Joe Jones likes to build practical robots… the type of robot that can do enough on their own to more than make up for their cost. His best known example is the first widely adopted consumer robot, the two-million-selling iRobot Roomba vacuum cleaner, which Joe, along with co-inventor Paul Sandin, developed while working at iRobot.
Jones is also the author of three books, including, Mobile Robots: Inspiration to Implementation.
He graduated from MIT, served for nine years as a research engineer at the MIT Artificial Intelligence Lab and worked as senior roboticist at iRobot Corp. for nearly 15 years, before both he and Sandin left in 2006 to found a company dedicated to the next great thing in robotics. Launched as QRobotics, the Groton, Mass.-based company has since been renamed Harvest Automation, Inc. and is working on ways to automate the labor-intensive process of growing commercial potted plants.
Jones, a 24-year veteran of the robotics business and holder of nine domestic and several foreign patents, Joe Jones is currently chief technology officer of Harvest Automation, whose brain trust includes not only Sandin and Jones, but iRobot alumni Charles Grinnell and Clara Vu. Jones agreed to discuss the venture recently with Robotics Trends Contributing Editor John P. Desmond
Robotics Trends (RT): What is the backstory on the founding of Harvest Automation?
Joe Jones (JJ): iRobot was a great place to work; it was fun working on Roomba and developing all that stuff. But, I am an inventor. I am always looking for the next thing.
When a company becomes publicly traded, which iRobot was after Roomba became successful, Wall Street wants you to keep putting money into the things that have already been successful. I was interested in branching out and finding new applications for robots, because I thought that while consumer floor care was a wonderful market, for me the question was what do you do next?
My thinking was that there are probably a lot of niches out there where a robot would fit perfectly, but you have to find them and they might not be related to consumer products. So we founded a new company to search far and wide to find the next place where robots would fit. That is the reason we founded QRobotics, just before Thanksgiving in 2006.
When we founded QRobotics, we did not know what the next application would be. We spent a few months looking for an application that would be suitable for robots. Once we found that, we thought the best thing to do would be to transform into the company that would exploit the application we found. So we became Harvest Automation in 2008.
RT: What is Harvest Automation doing?
JJ: We found this very cool and very apropos application for robots. If you go to a garden store, you will find many potted plants sitting around. All those plants came from container farms scattered around the country. The largest ones are hundreds of acres.
At a trade show we discovered that there is an operation everyone in this market performs a spacing operation. When you first put the seedling in a pot, it is okay for them to be next to each other. But after a while, the plants start to grow and if you do not move them, they damage each other.
[The growers] might move them from a greenhouse to a field or from one field to another, but [the plants] go from being close together to being spaced apart. For 50 or more years, the industry has been doing this manually. We thought this was an ideal application for robots.
RT: Where are you at in development of the product?
JJ: We do not have a product on the market yet. We have some prototypes, and we hope to develop the product and get it to market in the next couple of years.
We have knocked down many of the largest areas of technical risk; but we have a few more to go. After that, we will enter the development phase where we hope to build products reliable enough to work for customers.
RT: Can you describe any of the technical hurdles you have overcome?
JJ: When you put pots down in the field, they are supposed to be spaced in a particular way, so you have a nice pattern, either a grid pattern or hexagonal pattern for instance. The pots need to be within an inch or two of proper placement.
We believe we have resolved that in a cost effective way, with our system called ‘marked boundary’ and [by] dead reckoning. Errors can grow as the robot moves around, so we reference the robot to this boundary. It is a mark that the robot is able to detect. We have patents are pending on it.
RT: Can you describe an area of technical risk you have yet to overcome?
JJ: The robot has to find the pot, pick it up and put it down somewhere else. The thing we are working on now is to give the robot an accurate idea of where the pot is.
We have a way to do it now, but we are not satisfied with it. We are using discrete components designed for something else. So we will either redesign those or use a camera system designed to our specifications. All this is classified as industrial automation.
RT: Can you say a little more about the potential customers and the niche you are trying to fill?
JJ: Absolutely. There are 7,000 growers in the US who grow potted plants for resale. They range greatly in size. Most are below $1 million in sales but some are as much as $10 million in sales.
We have been in contact with 20 of the largest growers in the country, meeting with them, getting ideas, taking pictures and quizzing them. The reception has been great. It turns out this is a real sticky problem for the growers. It used to be the case where you could get teenagers in the summer to go out and work in the fields.
You cannot do that anymore. It’s a tough job, moving containers around morning to night. You have to do it rain or shine because plants don’t stop growing. It can be 120 degrees on the fields in the summer. The black cloth on the ground that stops weeds from growing makes it hotter.
It is difficult to keep the laborers once you have them, and as many as 80% of these workers are undocumented. So the labor uncertainty is very great. If the grower gets a call from INS (Immigration and Naturalization Service, now the Bureau of Immigration and Customs Enforcement), the workers might run away. If the growers do not have the workers, the plants just sit there. So the growers really would like to have an automated way of doing this.
RT: Is there any competition?
JJ: There have been attempts but no one has been successful yet.
RT: How are you funded?
JJ: We were self-funded the first year, then we took a “friends and family” round, and now we are in negotiations to get funding from professional investors.
RT: How many people are working at Harvest Automation?
JJ: Right now there are four founders and some consultants.
RT: What other kinds of robots are you working on?
JJ: We considered a lot of things and did a lot of brainstorming. We came up with 15 possible applications and as we thought about it, the pot moving application rose to the top. I do not want to say too much about what those other applications are.
RT: How does the future of commercial robotics in general look to you?
JJ: It looks strong.
Our vision of the future might be different from how robotics has been approached over the past two or three decades. The thing that will drive robotics is actual robots doing real work. The challenge is not to come up with a spectacular technology, like Commander Data from Star Trek. Instead, people will find an application in an industry that is just one technical step beyond what robots can do now. We are on that trajectory.
When we invented Roomba at iRobot, there was no robot you could buy commercially. It seemed that was the low hanging fruit. It worked out well, but we had a robot that could just roll around, so it can clean the floor, or mow the lawn if it had a blade. That was kind of unsatisfying.
The next thing robots can do is a bit of manipulation. We have a one degree of freedom manipulator. People have tried before to have general purpose manipulators, with for instance six degrees of freedom, and those are complex and unreliable. It seems like that is not the place to start; the place to start is the simplest thing for the robot to do.
RT: Can you explain the idea of degrees of freedom?
JJ: The robot can drive to a pot, pick it up, drive somewhere and put it down. It has one motor, so one degree of freedom. A general-purpose robot arm might have six motors—the waist, shoulder and elbow gives you three degrees of freedom. That allows you to put the end of the arm anywhere in space. Then with three more degrees of freedom in the wrist, it lets you point anywhere in space. That’s six degrees of freedom all together. Robots that have that are expensive, so we wanted to start with something simple.
RT: Do you have any advice for young people interested in robotics?
JJ: They should follow their interest, be it software, electronics, mechanical engineering, sensors or anything like that. And try to get a wide range of different experiences. We have often found that people who make the best roboticists have had a great variety of experience outside of engineering.