In this interview of the Robots in Depth series our host Per Sjöborg talks to Melonee Wise, CEO of Fetch Robotics, and lifelong robot builder and developer. Melonee shares how she first got into building things at a young age and how that led to studying mechanical engineering and leaving her PhD project behind to become the second employee of Willow Garage. She shares some personal anecdotes from the first few years at Willow Garage, including both successes like the PR2 and some less successful moments. Melonee also gives her perspective on the development phase robotics is in now and what the remaining challenges are. Related to that, she discusses what is feasible to deliver in the next five years vs. what her dream robot would be.
Fetch robotics website.
Melonee Wise Linkedin.
This podcast is part of the Wevolver network. Wevolver is a platform & community providing engineers informative content to help them innovate.
Learn more at Wevolver.com
If you are interested in sponsoring the podcast, you can contact us at richard@wevolver.com
Per: Welcome to the podcast version of Robots in Depth Episode 1 with Melonee Wise in cooperation with Wevolver. Robots in Depth is supported by Aptomica. Visit Aptomica.com to connect. Welcome to Robots in Depth. Today, I’m honoured to have Melonee Wise here. We’re going to talk about everything in robotics. Could you tell me how you started?
Melonee: I think that robotics for me has been very serendipitous actually. I think that I always had a love of building mechanical things. I think one of my earliest exposures to building mechanical things is my dad sent me to summer camp, well it was like a day camp actually not a summer camp but Legos and programming was the focus of the day camp. This this was when I was like eight years old so think 1989. You learned how to program Legos. I was like the ambitious kid in the class. I was like I’m going to make a plotter. I don't know if you if you remember but they had like Lego kits where you could make a plotter and draw things. I got overly into that. I think that I built the plotter really fast. I loved Legos so I was really good at building Legos very fast. Then I must have spent like the entire four weeks of that day camp programming the plotter to draw all sorts of ridiculous things. I got really into just building things and I had a very voracious appetite for just any kit that I could get my hands on. Chemistry kit, Lego kits, The Scientific America, I don't know if you so for Christmas I would like make my Christmas list out of Scientific America servo kits like the light following robots.
Per: I'm thinking we're seeing a trend here. We’re things are going to go.
Melonee: It actually is kind of funny because my viewpoint of all that after all of that interesting stuff that I played with I don't think I had like the greatest guidance into what I should be studying when I got older. I went through high school. I did what all the nerds do and I took all the classes like calculus and advanced chemistry and all of these things preparing me for college I guess. Then I decided to become a mechanical engineer. I had this preconceived notion in my head that mechanical engineers build robots but it never dawned on me for some reason that I should probably learn how to program. I went to University of Illinois and I studied physics engineering and mechanical engineering at the same time. It’s very ambitious. I don't know why I did it. All through the curriculum there was no robotics. I was like I want to build robots but there was no robotics curriculum really at University of Illinois.
In physics engineering they try and help you find what you're really passionate about whether it's experimental physics, theoretical physics. As you advance through the program you're trying to find what you're interested in. I took this set of seminar series on analog and digital electronics. You could build anything you want and I was like building a robot. I started building robots again. My best friend Derek and I, we were probably 19 or 20 and we just started building robots, anything we could get our hands on. We had no money, completely broke and we would go scavenge parts. Our first robot we built it was called Zippy and it was made out of plywood and some motors that we had salvaged. Even then there wasn't much robotics. Then at U of I there was a group of alumni who wanted to do the DARPA urban challenge. They got involved with the university. There was this opportunity to come to California to work on this autonomous car project. I was supposed to be doing my PhD and I was not.
Per: You are still in physics for the PhD?
Melonee: I was doing mechanical engineering actually so I did undergrad physics and mechanical engineering. Then I just kept going with mechanical engineering but Derek and I decided on our winter break to come out to California for two or three weeks. I ran into the person starting Willow Garage. He was like I'm going to build a robot company. I was like okay, that sounds cool. He’s like, you don't want to go back and do your PhD, do you? I was like I really should do that. He’s like no, people with PhDs that's not exciting. It’s like what are you going to do? Go back and do that for another couple years. You could be building robots right now. I made the most risky decision I would say my entire life. On a whim I decided to leave my PhD program and come to California and start at Willow Garage and I was the second employee.
Per: The rest is history as you say. It was a risky decision but I think you would know right?
Melonee: It worked out but I just can't remember my frame of mind that I just decided yes, I can drop everything and move to California next week. That sounds good.
