Ag Tech Talk Podcast: Using Lasers to Zap Weeds

---

The traditional crop inputs industry faces any number of challenges — regulatory agencies, “woke” consumers, and lawyers looking to make quick money at the expense of chemical companies with deep pockets. Add to that the LaserWeeder from Carbon Robotics. A relatively new offering, the LaserWeeder, appears part advanced agricultural tool and part weapon developed by the empire in Star Wars. In this podcast, we talk with Paul Mikesell, Founder and CEO. Mikesell talks about how and why he and his team developed the Carbon Robotics laser weeder and the future of robots in technology.

Here is the transcript for the podcast:

AgriBusiness Global: Hello, and welcome to Ag Tech Talk where we look at the agricultural advances shaping the future of the crop input industry. I’m your host, Dan Jacobs, Senior Editor with AgriBusiness Global. Today we’re talking with Paul Mikesell, Founder and CEO of Carbon Robotics, makers of a laser-wielding, weed-killing robot.

That sounds very futuristic. Welcome to the podcast. We appreciate your time.

Paul Mikesell: Yeah, you bet. Thanks for having me.

ABG: Well, before we even jump into the LaserWeeder, let’s talk about Carbon Robotics. What’s the significance? Where did it come from?

PM: The honest truth is that we needed a name for the company, and we played with a bunch of different words that we thought sounded cool, and I think carbon robotics just came out of that list. There’s nothing really specific about it other than we thought it sounded cool,

ABG: Fair enough. I just know there’s a lot of talk about carbon sequestration.

PM: There’re certainly positive environmental impacts to the laser leader itself. Some of it having to do with not having to cultivate so much for weed control, some of having to do with the negative environmental impacts of herbicide production primarily on the production side. There’s an incredible amount of greenhouse gases that go into the production of herbicides. But the truth of the matter is, we just chose the name because we like the way that it sounded.

ABG: Can’t argue with you there. And we’ll get into some of that other stuff you just mentioned in a little bit. Technology has been injected into so many parts of our lives – agriculture certainly included you know, in 10 years, or some undetermined time in the future. Are we going to see any people actually working in the field? Or is it going to be all robotics?

PM: Well, that’s a good question. I think that we will probably need people in the fields for a long time to come just because of the amount of day-by-day, or in some cases, hour-by-hour decision making that needs to go on about what’s going to happen in the field.

The crop markets on a daily basis can have dramatic effects of what the plan for the day is going to be. And so those kinds of business decisions will probably be made by people for a long time to come.

But one of the things that those people making those decisions need is more insight, more data, more evidence to make those kinds of decisions. So, I think one of the ways in which technologies generally on the farm, moving forward, will be helpful, not just talking about Carbon Robotics, but generally, is in getting that kind of information to the growers, so they understand what’s happening in their fields. And so that’s one of the, I think, more exciting things about agtech generally is that information about the field in real time is now coming. And that information is now getting to the point where it’s actionable. I think there’s been a lot of businesses that are trying to sort of take pictures and tell farmers how to farm. That turns out not to work very well but giving folks insights is the real path forward here for all of this, all this recent trend in data. So, we’ve been very focused on that as a side benefit of the LaserWeeder.

So, there will probably be people involved for a long time, but I hope that the kind of manual labor that are really expensive, hard to source, doing jobs that just generally are very fun, and have highly variable quality content — I think a lot of that stuff will become automated.

ABG: Fair enough. So, what caught my interest is, weeding obviously is a huge issue. Resistance to some of the chemicals you mentioned earlier. It is a problem in some places. So, you you’ve kind of gone a different route. You went with the lasers. Can you talk a little bit about how that came about?

PM: It’s kind of an interesting story. We decided that we could make an impact in farming because we came from an area where we really specialized in understanding computer vision and the way that machines generally understand the world around them.

That was how we started down this path of figuring out that we wanted to work in agriculture. It’s a pretty great environment for technology innovation, generally, because all of this stuff is happening on private land and controlled environments.

