high tech, fully automated, AI, masses of apples being processed, images taken of every apple, spread out, blemishes detected, it's graded from one down to a juicer, and each one's flipped off at a different angle.
• 2:10
  Speaker: They've even got robots that will then pack them and turn them around to put them in the best show for the apple.
• 2:15
  Speaker: So that is super high tech.
• 2:17
  Speaker: So the horticultural sector is high tech, but it's high tech when it's not in the orchard.
• 2:23
  Speaker 1: Right.
• 2:23
  Speaker 1: So more in the controlled environments.
• 2:27
  Speaker: Yeah, and I think with dairy, for me, the impressive thing about dairy is the voluntary milking systems, the robots.
• 2:32
  Speaker: You know, so I've been looking at those going, well, you know, that's impressive technology.
• 2:36
  Speaker: It's not just the tech that they've developed, but it's also the add-on, you know, the feed control and that whole monitoring of the animals.
• 2:43
  Speaker: So from that perspective, we're not saying, you know, I'd look at that as on farm, is very impressive in the dairy sector, which is one of the reasons why we didn't originally go into that area.
• 2:53
  Speaker: Because we thought you guys, you're well ahead of it.
• 2:55
  Speaker: And the other thing, which will lead on to why I think we can add some value, potentially, was people did come to us, and we have done a range of projects with dairy.
• 3:05
  Speaker: A lot of it was things like we've even got some projects going at the moment on human assist.
• 3:09
  Speaker 1: What's human assist?
• 3:10
  Speaker: Human assist means where some people are doing tasks, and when you look at the task, it's really onerous and difficult.
• 3:18
  Speaker: And if you just looked at it and said, okay, work with them and say, What's the problem?
• 3:22
  Speaker: They go, Oh, this isn't it.
• 3:23
  Speaker: And then you work and come up with some solutions and you brainstorm it, you build some prototypes, you try it out with the end user, and they go, you know what, this is better, or if you did this, it'd be better.
• 3:32
  Speaker: And then you can end up with something that would actually help them.
• 3:35
  Speaker: So one of the ones we're actually working on at the moment is something really that came to us from a vet, and it was actually ultrasonic testing for pregnancy.
• 3:43
  Speaker 1: Right.
• 3:43
  Speaker: And when they explained what they were doing, and they said, This is it, and we went, Wow.
• 3:47
  Speaker: And again, can you make it better?
• 3:49
  Speaker: So we worked with them and we've got a prototype, which we think is better.
• 3:52
  Speaker 1: How's it better?
• 3:54
  Speaker: It's basically taking into account what they were doing, which was pretty primitive, with just adding a bit more tech and a bit more design thought into what they've done and a few little adaptations.
• 4:05
  Speaker: And I'm not gonna say too much actually, because there's actually you don't want to give it away.
• 4:08
  Speaker: Well, the students who are working on this own the intellectual property.
• 4:11
  Speaker: These are very, very smart students.
• 4:13
  Speaker: These are A-plus students.
• 4:15
  Speaker: They own it and they are looking to develop it into a company.
• 4:18
  Speaker: So I'm gonna say they've come up with something very cool.
• 4:20
  Speaker: You may see in the future.
• 4:21
  Speaker: And we're gonna be helping them develop that into a product.
• 4:25
  Speaker 1: I think when you said human assist, I imagined the are they exoskeletons?
• 4:29
  Speaker 1: Brian Dalarou, who works at DairyNZ, I think he had one a few years ago, was trialing it out to help with milking and take some of the weight off.
• 4:37
  Speaker 1: But that's not what you're talking about, isn't it?
• 4:39
  Speaker: Actually, it is that as well.
• 4:40
  Speaker: So we have one at the moment, we're working in I've got to be careful what I say again, another one, which can be is related somewhat in the dairy uh sheds, is that we're working on a new area, which I can say is called compliant mechanisms, which is a way of having sort of complex motion, which is expensive to do, in something much simpler.