Per: I guess that's a fit between the person who asked you and what they ask you to do. I mean they just asked you. It was the right person asking you the right question and say, this just fits me. I mean somehow you knew you were going to do that.
Melonee: When confronted with the possibility of entering a robotics playground...
Per: How can you know? It’s not just going to happen. I guess the guy who asked you know that. He knows that this is a done deal. Would you talk a little bit about, you're now the second employee. There’s two guys, you and him there and you're going to revolutionize the world with robotics.
Melonee: Actually when I started it Willow it was me and two other guys. It was crazy because Willow was this big building and there were three of us.
Per: I've heard about this before that the starting team sits there in this huge space.
Melonee: It was funny because I don't know what we were thinking but we all claimed offices on opposite sides of the building. We’re just like oh that makes complete sense. We were working on the autonomous car and some of the people from U of I we're still coming out to work on the autonomous car. It’s kind of funny because my best friend came out with me to work on the car. He got the same offer hey, come work at this company. My friend was like you don't know this person. I'm going to go back and get a job. It’s funny because she ended out in California very close to me and eventually we convinced him to come to Willow Garage but it was very different in the beginning. There were three of us and then we got another person to join the team so then there were four of us. It was it was just so different. I can't even describe it. We had autonomous cars. We had like installed command centers in the autonomous car. We had a screen that came out of the ceiling and you just sit in the car and tune controllers and things like that as we were out in the field doing things. It was very exciting.
Per: I would imagine that.
Melonee: At the same time the founder also decided to start a boat project.
Per: Autonomous boat.
Melonee: Yes. They had a contractor doing that while we were doing the autonomous car project. Then eventually Steve Cousins joined so he was the sixth employee I think. Then that's when Willow went from like a ragtag bunch of individuals kind of building autonomous car and then there was a ton of us both thing to something that was starting to look more like a cohesive path forward possibly.
Per: Your start wasn't the garage. It was a huge space.
Melonee: Actually what's really funny is if you look at the original group of people, we actually started in a garage in Campbell for the autonomous car. They worked out of...
Per: I mean for being a car of course you were in the garage.
Melonee: We worked out of a friend of one of the alumni's houses out of the garage. It was kind of interesting because it did start very humble. From someone's garage to then moving to the Willow Garage space and then Willow Garage just grew from there.
Per: Then you went from the autonomous cars into other projects.
Melonee: The car project and the boat project had an interesting past. One of the things that happen with the car project is we had a DARPA site visit. Right before the visit started one of the engineers decided to program the firmware of the car but there was a ground loop in the circuitry and it fried all the electronics.
Per: Now you're dead in the water.
Melonee: There was no recovering and so that was kind of the end of our DARPA days. Then many of us moved on to work on the boat project. That was also not well managed. It was kind of like someone else was working on it and then all of a sudden we had to take it over. Some of the things that were done on it were not awesome. For example when the boat finally arrived and it was in the shipping container and we wanted to take it out to work on it. We pulled it out and all of a sudden we smelled burning. The person who had delivered the boat had wired all the solar panels backwards.
Per: As long as it was in the container it was okay.
Melonee: Can you imagine two people that were getting the boat out, checking it out and all of a sudden we smell this burning and the only thing we could do at the time because it was so hard to get the boat out that we just took cutters and we cut all the wiring out of the boat because it was really starting on fire.
Per: That's bad.
Melonee: I know. It was terrible and so the boat project was not long lived.
Per: It ended when you took the boat out of the shipping container.
Melonee: We wired it all back up and we did some tests, autonomous test runs in the bay because we had been working on the software in parallel with the hardware for a while but it became apparent that autonomous boats was not our strong point either. That’s when at Willow Eric and Keenan came in and pitched. We’ve created this PR1. It’s made out of laser cut wood and we could take this vision of a robot and turn it into the mainframe that is the computer equivalent for robots. That’s when a lot of the really big synergy started happening at Willow Garage. Suddenly we had a very strong focus and a direction. It was very ambitious. If you look back at how it started, I remember there used to be this, I bet you could probably find the cached page on Google. There’s this page on from PR.WillowGarage.com where it was like the PR2 timeline. If you look we were going to finish the PR2 and all of ROS 12 months. I mean really 12 months.
It was this crazy time because people were suddenly really excited and Willow was growing very quickly. We had gone from four people to six people to suddenly we had 15 people. Then we started the intern program then. The first summer we had 15 employees and 20 interns. It was suddenly this really energetic time. Many of those interns that summer went on to become full-time employees. It was just like this massive growth. ROS was iterating very quickly and PR2 hardware was coming on slowly but surely. All I can say is that in the first two years of Willow Garage it was electric. I don't think there's anyone who was there at that time that would disagree with that. I remember one of the employees telling me if I sleep under my desk robotics will move that much further.