Every action that happens on a farm farmer’s land is, has a dollar value associated with it. That means that if you build a machine with a demonstrated ROI then you could have a very successful company. And what that allows you to do is keep building new innovative stuff. It’s a pretty huge market, so It’s a great place to go into, because you could have a large impact.

That was kind of how we started. Talking to farmers, we realized that this weed control issue was getting pretty bad, because labor is continually getting harder to find.

A lot of this is being done by H2A migrant laborers. Those folks have, because they’re moving around a lot, you’ve got housing issues. There’s a lot to do with the workers’ comp overtime compensation. That’s very regional and kind of patchwork-y, minimum wage law is continuing to change, which is causing a lot of challenges for farmers to even know what their costs are going to be. So, putting people in the field to pick weeds is just not a very efficient use of time.

And then there’s all the issues around what herbicides do to the crops. They set the crops back quite a bit — what the downstream environmental impacts from direct spray and herbicides. The concerns around health effects, both directly to the growers and secondarily to the consumers concerns. As we mentioned before, about the greenhouse gas emissions from, or herbicide production, and then just general efficacy because there’s a rise in herbicide resistance in the weed population.

We saw these challenges going on. We decided we wanted to do something different, and the lasers came about from us, just running the experiments and trying some of everything.

I don’t know if it was insight or just randomly trying a bunch of stuff that got us to lasers. When we realized that was going to work, we started digging in really hard and figuring out how to put a machine together.

ABG: How long was that process to from the time you the concept to where you are now?

PM: we’ve been doing this for 4 years. And there’s been about 3 years in-the-field running machines, and that first year it was just a lot of experimentation trying just kind of everything we could think of.

There have been several generations of the machine. The final production for-sale version is the one that you can see on our website at carbonrobotics.com.

And we have all the social media — Twitter and Instagram, and Youtube, etc. But that machine is a 20-foot-wide machine, so it will run, if you’re doing 80-inch row spacing, they’ll do three rows of 80 inches. It’s configurable for anything down to 60 inches and up to 84-inch rows. But that machine is the one. That’s the production version.

It’s a tractor pull-behind implement. It’s powered from the PTO of the tractor and there have been a couple of generations before that that were self-driving and primarily are currently used as demo units now.

So, that whole process of figuring out how the lasers will work, recognizing it, making it feel ready … that’s been four years total.

ABG: Okay, so can you give us kind of an overview of how it works. How many lasers are we seeing? How do they make sure they hit the weeds and not the crops?

PM: There are 30 150-watt lasers. It’s a pretty large amount of output power. There’s a there’s a computer running that is, looking at cameras that see the whole field horizontally – the whole stripe of the field. The machine is over. which will be typically three or six rows, depending on your row configuration.

It’s finding all the weeds. There’s a targeting system for each of those lasers.

Again, there’s 30 of those. That targeting system, through a series of interesting and clever optics and some servos are able to control where the beam is going to land on the ground. So, the laser beam is going to hit, and there’s a camera that can see what the laser beam path is going to be.

It’s configured in a way, so the laser beam looks like it is hitting the ground from the point of view of this camera in the same area of pixels every time. And so that’s how that camera knows what it’s going to shoot.

From the perspective of that camera and the target location and lasers are all controlled through optics. It’s all a kind of interesting science project to make that whole thing work.

And what happens then is that targeting camera is positively identifying the weeds where the laser is moving to, and kind of fine-tuning the targeting over time, because the machine is also moving right? So, we’ve got to identify the weed and then stay on that target while the machine is moving. And that’s all done through this computer vision system with this targeting system I’ve been talking about. That whole thing happens visually, and it’s based on some pretty interesting work that we’ve done on the computer vision side.

ABG: So, you’ve had to sort of train the computer brain to identify what a weed is.