• 5:00
  Speaker: And it's called compliant mechanisms.
• 5:01
  Speaker: We have expertise in that area.
• 5:03
  Speaker: And we've had students working on it, and we're looking at things like if you are bending down, these things will naturally take the load off your leg.
• 5:10
  Speaker: No, like you're really straining, they'll take the load off it like a spring and will return it.
• 5:14
  Speaker: But it's done in quite a clever way using these compliant mechanisms, and there's loads of those, like even in the robotic area, these compliant mechanisms, we're looking to replace complicated, expensive grippers with much cheaper grippers.
• 5:26
  Speaker: And we've got loads of prototypes, okay, which we're working on, and we love them, and we think cost is a big issue with automation.
• 5:33
  Speaker: If you can make one part of it that's a critical part, much cheaper.
• 5:37
  Speaker: And also so that if it wears out over, well, we've been testing these things, there've been millions of cycles, but you know, ultimately, you've then got to replace them.
• 5:46
  Speaker: It's a cheap replace.
• 5:47
  Speaker: Yeah, you're not going, oh, the $5,000 bill for a replacement, which it would be in effect.
• 5:52
  Speaker: So you literally be, oh, just a hundred bucks or two hundred dollars or something.
• 5:55
  Speaker: Nice.
• 5:56
  Speaker: So that is in the human assist and the sort of clever design area.
• 6:02
  Speaker 1: So you haven't yet done your talk today.
• 6:04
  Speaker 1: What are you going to tell them?
• 6:05
  Speaker 1: What are you going to show them?
• 6:06
  Speaker: So I'm going to show them the developments in horticulture, which was definitely behind dairy.
• 6:12
  Speaker: We've had some technology changes that have enabled us to do things that you couldn't do several years ago.
• 6:18
  Speaker: In particular, if I'm talking about AI, but there's two things in AI area.
• 6:23
  Speaker: One is called machine learning, and the other one's a large language models.
• 6:26
  Speaker: The work that's been done in horticulture so far has been really focused on machine learning, which is using images.
• 6:33
  Speaker: And basically, if you went into an orchard, you tried to do a task, you A had to find something and identify it, you then had to go to the location and do something.
• 6:41
  Speaker: That was really difficult seven or eight years ago.
• 6:44
  Speaker: With machine learning and something called convolutional neural networks, it meant that you could train a neural network to identify, for example, kiwi fruit.
• 6:52
  Speaker: And once you've got that, you could look up and take an image and say it would find the kiwi fruit by using stereo vision, which is just like the way we have our depth, you would then get the location.
• 7:03
  Speaker: So once you've identified something and you've located it in an unstructured environment, you can send a robot arm to go and do something to it.
• 7:10
  Speaker: That was the game changer.
• 7:12
  Speaker: And globally, if you looked at research in this area, it's gone crazy.
• 7:17
  Speaker: Everyone on their dog is doing it now.
• 7:19
  Speaker: However, this is where the Kiwis are great.
• 7:22
  Speaker: Kiwis just don't go and do this.
• 7:24
  Speaker: They set a target and they say, Well, what do you need to do it in?
• 7:28
  Speaker: What speed do you need to do it in?
• 7:29
  Speaker: Right.
• 7:29
  Speaker: And they say, Well, harvest the kiwi fruit, it's got to be one fruit a second.
• 7:33
  Speaker: Like this.
• 7:35
  Speaker: And other researchers, they they're just delighted when they manage to harvest one or something.
• 7:39
  Speaker: Yeah, they all give themselves a pat on the back, it's taking 35 seconds.
• 7:43
  Speaker 1: It's not violent.
• 7:44
  Speaker 1: That's not replacing anybody.
• 7:46
  Speaker: No, it's not.
• 7:46
  Speaker: So we we are literally starting with what is the commercial requirement and then driving us to try and achieve it.
• 7:54
  Speaker: And that's a just a different attitude that happens in other research organizations.