Per: Because they didn't have to commute for whatever.
Melonee: It was just one of those times where you felt like you could really impact robotics.
Per: You certainly did. The PR2 is a magnificent platform and its really brought robotics so much ahead. So many people have started working on it. You have an enormous depth in robotics and the series is Robots in Depth. Could we just forget all this and just talk about robotics in general. Where are we now in robotics maybe related to the computer industry? Are we 1989, 1992? Then of course we're going to talk about the future too.
Melonee: I think the thing is if you look at robotics we're still in the infancy. If you look at the history we were talking about robots almost 40 years before we had even the simplest robots. That’s like 1940, 1950 we had the first robots. They were the turtles. Then we jump forward 30 years, we're in the 1970s and we have robots that take many hours to cross six feet of space. When you look at what they accomplished it's just mind-blowing. When they show some of the first shaky demos, if you listen to the video they're like oh yes, we're communing back we're communicating back to this huge mainframe that takes up this entire wall with radio. You’re like what?
Per: We're so spoiled with all the things that already work like communication, batteries, computers on a chip and whatever have you.
Melonee: One of the big things that happened in between in the early 1960s and the end of the 70s is that we were starting to introduce computers. Some of the first robotic arms were run on computers. The transformation math that we use and depend on today was kind of understood during that time period and even allowing us to start running robotic programs on the limited computers that we had. I think in that moment computers and robotics were kind of constantly pushing the limits of what we were capable of. At the same time that computers were in their infancy of some sort, robotics were kind of pushing the boundaries lockstep. Then one of the things that happened is you look at the 80s and the 90s and all of a sudden this massive diversity happened in robotics whereas computers seem to be kind of driving towards personal computer and other devices that we now rely on today. If you look at pictures of computers it looks like the evolution of man. It looks like we're all going towards walking upright and it looks like we're all running towards making tablets and phones. In robots it was like...
Per: More of a Precambrian explosion.
Melonee: Yes, it was. Suddenly you had people working on walking, people working on insect-like robots, all of this huge diversity. There wasn't a cohesive vision towards making a platform in robotics. PR2 is one of the big platform pushes in robotics that I know of. Maybe I'm wrong. In between the end of the 1970s, early 80s and 2010 we didn't move forward cohesively as a group. That’s for many reasons. We didn't standardize on hardware. Some of the hardware was innovating very rapidly stereo vision cameras, LiDAR sensors. In computers it was kind of like we know what we need to make computer work. These are the input devices. These are the possible output devices. Maybe you don't need any of that. Maybe you just need the input devices and the program but with robots even the simplest robot has almost infinite configurations when you think about it. Where do you place the sensors? What kind of sensors do you place? How do you use those sensors? What are they sensing?
Per: What is the purpose of the robot? Robotics is deeper, more powerful but it's also much harder.
Melonee: I think the other thing is that when people look at robots and then they look at their computer. They’re like oh, my computers can do all these amazing things but when you look at, computers have had almost 30 years to evolve even the software that makes them work. If you look at robots there was a lot of effort into different pieces of software but there was no cohesive vision. When ROS came along and tied some of that together and I wouldn't say that there weren't other frameworks that existed before us and that ROS is the ultimate solution but it suddenly created a cohesive community that enabled these leaps to happen in robotics. I see 2010 when the PR and ROS were released together as kind of a springboard in robotics. Maybe I have that that vision because I was there participating in it but I think you can truly see the impact of it. I really feel that we're still in a very infant stage because in all aspects of what robots can do because even if you look at the progression of the robotic arm. Let’s just slice everything down to probably the one thing that is probably the most advanced in terms of what it's been doing, the tasks it's been doing, the software that we have for it.
In 1961, Unimate hits the scene. GM, they're taking very hot things out of furnaces with the Unimate arm and that arm is not even computer controlled. It’s controlled by tape or magnetic tape. It’s just like a play and record. Fast-forward, 80s we're using robotic arms with teach pendants. We’re doing some coding. Then we continue to do that, we continue to do that and we continue to do that.
Per: No real innovation. I guess the innovation was more in under the hood than in the details because to adapt it to the market we're talking a lot of automated production in car factories and other factories.