PM: There’re a couple of those things running. The term is a is a neural net. a deep learning neural net, which just means you could think of it as a giant matrix – they call them neurons in the AI world. Think of it as a giant array of little micro decisions. And over time those micro decisions learn, as a group, how to do things like detect weeds, detect or detect crops, etc. And there’re a couple of different versions of those neural nets running.

There’s the one that finds the original weeds. And then there’s the ones that do the fine-tuning target camera adjustments after the laser has found the target. And that whole thing is running. It’s a process that really would not have been possible seven to 10 years ago we didn’t know how to do this kind of work in the computer vision world. And AI and deep learning neural nets that’s really brought about a big change in the way that computers are able to do things, see the world around them.

The funny thing is agriculture and farms is really the sort of perfect environment for this kind of advanced technological deployment. And I say the funny thing about that because people don’t usually think of farms as being an advanced technology area, but it actually really is. And that’s come about in the last five years. A lot of people don’t really realize it. Some of the most advanced computer vision and robotic machines are out on the farms right now.

ABG: That learning process – is it for farm, or once it’s learned something, is it applicable across the field, so to speak.

PM: Yeah, once it’s learned, it is applicable across all farms. We are gathering more data every day. Every picture we take helps make the system better for everybody. So, it is continually getting better from the information we’re getting from a diverse number of farms.

Once learned, once you deploy a new LaserWeeder odds are we already know the crop and the information about the soil to be able to make good decisions, and then we’ll continue to make it better over time from images from that specific location.

ABG: So, what’s actually happening is the when the laser spots that weed.  Is that a permanent death of that weed. It’s not hitting the roots. Does it still kill the weed entirely.

PM: If you burn out the meristem, the area that’s the undifferentiated meristematic growth cells of the plants – if you hit that with a laser or some other high density, energy source, you will explode the cell walls of the plant itself and prevent it from growing. That’s how the system works, and we have been like I mentioned doing this for several seasons now.

So, the efficacy has been proven time and time again. What’s happening is the plan is not able to grow anymore and will die. That organic material then goes back into the soil and can be used and consumed by the other by the other crops.

And so it means that we’re not relying on chemicals, which have their own run-off issues. We’re just deploying heat energy directly into the plant to kill it. And it does kill him.

ABG: How efficient is the weeder?

PM: I would say, we’re effective in in normal operation at about 88% of the weeds. That’s our benchmark. There are variations in weed density across the width of the machine. A lot of it is: How quickly can you get as you’re cruising through that field?

ABG: Okay, if I remember reading on your website, I think it said about covers about two acres at about, roughly one mile per hour. It’s not exactly setting any speed records there. Is that something that’s going to improve over the next few iterations of the machine?

PM: We are continually working on performance. There’s a certain amount of energy that needs to be pumped into a plant to kill it. And 30 lasers sounds like a lot of output power. But there’s also a lot of weeds, so we’re continually working on this. I wouldn’t be looking for a 2X performance improvement in the next six months or so, but over time we will continue to chip away at that speed.

ABG: Given the current parameters, is there a certain size farm that this is a best use. If you have a 10,000-acre farm, it’s going take a heck of a long time to get through all that.

PM: It typically anything that’s 800 acres or more has been good for us.

Last time I looked at our average farm size it was, I think, 4,000 acres. (For farms with) tens of thousands of acres, no problem. It just means you need more than one laser.

ABG: All right. I think I read the ROI is about one to three years, obviously depending on the size of the farm and the machine ̶ all that kind of stuff.

PM: Typically, one and a half years to 3 years is the ROI that we really look for and focus on. And that means the amount of savings per year, if you divide that savings amount into the cost of the LaserWeeder, that means that within 3 years the LaserWeeder pays for itself. So, that’s how we that’s how we look at it is that.

We need to make sure that growers will get a payback period. That’s 3 years, and that they get a positive ROI and every acre and then it’s just a question of how quickly we can build them.

ABG: Okay, I think I remember seeing somewhere on your website the ones for this past year 2022 we were sold out. I don’t know how many you built, but you know selling out is probably a good thing.