• 7:58
  Speaker: They end up trying to focus on some little bit of tech.
• 8:00
  Speaker: We actually end up on trying to focus on solving the problem.
• 8:03
  Speaker: And that way we don't care what we use as long as we can solve it.
• 8:07
  Speaker: I'll talk about the partners, it's not just us, but there's so many universities involved that we work with.
• 8:12
  Speaker: The old what we're called CRIs, like you know, in horticulture, plant and food research, the funding bodies like MB, Callahan, all the companies that do manufacturing, a big shout out in particular for Robotics Plus, who've done so well that they've now been brought out by um Yamaha Agriculture, but it's all still manufacturing their products in New Zealand.
• 8:30
  Speaker 1: Wow.
• 8:31
  Speaker: Of course, the end users, the growers who we have to work with, now that what we call co-design.
• 8:36
  Speaker: Everyone is dragged out to the orchards, the vineyards, whatever it is, get your hands dirty.
• 8:43
  Speaker: You're told what you do by the experts.
• 8:45
  Speaker: We practice it and go, what?
• 8:47
  Speaker: Take the video footage, discuss it, and we discuss also with the people who actually do the job and figure out what are the problems and then look at solving them.
• 8:55
  Speaker: And it completely changes the way you think about something.
• 8:58
  Speaker 1: And then you talked about prototypes.
• 9:00
  Speaker 1: So I presume you're going back with your prototypes.
• 9:04
  Speaker: We have loads of prototypes.
• 9:05
  Speaker: Because you know, I think we have our what was the toughest one to do?
• 9:08
  Speaker: And just on the kiwifruit front, when we got started doing the robotic harvesting, it was great, right?
• 9:14
  Speaker: We're really happy.
• 9:15
  Speaker: But we noticed that the cool end effectors that we'd made.
• 9:17
  Speaker: The what?
• 9:18
  Speaker: Sorry, end effectors, the things that grab the kiwi fruit and do the thing.
• 9:21
  Speaker: We came up with one version.
• 9:23
  Speaker: We thought in our heads it was really good.
• 9:25
  Speaker: And when it went up, as it went up to get the kiwi fruit, it would actually bump off another kiwi fruit.
• 9:28
  Speaker: That's not good.
• 9:29
  Speaker: No.
• 9:30
  Speaker: Because once it's on the ground, it's had it.
• 9:31
  Speaker: So we thought, oh, we thought we'd be really clever there.
• 9:34
  Speaker: And so we then went, okay, how do we fix that?
• 9:36
  Speaker: Try and fix that, try and fix that, try and fix that.
• 9:38
  Speaker: And finally, it took so many iterations to end up with what we call the Pringle, and it's great.
• 9:44
  Speaker: This is it.
• 9:45
  Speaker: We go, that's the solution.
• 9:46
  Speaker: But anyone who thinks, oh, you just go and get a gripper and you solve it, you're dreaming.
• 9:50
  Speaker: Yeah.
• 9:51
  Speaker: Every problem, it's hard to solve it initially and then to refine it to the point where it's useful.
• 9:56
  Speaker: Because we would actually get all the stats on how many each gripper would knock off a kiwi fruit or whatever, and its harvesting speed and gripper force and everything.
• 10:05
  Speaker: And finally, we found, you know, after a lot of developer work and brain learning, we came up with uh a good solution.
• 10:12
  Speaker: So the idea that you could just go into any environment and get some sort of solution like that is dreamy, which is why it's very impressive to see like the robotic milking systems.
• 10:22
  Speaker: They've solved many, many problems to get that to that level.
• 10:27
  Speaker 1: It sounds like you're talking to the dairy sector, not just farmers, but also, you know, the researchers, the developers.
• 10:35
  Speaker 1: What do you want us to learn from your group, from the work you do, from horticulture?
• 10:40
  Speaker: One of the things that we learnt was that when we started, we would have people coming to us, growers and the like, with let's say a standard orchard with 3D trees.
