Vision Robotics Corporation (VRC) was founded in 1999 with the goal developing vision-based technologies to produce practical, autonomous robots. The company initially developed consumer systems, but entered the agricultural market in 2004 with a feasibility study for harvesting oranges. In 2011, VRC was approached by lettuce growers to develop a lettuce thinner. In a few months, the company built its lettuce thinner, which was released in 2012 and since then has collectively thinned hundreds of thousands of acres.

Tony Koselka, co-founder and Chief Operating Officer of Vision Robotics Corporation is an award-winning design engineer who holds 21 US patents. He was recently interviewed by Joanne Pransky for Industrial Robot. That interview is reprinted below, with minor modifications. The original interview can be found HERE.

---

Joanne Pransky: Can you describe the technological evolution of VRC’s products and systems, highlighting key years and developments?

Tony Koselka: Our first product was a commercial vacuum cleaner. We started thinking of a home sweeper because that was low power and easier, but we found commercial partners that made vacuum cleaners.

The first product was a canister vacuum cleaner. It had the power cord on the back, so it was full power, with a little tiny cleaning head that was square and could get under everything. All the other vacuum cleaners were round because round systems cannot get stuck, where square things can. We were working with a couple of large manufacturers, and we beta-tested our vacuum cleaner in people’s homes beginning in 2008.

Though it got very good reviews, our partner decided that it was the wrong time to introduce a $2,000 vacuum cleaner, and particularly during a financial crisis, so the project got canceled. In the meantime, around 2004, a grower came to us and asked us about picking oranges and that’s how we started in agriculture.

We did a lot of work with counting apples and oranges on trees. Using cameras we were able to identify and count greater than 95 per cent of the visible fruit with very good sizing accuracy. We were close enough to a product, but the industry couldn’t put a monetary value on what the data that was going to be provided them, so the crop-load estimation system project died slowly. However, growers still get excited about it, and we are frequently approached about the technology.

We then experimented with a grapevine pruner and worked on that for a number of years. Eventually, somebody came to us with a request for a lettuce thinner and with our core technology, we whipped the lettuce thinner out in six months. That is now our main product.

One of our core design philosophies is to make the products very modular with the key electronics essentially being black boxes for the growers. Because of this, many of the owners have modified their thinners, with several of them designing and fabricating new frames. We have thinners that are sometimes in smaller configurations, and sometimes, multiple thinners are assembled into a single bigger one. We like the fact that our customers are comfortable doing this and that our equipment is so simple that these reconfigurations work.

There is a next-generation lettuce thinner that I call V4 – it’s the fourth major design. It looks like it’s going to be a huge win. It’s much lighter than competitive systems. It’s wireless so the wiring is simple and it’s easier to adjust. These are things that the growers really appreciate. V4 was released in 2019, and we’re presently building a batch order of four machines.

We’ve had an idea for some mechanical cultivators and weeders, and we worked on one with another company a few years ago, in which we literally dropped our lettuce thinner software into their mechanical system and it pretty much worked. We had some suggestions on how to make it a better product and nobody wanted to hear it, so when we got some time and money, we finally developed the mechanical cultivator ourselves. A lot of the technology being incorporated into today’s agricultural robotics was thought of and tested back in the 70s and 80s, so it is hard to get meaningful patents. It’s all been done before, but we think with our mechanical weeder we have a potential intellectual property that would give us a sustainable competitive product.

Pransky: How is VRC funded?

Koselka: We are privately funded. At our inception, we raised start-up money from friends and family. Then, the vacuum cleaner partners funded us for a couple years. We also received a few government grants. The grapevine pruner project was funded by a consortium of mostly vineyard owners through its own LLC. Currently, our money comes primarily from product sales.

Pransky: Is VRC profitable?

Koselka: In 2018/2019, we were profitable, but we had several years that we were not. We made it through the lean years by tightening our belts. Admittedly, fundraising is not our strong suit. My partner and I would rather do the technology, and we never got the right marketing people to secure outside funding.

Pransky: Does VRC want to be involved in the mechanical design of robot systems or would they prefer to find “mechanical” partners?

Koselka: Actually, we have used mechanical partners and have found it is often a mistake. We are primarily a software company. My co-founder, Bret Wallach, and I are both systems people. While much of the technology we use is not difficult or is at least well understood by people in the field, we find that the final products tend to benefit from understanding the entire system, so we find it better to do it ourselves.

It’s not our desire to manufacture as we prefer to come up with the products and do the technical development. We outsource most of the manufacturing and try to only perform assembly of the electronics. At this time, we are still purchasing and coordinating the components, but we are willing to transition to an even more turnkey system. Our quantities are still small enough that we can control it. If we start selling 100 cultivators a year, we will need new systems.

Another core philosophy is to use stock parts whenever possible. This enables our customers to largely repair and upgrade their machines themselves. Unfortunately, this is not practical for most of the electronics, but we keep our eyes out for opportunities. Currently, we envision the cultivator to be an autonomous mobile robot rather than being mounted on a tractor. I don’t want to design a drive system if I do not have to. There are many of them out there, so we are looking for an electric drive subsystem partner.

Pransky: The weeder and thinners use conventional agricultural machinery like tractors. In Europe, they have concerns about impact damage to soil caused by heavy machinery. Is there scope for using multiple smaller and lighter ’robots’ instead of tractors?

