The Pransky interview: Dr Paolo Pirjanian, CTO of iRobot Corporation

Pransky: Tell us about the technology you spear-headed at Evolution, highlighting key years, developments, partners, etc.

Dr Pirjanian: When we started Evolution Robotics in 2001, we realized that one of the first key building blocks for robotics was autonomous navigation. At that time, the state-of-the-art in the autonomous navigation field was mainly being driven by people like Sebastrian Thrun and others from Carnegie Mellon University who were using expensive laser rangefinders such as SICK, and developing that with simultaneous localization and mapping (SLAM), a probabilistic approach to building a map while figuring out its location.

Early lab results showed that this can be done, but with hardware costs of tens of thousands of dollars. At Evolution, we were focused on going after the consumer market, and even in the other markets, $10,000 was not a deployable solution. We therefore decided that we needed to find a much more cost-effective approach to solving this navigation problem, and my intuition for this solution was the use of cameras. Cameras, which were just getting embedded into cell phones, cost less than a dollar in high volume, and we knew we could ride that curve of volume, compared to a $20,000 laser range finder. The challenge, of course, was to develop algorithms that would be able to interpret the images properly to build a visual map of the environment, and then to use the map for the robot to know its location.

We started developing these algorithms and then I stumbled upon Professor David Lowe from the University of British Columbia who had developed and patented a visual recognition algorithm, known in the industry as the scale invariant feature transform (SIFT). SIFT was a paradigm shift in the computer vision community, and although I saw this algorithm as rudimentary, I also saw promise in it. Evolution immediately licensed the SIFT patent and we then incremented an algorithm to develop the vision pattern recognition technology and combined it with SLAM technology. We called it visual simultaneous localization and mapping (vSLAM), a term we coined. The algorithm builds visual landmarks and these landmarks become the anchoring points into the real world. Whenever the robot sees these landmarks, it is able to recognize this position very accurately and figure out where it is.

We implemented vSLAM and at the time we implemented it, the sensor was inexpensive, but the computation required a lot of processing. Even with the Pentium grade processors we were using, the solution was orders of magnitude more cost-effective than the state-of-the-art laser-based SLAM.

However, while we were working on the optimizations for vSLAM, we could see it would take years of work before we could get the cost lowered to fit into consumer level products. Under frustration, I started thinking about what we can do for an alternative solution that could meet the consumer price points needed today, rather than having to wait another three or four years. Out of this, NorthStar® was born and although NorthStar® is not as elegant a solution as vSLAM, it was a very practical way of solving a hard problem at an extremely low-cost. NorthStar® takes infrared lights or beacons and projects infrared spots onto the ceiling that act as GPS satellites. An optical sensor on the robot tracks the position of these spots, and by way of triangulations, the same way as GPS works, figures out the position of the robot very accurately.

In the consumer domain, if the product is retailing at around $200 (the price Evolution's Mint floor cleaning robot was launched at in 2010), that means the builder's opportunity cost cannot be more than $40-$50 for the entire product, so navigation has to be a sub-fraction of that. We got NorthStar® to that point and then NorthStar® became Evolution's flagship, the world's lowest cost navigation technology even today by far, by orders of magnitude. We then developed the Mint robot based on NorthStar®, to have a robotic floor cleaner that maps the environments, systematically covers it, and doesn't miss a spot.

Pransky: You went from Chief Technology Officer (CTO) to Chief Executive Officer (CEO) of Evolution. Please describe which one you enjoyed more and why.

Dr Pirjanian: I did enjoy being the CTO very much. I was frustrated with the fact that we were unable to gain business traction, and I think most scientists and engineers at heart are problem solvers. I could see there was a problem in figuring out the business, and I started finding myself spending more and more of my time in business development. Although this had nothing to do with science, the same skill sets of problem-solving and analytical thinking were utilized. That's what drove me more and more towards the business side and the CEO position.