• 10:52
  Speaker: And they go, Oh, we want to automate this.
• 10:54
  Speaker: And we would say, Sorry, we're not doing that.
• 10:57
  Speaker: And the reason was that the future is in linear two-dimensional structures for trees.
• 11:03
  Speaker: We're growing apples.
• 11:05
  Speaker 1: What does that mean?
• 11:06
  Speaker: So instead of having round trees, like you do, the lovely little vision we have of it, they're like giant uh vines.
• 11:11
  Speaker: It's like a vineyard, but on steroids, you know, like they're four meters high.
• 11:15
  Speaker: And and all the apples are are lined up along the wires.
• 11:21
  Speaker: So they're all nicely lined.
• 11:23
  Speaker: Now that wasn't done for robotics.
• 11:27
  Speaker 1: No, I can see how that would make a lot of tasks in the orchard a lot easier.
• 11:32
  Speaker: Complete game changer.
• 11:33
  Speaker: Yeah.
• 11:34
  Speaker: It wasn't done for us, it was actually done for solar gain.
• 11:37
  Speaker: If you lined them up row on row and pointed them towards the sun, you've got giant apple solar panels.
• 11:42
  Speaker: And the yield went up dramatically.
• 11:44
  Speaker: And I'm sure the people who are experts in FOPs saying, Mike said that all wrong.
• 11:48
  Speaker: You know, we didn't anyway.
• 11:50
  Speaker: That's the take on is to get a massive increase in yield.
• 11:53
  Speaker: Loads and loads of orchards are moving to that type of structure.
• 11:57
  Speaker: But it means for us, you can automate, you've got a much more controlled structure to do it.
• 12:03
  Speaker: I think the same thing's happened in dairy.
• 12:05
  Speaker 1: Okay.
• 12:06
  Speaker 1: Tell me more.
• 12:06
  Speaker: And I'll go, no, I'm gonna say this.
• 12:08
  Speaker: The fenceless farms.
• 12:09
  Speaker: Right.
• 12:09
  Speaker: When we were looking at doing some work on dairy, for example, one of the projects that we had was killing weeds like thistles and things.
• 12:17
  Speaker: You know, it came from one of our dairy experts at the university who works on soil, but he said, Hey, they come to me with this problem.
• 12:24
  Speaker: We had a look and said, You're not gonna be able to get around.
• 12:26
  Speaker: There's fences everywhere, there's gates, there's you know, it's a nightmare.
• 12:30
  Speaker: Oh, can you do it with drones?
• 12:31
  Speaker: I'm not a drone fan, sorry.
• 12:33
  Speaker: So, you know, someone actually got money to do this with laser beams.
• 12:36
  Speaker: Was it dairy and Z?
• 12:37
  Speaker: I don't know.
• 12:38
  Speaker: I don't know.
• 12:38
  Speaker: Anyway, some time ago, we all went, This is crazy.
• 12:41
  Speaker: So we never did it.
• 12:42
  Speaker: You know, you look at the VMS and you look at the way you control entry to a VMS with voluntary milking system, so that's robotics.
• 12:50
  Speaker: Yep, right.
• 12:51
  Speaker: Love it, very, very cool.
• 12:53
  Speaker: People love it, people hate it.
• 12:55
  Speaker: I'll come to that in a minute as well.
• 12:56
  Speaker: And then you've got things like Holtero are one of the sponsors, so the collar system.
• 13:01
  Speaker: I know people call it the collar system, there's more and more being developed.
• 13:03
  Speaker: Awesome, because you can suddenly control it, you can take the fences away.
• 13:07
  Speaker: That means that the autonomous vehicles can go into the farm.
• 13:12
  Speaker 1: Yeah.
• 13:13
  Speaker: And that means they can go and perform tasks.
• 13:15
  Speaker 1: Yeah.
• 13:16
  Speaker: You're on the bikes or they're on the quad bikes, they're off doing stuff because that's how they have to go and collect information or collect this or go and expect that or go and do a task.