PM: We have (sold out). We are continually trying to bring on more manufacturing capacity. We have machine slots available in the second half of 2023. Right now. The first half of 2023 is sold out. Manufacturing is still difficult right now. Supply chain is still messed up. It’s hard to get parts we use.

Two or 3 years ago I used to be able to get quantities of 10,000 the next day, no problem. Now it is like 18 months. So it’s still pretty bad and there’s a lot of work to do, and exposes how (challenging it is to obtain) base level technological components. I’m talking about integrated circuits and microcontrollers, etc. Everything that goes on us on a circuit board. Those components come from overseas probably 100%

We don’t do much manufacturing electronics manufacturing in this country anymore. That means that we’re exposed, and there’s a lot of shock still working its way through the system. It’s still incredibly hard to get parts, and that’s part of the reason why our order book is so far up.

ABG: Yeah, that’s certainly an issue we’ve talked about over the years. Does this (device need to be used) early season when the crops are still low. Corn grows several feet tall later in the season.  Can the weeder still be used then?

PM: Earlier is better. Smaller weeds take less time to shoot and kill, so, getting in quickly is good. It depends on the crop. Onions, which don’t shade out, so they continually have new flushes of weeds.

We’ve been in onions up until a week or two before harvest. (The machine can handle) a crop that’s growth cycle keeps it below three feet. (After that), the canopy of the crop will prevent sunlight from getting down below and weeds will basically stop growing at that point.

For those (taller) crops we’ll do a single pass, or maybe two. It’s crop dependent. We’ve been primarily focused on vegetables and specialty crops. So, corn has not been a big focus of ours for economic reasons, anyway.

ABG: The growers have been using this LaserWeeder. What are you hearing from them? Are they happy? Are they expressing any concerns?

PM: Most of our challenges are about getting machines out into the field building them fast enough. Growers have been happy with it. The cost savings are great. The technology is interesting. We’re allowing growers to do things like plant more densely because the LaserWeeder’s accuracy, and how close it can get to crops and still kill weeds (allows for) more stand count per acre. So, there’s some positive benefits beyond just killing weeds that come from the fact that we don’t have any crop inputs so you’re not putting a pesticide on the ground. You don’t have to till, so you’re not spinning up that topsoil. You’re not getting micro-bacteria onto the leaves of your leafy greens. There are a lot of secondary benefits that we didn’t even realize when we were building the initial product that have that we learned about. That’s been exciting for us.

What’s the old line? If you’re not moving forward, you’re falling behind. What’s next for the LaserWeeder? How will it be different a couple of years from now.

We’ve been focusing on some other things we could be doing. The software we are (developing) now is for thinning. So you can use LaserWeeder for thinning.

When you don’t get a 100% germination in your seed, so many put more seeds down to get good stand count throughout the field. But if you do that, if you overseed what it means is, you’ll have some crowding issues and areas where you do get good germination across a higher distribution of the seeds.

And so there’s a process called thinning which is, get rid of some of those extra plants that will cause crowding issues. And so, LaserWeeder, can also do that. So that’s what we’re working on, and then just sort of generally other technologies on the farm and areas that we can help with. We’re starting to focus on some of those other projects.

ABG: What else do we need to know?

PM: It’s been kind of a labor of love and passion over the last four years. If you’re interested in the LaserWeeder, check out Carbon Robotics. We are the one and only LaserWeeder.

ABG: And I think you mentioned the website earlier. That was carbonrobotics.com.  Okay. Terrific. Thank you, Paul, For your time today. We appreciate your thoughts on technology and the future of agriculture. Join us next time for another installment of ag tech talk where we’ll talk with another expert improving the ag industry through technology. Thanks for listening.

PM: And I’d also like to point out that we have had enough demand for the LaserWeeder, now that we’re starting to sell some units in the 2024.

That speaks to how all the products have been doing. A lot of this has been people seeing it run in their neighbors’ fields and things like that. So, things have been going great, and growers are generally very happy with the machines.