Pransky: Describe the background in which iRobot Corp came to acquire Evolution: How and when did the partnership begin and develop? What was the primary reason iRobot acquired Evolution? How many employees were working at Evolution and what has happened to their jobs after the acquisition?

Dr Pirjanian: All the employees (now iRobot employees) are there in our Pasadena, California office.

Colin Angle, the CEO and co-founder of iRobot Corporation, and I, have known each other for many, many years, while I was in school, and before iRobot was a company. When I joined Evolution Robotics, we talked a lot at every robot forum and we followed each other. I admired Colin for the business platform that iRobot had developed, and the fact that he was persistent and so passionate about robotics. Most entrepreneurs would have given up a long time ago. Colin stuck with it. When I got involved in the business, I started gaining more appreciation for what iRobot had accomplished from a business perspective, and I think Colin had gained a lot of appreciation for what Evolution had done by way of technology development.

When I became Evolution's CEO and launched Mint, Mint was a success story in terms of the revenue. The revenues grew to a point where Colin could make an offer that the venture capital (VC) partners at Evolution could find acceptable. He gave me a call in 2012 and asked what I thought of putting the companies together and that's when we started talking.

Pransky: Is having the intention of selling a company to a major player like iRobot a good business strategy for a start-up company?

Dr Pirjanian: I do believe that generally it is extremely important for any startup to consider their exit strategy early on. Particularly, startups with VC backing need to think about how they create value for their investors and certainly an acquisition is a common way to accomplish this. Many startups begin by understanding the needs of big players in their field and “design” the company with the goal of someday selling to their competitors or dominate players in the field. It is however, important to not lose sight of your long-term mission, and focusing solely on building a company that will be acquired should not come first. Focusing on solving real-world problems and adding real value is what will create the opportunity to be an acquisition target.

Pransky: Please describe the Mint's technology vs iRobot's technology and the product differences.

Dr Pirjanian: The synergies were everywhere. The Mint, now called Braava, was very complementary for the Roomba® and Scooba business. Roomba is a vacuum cleaner and the Braava was designed for mopping hard surface floors. The revenue provides an additional revenue source for iRobot. iRobot has international channels for the distribution of the Roomba®, so from my perspective, it was going to accelerate the business growth for the Braava by plugging it into the channels that iRobot had already developed. And, we are seeing that growth.

In addition to the product side, Evolution's navigational technologies, NorthStar® and vSLAM, are also very complementary to iRobot's technology. On the high-end of navigation is RP-VITA, an FDA-approved robot for hospitals that uses laser rangefinders and a variety of other sensors. NorthStar® and vSLAM broaden the range of iRobot's other product categories, including those from the home robots business unit.

Finally, the teams are very complementary. We had the same passion; we had the same roots really, though we were coming from very different directions. During the due diligence process and the negotiations, as I got exposed to more and more people from iRobot, I realized we were different but at the same time, very similar. The similarities were that we had a very common passion and we talked the same technical language.

One of the things I struggled with the most when I was CEO of Evolution, was when I was interfacing with the VC community or the retail customers, they did not understand our vocabulary. When I talked to them, I needed to be able to talk about the benefits rather than the technology. For instance, when we launched Mint we made an explicit decision not to use the word “robot” in the description or positioning of the product because in several focus group studies we learned that many consumers relate “robot” to a sterile, industrial device which would have not been great for positioning a consumer product that needs to work in the home. Another example is that when talking about localization and mapping technologies to the consumers, we chose to talk about the benefits, e.g. “NorthStar tracks when it cleans so it doesn't miss a spot!”

Pransky: Are there any new emerging technologies that combine Evolution's NorthStar®, vSLAM, and iRobot's navigation platforms?