• 13:23
  Speaker: Suddenly you've opened it up to a completely new world of potential automation, which is what is happening in the horticultural sector.
• 13:30
  Speaker: Right.
• 13:31
  Speaker: Yeah, because there are no fences.
• 13:32
  Speaker 1: Okay.
• 13:33
  Speaker: If you go to a big orchard, there's massive open areas for vineyards and orchards, dairy.
• 13:38
  Speaker: You know, you can go all the way around these things without having to open a fence.
• 13:41
  Speaker: Then you move to another one.
• 13:43
  Speaker: But you know, while you're doing a performing the tasks, you've got a lot of space that you can move around without having to do all this gate opening business.
• 13:51
  Speaker: Yeah.
• 13:52
  Speaker: So I think that's one of the game changers.
• 13:54
  Speaker 2: Yeah.
• 13:55
  Speaker: Potentially.
• 13:55
  Speaker: The other one, which is a really new thing that we're working on.
• 13:59
  Speaker: So if you looked at the horticultural sector, most of it, you could call it almost muscle memory repetition, which is like if you're harvesting something.
• 14:06
  Speaker: You'll go up and you can almost go to sleep while you're doing it.
• 14:09
  Speaker: Very easy for a robot if it's using, for example, the machine learning that we talked about to identify and harvest.
• 14:16
  Speaker: But when you have a task that needs more thought, so a really good example of this, there are many, many examples, but the one that is a really good one to look at is pruning a vine in a vineyard.
• 14:28
  Speaker: There are literally tens and tens of millions of vines in New Zealand that have to be pruned every year.
• 14:34
  Speaker: There are five or six critical cuts.
• 14:37
  Speaker: If you make the right cut, you get good yield, get good growth, and you get long-term well-being of the park.
• 14:44
  Speaker: If you make the wrong cuts, you can get the opposite.
• 14:47
  Speaker: So, how do they do it?
• 14:49
  Speaker: You have a few experts.
• 14:50
  Speaker: I mean, these are real experts.
• 14:52
  Speaker: They won't argue amongst themselves, as experts would, about what's the right way to prune.
• 14:56
  Speaker: And we were there once where we had two experts arguing about the right way to prune with all the researchers there.
• 15:01
  Speaker: We're trying to work out what they want to do.
• 15:03
  Speaker: And there were some workers who were semi-skilled who had been sort of trained up, and they were looking at it and going, Well, they should have done it like this, shouldn't they?
• 15:09
  Speaker: What?
• 15:09
  Speaker: And they went, Oh, that's not right.
• 15:11
  Speaker: And then he went to the next one and went, That's not right.
• 15:13
  Speaker: Before we knew it, we're standing around.
• 15:15
  Speaker: This guy is running down the vine to stop the workers from doing any more pruning.
• 15:21
  Speaker: Wow.
• 15:21
  Speaker: Because they were doing it incorrectly.
• 15:22
  Speaker: This is a problem that you've now trying to get someone to train someone to do this.
• 15:27
  Speaker: But if you can put it on a robot, that robot will do exactly what you want.
• 15:32
  Speaker 2: Yeah.
• 15:32
  Speaker: Now you can do it with algorithms.
• 15:34
  Speaker: So you can say, what are the algorithms to prune the vine?
• 15:37
  Speaker: And that has been done, but not really done properly, in our opinion.
• 15:42
  Speaker: Actually, not at all properly.
• 15:44
  Speaker: It needs a lot of work.
• 15:45
  Speaker: And there's another way of doing it, which is called narrow physical AI.
• 15:50
  Speaker: So, what it is, we humans are unbelievably good at doing tasks.
• 15:54
  Speaker: We can pick something up.
• 15:55
  Speaker: You know, with with the thing that you just take for granted, you pick up a glass of water or you do this or whatever.
• 16:01
  Speaker: We're really, really good at it.
• 16:03
  Speaker: We just take it for granted.