Melonee: The way that we interfaced with those devices had minor evolution between 1961 and 2014. Now we're starting to see some innovation. If you look at the UR5 they've really innovated the interface for how you interact with that specific arm. Now with ROS-Industrial we're starting to say hey, there's all this research and this information that we have, this motion planning, collision avoidance, how do we make that available to these robust pieces of hardware that we've been using for the last 40 years. I mean we have so much more capability but we've yet to integrate it into these systems and how do we do that. I think that even in robotic arms we kind of stagnated for a while. I understand there’s safety and all sorts of other things and how we interact with the devices. Today it’s a lot easier because we have tablets because we have faster computers. Without those things it made it very hard to do much more than have a program logic controller that controlled these robotic arms. I think that robotics arms are a great example of the infancy of robotic technology and showing that we've come a very long way but at the same time but we've only come a short way in a very long path.
Per: If we look at the future, I mean cool robot you'd like to buy and you think with your extensive knowledge in the area that somebody could actually make that work. We could have that in three to five or ten years what would that be? I mean from the vacuum cleaner robot, the lawnmower robots, people have those already but what would be that next thing you think?
Melonee: The next thing I think is feasible or the next thing I really want?
Per: Let's do both. Let’s start with the one you want.
Melonee: The one I want I would just love to have a robot to clean the house.
Per: Pick up stuff not vacuum cleaner and sort and order.
Melonee: I think that's totally not achievable.
Per: What do you think would be the main challenge, if we select a very small scope not do the hard stuff of deciding whether to wash this t-shirt in 40 degrees Celsius or 60 degrees Celsius but is there any space there we could explore because I do want one myself too just like you want it.
Melonee: I think there are several key challenges to doing robots that we have in our home before we even talk about what the task is. Many of those key challenges are things like understanding social context, cultural context. If you and I have homes, we most likely have a bedroom, we most likely have a kitchen, those are simple concepts. When you come home what do you do? You take your keys out of your pocket. You place them somewhere.
Per: Charge the phone, take off outdoor clothing.
Melonee: What I do when I get home totally different so how do you semantically convey even that notion to a robot. What does it mean for the robot to live in your home?
Per: As supposed to any other home.
Melonee: First there's that context. Say the robot you we have is it's the come home robot. When you come home it does all that stuff for you? It has a little bin you give it everything and it's supposed to know what to do. I mean seems like a simple task but how do you even start to teach the robot what it means to come home and what it's supposed to do.
Per: I mean it has to be inferred from my actions by the robot because we can't expect people to program the robot. I mean if we could give them the manual and they program it that might be manageable but then we have a market of four. I mean the robot has to look at me and my behavior and deduce what I won't want to happen. I understand that that's a huge challenge. Do you see any progress in this domain?
Melonee: I think there are some progress. There’s this whole notion of semantic mapping. Now that we have the capability of dealing with very large databases it makes it much easier because then you can basically create a map and a database that represents your entire home not just where things are XY but XYZ and understanding and the tagging with that mapping be much more dense and complete because it's not just a bowl it's a white bowl. It’s made of porcelain. The robot needs to know it's fragile right and how it interacts with that is very specific. This is where some of the cultural things come in. How we keep our homes is very cultural. I think one of the next really big barriers long term is another example of when you when you have a programmer designing something and then they only test it on themselves. I think that that challenge is so far away removed from the challenges that we're trying to solve today that I think that when you think about the idea of bringing a robot into your home in the kind of generic concept text it's a very hard problem but I think there is some medium ground. What if I said to you know what? You can have a robot in your home and it will pick up your dishes and put them in the dishwasher and wash them and put them away.
Per: I’ll say here's my credit card.
Melonee: But all your dishes have to be bright orange.
Per: I'm okay or that. My wife, she'll also be okay with it for I mean say the everyday life. When we have a dinner party not so much for everyday life we are going to do that trade off and saying that if it's that kind of a day we're stressed, we have a limited time we'll do the orange plates and the robot will deal with it. If it's a Friday evening or a Saturday evening we'll do the nice dishes and the robot will just leave them alone. I think if we have a choice we're going to gravitate towards the orange plates and suddenly orange plates are going to be the standard and nobody knows why.
Melonee: I think that's an example of how I think robotics technology is going to make it into our everyday lives is that we're going to make these concessions. We already do it. People don't really want to admit it but many people clean-up for their Roomba. At Willow we did need finding and I interviewed many individuals who talked about how they cleaned up before their Roomba or their electronic vacuum. They would tuck in the tassels and they would pick up everything off the floor.
Per: You do have to pick it up from the floor otherwise it can move stuff.
Melonee: If you had come to someone before this robotic vacuum came out and said look, this little guy here is going to vacuum your whole room but you have to clean before you vacuum.