Dr Pirjanian: Navigation has reached a level of maturity which is going to influence a lot of our products. iRobot's Ava 500, a remote presence robot for the enterprise market, is at the high-end of the market and iRobot's Braava (formerly Evolution's Mint) is at the low-end of the market, incorporating navigation and smarts. These robots are becoming really intelligent and aware of their environment, and you will see more of that to come from iRobot.

The next stage, not only for iRobot but for the robot community in general, is manipulation. Physically being able to interact with the environment is what's missing in robotics now. Navigation will take us to a certain extent and allow us to grow, but what follows that is going to have to do with manipulation. Manipulation is the next wave and absolutely something that we have a keen interest in and internal research and development investments in.

Perception and algorithms that can help interpret the environment – the context to support appropriate decisions – are key steps to not only supporting navigation and manipulation, but to also supporting steps beyond that, including human-robot interaction which involves much richer and more reliable voice recognition, understudying language, gesture recognition, etc. We believe computer vision as well as 3D sensing technologies will play a significant role in this. In addition, we believe cloud computing, combined with machine learning algorithms to harvest and learn from the wealth of information online, will play a significant role in robotics in the future. However, the “missing link” in robotics currently is the ability to physically interact with the environment which I would generally categorize as manipulation.

Pransky: You are now CTO of iRobot Corp. What is it like being CTO of a company that has 500 employees and almost 500 million dollars in revenue?

Dr Pirjanian: It's fun. I'm enjoying it because it's a bigger platform that can have a bigger impact on the world. I think iRobot's business and technologies platforms combined with Evolution's technologies and team is a really good combination that can fuel a lot of things. This was a really good marriage I would say, and I'm enjoying it from that perspective. Also, I'm enjoying the fact that I don't have to deal with certain things that I used to deal with as a CEO of the company, so I can focus more on the things that I really enjoy doing.

Pransky: How many Mints have been sold and how many iRobot Braavas have been sold?

Dr Pirjanian: While we can't provide details on the actual breakdown, we can say that it's been well received internationally as well as in North America. I think that one of the things that has been great about this product is that it came at a time when iRobot was expanding on a global basis, and we were filling out our portfolio. Different markets have different needs and the Braava platform has been perfect for us to be able to complement some areas of the world which require a different type of cleaning. Vacuums are not by any means common place and standard across the world. For example, South America and Asia, a barefoot culture, have mostly hard surface floors.

Pransky: What is the biggest mistake you made/greatest lesson you learned throughout your career?

Dr Pirjanian: From a technology standpoint, the biggest lesson I've learned is you have to ground your research and development into reality, which is a contrast to what I used to do in academia.

For instance, at the time in 2001 when I started the company, I had an advisory board of the most prominent robotic researchers/professors from around the world. I remember we were talking about navigation, localization and mapping and they said, “Solved. It's a solved problem”. But until today, 12 years later, it is only now that we consider it being mature enough to be integrated into products. What academia considers as theoretically solved – and I came from that background – and solving something in practice are two completely different things which may require a completely different paradigm shift, which is what we saw, for instance, with vSLAM or NorthStar®. A laser range finder solution is still today costing on the order of thousands of dollars, but if I want a $10 solution, I need to figure out vision and pattern recognition into something completely different. That lesson for me has formed my approach to innovation and the way I define it. I say that innovation has not been fulfilled until your technology is contributing into a successful, profitable business, as it's not until then that you really have solved the problem. I feel I'm truthful to the scientist in me as well when I say this, because if I'm not able to take research work and translate it into profit for a company or for any entity; or, have a positive impact on society, I won't get to do this again. If technology stays in the lab, I have had zero impact on the world. It has to translate to some value for someone. That's the biggest learning I would say I've made, and in getting there, I have made a lot of mistakes to learn that.

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?

Dr Pirjanian: I think it's a really good combination to have someone with strong technical background and degrees, to also gain an understanding of business. So, complementing a technical education with business courses in business or entrepreneurship I think will ground them and prepare them much better for going into a commercial entity. That's where magic happens; at the intersection of technology and business.