• 16:04
  Speaker: If you try and get a robot to do that with a decision, it's really, really tough.
• 16:09
  Speaker: So to try and solve any particular problem isn't really broad, it's actually really narrow.
• 16:15
  Speaker: But you can break down a lot of agricultural tasks, whether it's in horticulture or dairy, into these narrow AI tasks that you have to do.
• 16:23
  Speaker: So that's the future is narrow physical AI.
• 16:27
  Speaker: And for example, what we're trialling at the moment, and we've already got it to the point where we have a robot which can drive down to a vine row with an arch and go over the vine.
• 16:39
  Speaker: It can stop at the vine, it can then scan the vine.
• 16:42
  Speaker: Our partners in Auckland Uni, we work really close with Auckland, have been scanning the vines and creating a model of the vine, a virtual model of the vine, to the point where you could make decisions on it and you'd know each location in 3D space.
• 16:58
  Speaker: So the robot can go and do the cuts.
• 17:00
  Speaker: You've got to figure out where to do the cuts.
• 17:02
  Speaker: If you take an expert who can then talk through what they would do, as they talk through, you can convert the spoken language into text.
• 17:12
  Speaker: We've all familiar with large language models, Copilot, ChatGPT or Gemini or whatever.
• 17:18
  Speaker: But they're too large.
• 17:20
  Speaker: They're very broad.
• 17:21
  Speaker: And there's a thing called RAG, R-A-G, which is where you can take all this other information that you've got, like the Vine data and the requirements, and combine it with a large language model to then generate the computer algorithms that the robot will then prune to.
• 17:36
  Speaker: So you've now combined what the expert wants with what the robot's going to do.
• 17:41
  Speaker: This is frontier stuff.
• 17:42
  Speaker: We're working on drawing the block diagram for this for the talk later as a sort of thing.
• 17:46
  Speaker: I'm going, yeah, but should that be and it is that we we got a very smart person working on this at the moment.
• 17:52
  Speaker 2: Yeah.
• 17:52
  Speaker: We think within a year we'll have the first proper prototype where we'll be at a trial.
• 17:58
  Speaker: This it will be rough as gus.
• 18:00
  Speaker 1: Very cutting-inch.
• 18:03
  Speaker: It's crazy.
• 18:04
  Speaker: And we're going, oh, but suddenly imagine on a dairy farm that you've now got lots of narrow AI.
• 18:11
  Speaker: And when it's a task that needs to be done, it's the farmer who knows what they want to do.
• 18:16
  Speaker: You may have some default values in it.
• 18:18
  Speaker: This is a default, oh yes, all right.
• 18:20
  Speaker: No, no, I want it to do this.
• 18:21
  Speaker: You talk it through, it redevelops the algorithm.
• 18:23
  Speaker: This is a bit Star Trek-y, but you're definitely going to see this work in the future.
• 18:28
  Speaker 1: Okay, so you're talking about some potential stuff that's some way off.
• 18:33
  Speaker 1: Is there anything that dairy farmers who have come to this conference or who are listening could pick up from horticulture right now in terms of learning or technology?
• 18:43
  Speaker: I don't really want to go down the app route so much because a colleague I work with, he's a real industry guy, Nick Pickering, and he did a lot of work with Zespry on what we call system of systems.
• 18:54
  Speaker: It's the fact that everybody is collecting data and providing with an app and providing with this and providing with this.
• 19:00
  Speaker: It's all over the place.
• 19:01
  Speaker: Where is the integrated system?
• 19:03
  Speaker: Yeah.
• 19:03
  Speaker: So uh Nick's big thing, and he insisted if you're going down there, you need to talk about that.
• 19:07
  Speaker: I said, All right, Nick, I'll mention it.
• 19:09
  Speaker: And he goes, System of systems for dairy, where is it?
• 19:13
  Speaker 2: Yeah.
• 19:13
  Speaker: The thing that's close is this interoperability of data, edge computing.
• 19:19
  Speaker 1: What's each edge computing?