Per: It kind of defeats the purpose.
Melonee: When you think of it in the logical sense that seems totally foreign to us. Why not just get the vacuum out? When you think of it in the if I just keep everything off the floor every day this guy is going to run around and do this for me and I don't have to schlep this vacuum around. You’re like oh that that seems reasonable.
Per: It's a thing we do two together, the robot and me. I understand he the robot can't do everything and I don't want to do everything. I heard actually that in a family where they had teenage kids it took the kids only a few months to realize if they keep the floor clean of stuff the robot went in there and cleaned the room. They did that autonomously. They just realized that instead of the parents telling him to pick stuff up the fact that if they did the robot cleaned their room and they liked. That was the dream thing maybe, the robot that can do these things at home. Where do you see the feasible space? If I told you here's a bunch of money make me a robot that can be on the shelves for Christmas 2018 or maybe 2019. I'll give you another year there, the feasible one.
Melonee: I think the feasible robot is autonomous cars honestly.
Per: But that's not going to be on the shelves.
Melonee: I know it's not going to be on the shelf but you asked me what technology I think is feasible and I’d like to see that a lot because I think it'll revolutionize society in some sense.
Per: I mean just from the fact that cars kill a million people a year worldwide and then how many gets seriously injured.
Melonee: Yes, that's why I think it's very important. I can't wait for that to happen. Something that is available in 2019.
Per: We'll skip the production stage. We’ll say when you're done with it it's on the shelf. You’re probably better equipped to answer that question than anyone in the world. It doesn’t have to be general use. It can be for a particular person. I think this is the multi-billion dollar question.
Melonee: I mean the problem is that a lot of the things that that you would build cost too much to the consumer to bring into their daily life.
Per: I also think that if you had this pick stuff up cleaning robot or the clear kitchen robot with the orange plates I think you'd be surprised how much people would actually pay for that stuff.
Melonee: There's lots of there's lots of examples of that like HERO robots, Heathkit HERO robots, there were more than 10,000 sold and the equivalent price was almost $15,000. There’s there some good I guess history for demonstrating that expensive robot kits will sell into I guess hobbyist/consumer markets. I mean the sweet spot for consumer product that you want to put on the market, it seems like the tolerance is for everyday consumer items may be in the hundreds of dollars for luxury items or things that people like TVs, I mean the upper limit is like $5000. When you look at what we can deliver for that price it's not very much.
Per: It still has to come down.
Melonee: Yes because you need economies of scale. That’s the biggest problem. I think a lot of the technology that you'll see that people probably want to buy will be for specialized markets like autonomous wheelchairs.
Per: Less price-sensitive of course.
Melonee: Those are good applications I see on the horizon, prosthetic limbs, medical devices that we are not that price sensitive to I think are definitely capable in the next five years. I don't see that is a big challenge. I’ll say this Brian Gerkey once told me that that roboticists are the worst at coming up with applications for their technology. I totally agree. I think that sometimes it's you know so much of what can't be done that you convince yourself it can't be done. Sometimes I feel like I'm a dangerous person asked what is and what isn't possible because I'm a pessimistic optimist.
Per: You look back at the time when you did that timeline of 12 months until ROS or PR2 is done and you say, I’m going to add a few years to the next project but I think also this is a very interesting insight where you have this experience and it's like that decision of not going back to grad school and finishing your PhD. When you find that idea or somebody introduces you, you can think about it and you said wow, I can actually do that then you're going to have a very solid footing for that feeling. You can't really articulate it and you really explain it but you just know that that's going to be possible and there is going to be a market and it's going to be used then it's going to be great.
Melonee: I think some of the cynicism or the pessimism that I have is like so today, if you pick up a gallon milk accurately how much do you think that costs?
Per: No idea but quite a lot of money I would presume.
Melonee: Probably $40,000. That’s just the arm.
Per: Then you need the sensors.
Melonee: A gallon of milk is pretty heavy. If you set that as the bar of something that you have to pick up in the home.
Per: I mean it's not that heavy. We’ve got stuff in our homes that are in that weight spot. This been fantastic. Thank you very much. I hope you liked this episode of the podcast version of Robots in Depth. This episode is produced together with Wevolver. Wevolver is a platform and community providing engineers informative content to help them innovate. It is how engineers stay cutting edge. Aptomica is the founding sponsor for Robots in Depth. Aptomica runs anything in modular robotics. Dream, rent, build. Visit Aptomica.com to connect. I am your host Per Sjöbor. Until the next episode thank you for listening.