Koselka: Absolutely. In fact, that’s why I said we need an electric drive system because we want a smaller, lightweight weeder that is self-propelled. A weeder around lettuce is limited to a top speed of about three-quarters of a mile to a mile per hour, simply because you’re plowing through dirt. Operating much faster than that starts to throw the dirt, which gets in the plant and either kills small plants or gets stuck on the inside, which causes potential damage and safety issues as the plants grow around the dirt.

But on the big lettuce thinner, where it’s five minutes down the row, we don’t want to be in charge of turning a 20-foot- wide machine at the end of the row. That turning time relative to the straight-ahead time is so high that it did not make sense to make it self-propelled, so once again the right product is an exercise in system design.

Pransky: Europe also has concerns about the use of weed killers and fertilizers. Could the weeding and thinning be done purely mechanically?

Koselka: What I have been calling the cultivator is also a mechanical weeder… we use the terms interchangeably. There are already a number of mechanical weeders on the market already, but none seem to be as good, which is why we are excited about our idea. We initially looked at a mechanical system for thinning lettuce, but that is harder to do while working around the small delicate plants. Instead of being able to thin at two-three miles per hour, it would have been more like one-half to one mile per hour, which essentially doubles or triples the cost of thinning.

As to the rest of your question, we do not specify the chemical, but most operators thin with fertilizer, not herbicide. The fertilizer kills the small lettuce plants. When the fertilizer used during thinning, is included in their fertilizer plan, they actually end up using five or ten per cent less fertilizer, but get the same amount to the plant because we only spray a four-inch band right around the plants with none of it going into the furrows. There are also organic growers who use organic herbicides to thin. We looked at lasers, we have looked at flames, but it turns out spraying is just way better for thinning.

Pransky: Can drones be used for mechanical thinning?

Koselka: I guess the answer is yes, but why? I’m amazed by the drone business. There is all sorts of money going into high-tech agriculture, and a lot of it is going into drones. People have proposed drones for harvesting fruit and for thinning and weeding. Why would you want to do that? Things floating in the air are much harder to control for performing any precise action than something sitting on the ground. In fact, a drone could spray, but having the ground and stability is just so much simpler.

Pransky: Field sizes in the USA tend to be much larger than in Europe. How does this impact the cost effectiveness of the various approaches?

Koselka: Significantly. Lettuce thinners thus far have been limited to Arizona and California because they are the only regions with large enough lettuce fields for the lettuce thinners to be practical.

Our earlier designs on lettuce thinners had a minimum capacity of three lines, but six was the practical smallest size. In round numbers, that can thin more than an acre per hour. Given the cost of the machine, thinner design has a minimum farm size of several hundreds of acres for viability. Our latest design is more modular, with a computer for each line. The rest of the system is also simpler, so potentially is is a viable solution for smaller farms. There is a grower in Australia who is interested, but then you run into other issues such as spare parts and training. All in all, I would like to move beyond California and Arizona, and I’d love to be overseas someplace.

Pransky: Are there any armed mobile robotics platforms currently available that are suitable for agricultural use?

Koselka: There are a few of them out there presently being used in vineyards and a number of prototypes of various systems. I have seen a video of a solar-powered spray weeder that uses an arm to spot-spray small areas.

Pransky: Does Vision Robotics use students to help in product development?

Koselka: We have had interns off and on and worked several projects with different universities, but we don’t have any right now. Students in general are really interested in robotics, and the ones that come from farming families are especially interested in agricultural robotics.

Pransky: What do you think PhD and masters of engineering students should be doing while in school to prepare them best for the commercial side of robotics?

Koselka: I am a mechanical guy and many of the mechanical design programs are really good. They have hands-on projects that give very good experience. My advice is to remember that commercial applications require mechanical, electrical and software coupled with an understanding of the true requirements. Successful products solve an actual and specific need rather than simply being cool.

Pransky: What do you see for future applications and enhancements to your products?

Koselka: One of the reasons that John Deere has invested in robotics is they want to spray less chemicals. Roundup Ready* crops have been a boon for yields and for maintaining the cost of growing, but have resulted in the spraying of lots of Roundup. The next step for John Deere and its competitors is to target-spray, where instead of spraying the whole field, you spray less chemicals and just where the weeds are. That’s what we’re doing. We’re a small company that has chosen to focus on specialty crops such as lettuce, instead of going into the big-acreage row crops. Our mechanical weeder should help eliminate the need for some herbicides.

People are putting lots of money into farming and agriculture, and there’s going to be a great deal of change over the years.

(* Editors Note: “Roundup Ready” crops are crops genetically modified to be resistant to the herbicide Roundup, the brand-name of a herbicide produced by Monsanto.)

---

Joanne Pransky has been an Associate Editor for Industrial Robot Journal since 1995. She was also one of the co- founders and the Director of Marketing of the world’s first medical robotics journal, The International Journal of Medical Robotics and Computer Assisted Surgery. Pransky also served as the Senior Sales and Marketing Executive for Sankyo Robotics, a world-leading manufacturer of industrial robot systems. She has consulted for some of the industry’s top robotic and entertainment organizations, including Robotic Industries Association, Motoman, Stäubli, KUKA Robotics, ST Robotics, DreamWorks, Warner Bros., and for Summit Entertainment’s film Ender’s Game, in which she brought never-seen-before medical robots to the big screen. She can be contacted at joannepransky[@]gmail.com.