• 19:20
  Speaker: Edge computing means when we went to collect data in an orchard and we were mapping out all the kiwi fruit, when we did a certain number of orchards, we hit over six terabytes of data.
• 19:30
  Speaker: Cool.
• 19:30
  Speaker: Crazy numbers.
• 19:31
  Speaker: So we had a stack of computers which we had storing everything.
• 19:34
  Speaker: We had to make our own mini supercomputer.
• 19:36
  Speaker: And we went, this is nuts.
• 19:38
  Speaker: So edge computing is that as the data's coming through, you process it, which is why it's on the edge, you take out the key information that you need, and that's what you can do.
• 19:48
  Speaker 1: Throw the rest away.
• 19:49
  Speaker: Throw the rest away.
• 19:50
  Speaker: Because you don't need all this micro data, you just need the key ingredients.
• 19:53
  Speaker: So if you want to create hotspot maps or deep information where you need to go and do a task, do it that way.
• 19:59
  Speaker: So that means you go from terabytes to megabytes, which means suddenly, you know, you can start to get something useful out of it.
• 20:06
  Speaker 2: Yeah.
• 20:07
  Speaker: So that's one thing.
• 20:08
  Speaker: And the other one is about hardware.
• 20:10
  Speaker: If everyone develops all their own hardware, you're going to notice so much high-tech hardware, it's not cost effective.
• 20:16
  Speaker: You look at the mobile phone, the mobile phone's one bit of kit with everything on it.
• 20:20
  Speaker: So data interoperability, a system like, you know, we use Android on your phone or whatever, edge compute to reduce the data that you're storing, and the hardware such that it's modular.
• 20:32
  Speaker: You start to do that, and that is something that can be done now.
• 20:34
  Speaker: Before I came down, spoke to some experts.
• 20:37
  Speaker: And it looks like there's some good opportunities for that data interoperability going on right now.
• 20:42
  Speaker: I think it originated in New Zealand, ended up going overseas, and I think it's on its way back again.
• 20:47
  Speaker 1: Oh yeah.
• 20:47
  Speaker: I don't know, because it's not, you know.
• 20:49
  Speaker 1: Yeah.
• 20:49
  Speaker: I've just what someone who works in this field in, I think they're here at the conference, he's told us about.
• 20:54
  Speaker: So that's something that could be done.
• 20:56
  Speaker: The other one is the vehicles.
• 20:58
  Speaker: I think you would be able to get vehicles to go out and do some tiles like surveying and giving you information without you having to go out and do it yourself.
• 21:05
  Speaker: That's pretty close, as long as you get rid of the fences.
• 21:08
  Speaker 1: A lot there, Mike.
• 21:10
  Speaker 1: A lot of big ideas and exciting ideas.
• 21:13
  Speaker 1: I'm sure the farmers that have come along to this conference feel very inspired.
• 21:16
  Speaker 1: I feel very inspired and curious about what's been going on in horticulture.
• 21:20
  Speaker 1: Thank you very much for that glimpse into how robotics might be able to be applied even more in dairy.
• 21:28
  Speaker 1: Yeah, enjoy the rest of the conference and I hope your talk goes well and you don't get torn to shreds.
• 21:33
  Speaker 1: Thanks.
• 21:34
  Speaker 1: I hope not too.
• 21:35
  Speaker 1: Thank you very much, Mā te wā.
• 21:38
  Speaker 1: If you'd like to get connected with DairyNZ's latest advice, research, tools, and resources, whether it's reading, scrolling, listening, or in person, you can visit dairynz.co.nz/get-connected, and don't forget to hit follow.
• 21:59
  Speaker 1: To keep up to date with our latest episodes.
• 22:02
  Speaker 1: As always, if you have any feedback on this podcast or have some ideas for future topics or guests, please email us at talkingdairy@ dairynz.co.nz.
• 22:11
  Speaker 1: Thanks for listening, and we'll catch you next time on Talking Dairy.