I was lucky enough to interview Steve Cousins. Steve has a deep robotics background. He was the CEO of Willow Garage for 6 years before starting Savioke, maker of the robot, Relay. Relay is in several hotels, delivering items to guest rooms. The next time you stay at a hotel and call the front desk for toothpaste, you might meet Relay. That would be cool. It’s also a beautiful robot.
I invited Steve on the show to tell us more about his very interesting and robotic-rich background and exactly what they’re doing at Savioke.
Here are some other things we talked about:
-Did you like to build stuff growing up?
-How did you develop the tech for Relay?
-What were some of the greatest challenges navigating a hotel? The answer makes sense but I didn’t think of it.
-How have people responded to Relay?
Dave Kruse: Hey everyone. Welcome to another episode of Flyover Labs and today we are lucky enough to have Steve Cousins with us. And Steve is the CEO of Savioke, which is the maker of the robot Relay and it’s quite cool what really they can do. Essentially right now Relay is working with several hotel and they deliver items for gas. So let’s say you’re in a hotel room and somebody wants – if you want toothpaste you call the front desk and the front desk will put a tube of toothpaste and Relay the robot and the robot will deliver it to your room. So it sounds quite Sci-fi which I like and so that’s why I invited Steve onto the show to tell us more about his very interesting robotic rich background and exactly what they are doing at Savioke. So Steve, thanks for joining us today.
Steve Cousins: Yes, thanks for having me.
Dave Kruse: So before, yes and like I mentioned in the intro, I mean you have quite a rich background. Can you tell us a little bit more about your background?
Steve Cousins: Probably the highlight is the last company and this company. So I ran a company called Willow Garage which was – we sometimes call it a think tank or an incubator. Somebody called it once a halfway house where they would research for something they are waiting to start off. But Willow Garage was a pretty amazing place. It was funded to basically have impact first and worry about return on capital second, so not a typical start-up and as a result we were able to attract researchers who may be willing with the more risk adverse than starting a startup before this robotics boom that we are in now had really gotten going. We started with kind of random projects. There was an autonomous car project that had some starting to do with the Founder. There was an autonomous book project that he also started and when I came on I kind of rounded out the portfolio by bringing in projects from Stanford called the Personal Robotics Project. And after a few months we decided you know, let’s just focus on the personal robot, something we can really be world class at and so we did and there’s two parts to that. There was this amazing robot called the PR2, Personal Robot 2 which was – in hindsight we went back and renamed the Stanford prototype the PR1, so the PR2 was the second generation of this and then there was the software to go with it. Really one of the, probably the primary goal of this project was to make it so that all the research labs around the world who are working on robotics could share a quote. And that was something that really couldn’t happen before that because there was no common hardware platform and then there was – there were a bunch of different kind of open source software platforms but all pretty much created by a few grad students at a lab somewhere. No major serious investment. And so what we did is took the fact that we had the ability to build a world class team. We built this PR2 robot and we also put about 100 to 150 engineer years into the robot operating system and really had it become just dominant in terms of robotics, open source robotics software, because basically you know it’s very hard to put that kind of energy and effort into something, unless you are very well funded and so we took advantage of that fact and created the software that’s now kind of knitting the world of robotics together.
Dave Kruse: Yes, interesting and that was a big contribution that you guys gave to the robotic community. And can you just give an example of you know what – how robotic operating system enabled a researcher or a robotic developer with using the operating system that they couldn’t do before?
Steve Cousins: Yes, I mean if you think about the problem of building a robot, right, you have – I mean first you have the mechanical and electrical engineering pieces and so you’re going to put together some hardwire that can move in certain ways. But then the next problem you have is you’ve got software all the way up the stack. So from device drivers to control the individual motors or individual sensors to the communication system to put those together, to the algorithms on top of that, that motion planning and other kind of higher level robotic actions and perceptions and so really it’s hard for anybody to be an expert all the way up and down, even the software stack, much less the hardware stack. And as a result what Ross does is he lets you be good at whatever you’re good at, right. So if you’re good at robot perceptions, then you can get all the rest of the stack, the lower level stuff for free and just add your own perception stuff and have a working system that emphasizes what you’ve done, but the rest of the system at least works. Somebody else might come in and say, oh I’m really good at motion planning and I’m going to focus on that aspect and so they are going to use you know the off the shelf perception stuff that’s part of Ross and they’ll get something that also works and it will be different in another way and so there’s an opportunity to basically get going and do your first prototype. And if you’re trying to grow a business or get a PhD thesis, either way right, you can’t solve the whole stack before you get going. So if you’re trying to grow a business, you can get a working prototype really fast and see if your customers like it and see what areas you need to most improve it and then focus on those areas in order to get better. If you’re doing a PhD thesis, you can have a working system that turns over and you improve the part that is going to be your thesis and the rest of it just is there to show off. So it’s a really cool as kind of a scaffolding that you can then add depth to.
Dave Kruse: Yeah. You know that’s super cool and part of the reason why probably robotics well, it’s taken off, but it will be more so in the next few years. Well there’s probably different reasons which I’d like to ask about that too, but before we jump into how Relay kind of came about in Savioke, I always like to ask and especially with someone like with your background. When you were growing up, did you like building stuff? Were you curious or when did you get involved with building stuff?
Steve Cousins: So I almost happened to grow up just at the beginning of the personal computer revolution. So I came out of – I was in high school, I missed punch cards by one year in high school and I missed punch cards by one year in college, right. So I mean this is a good thing, because punch cards surely slow you down. I did actually get to experience paper take for a year or so in high school. So this – I guess I’m beating myself. But the point is, I had this opportunity to do automation, but not so much robotic automation. I was able to do for example – I got word processors in the early days and so I was like Oh, this really does save me time. I don’t have to retype my draft of my paper. I can just fix the errors and print it again and get a perfect draft out. So that was – that sounds really mundane today, but it was a big deal back in the day. And the first robot that I really got to play with was a robot arm in college where I wanted to make the robot write, like draw characters, draw picture. The problem was it was a really shaky cable controlled robot arm, not very precise and it didn’t have any way to put force control and so I quickly experienced the limitations of early robotics and so at that point I just you know didn’t do more with robotics for a while. Then I got to Stanford and there was a robotics class and a lab, so I took that and in that class we were trying to make this robot navigate from Point A to Point B, almost exactly what Relay does, except that it didn’t have the sensors that we have today. It had like a ring of 16 sonars and if you think about trying to navigate with that, you know we actually got it to go down the hall and turn left at the intersection and kind of stop roughly in front of a door, but you know we were within, I don’t know precede accuracy or something and we got lost all the time. So you know it really makes – and so at that time it was like well, there is no way we can pursue this. But now you can see that its – we’ve come a long way in essentially the last 20 years to where you could give lots of robots in both indoor and outdoor, in autonomous cars and really the field is popping because the technology has kind of all come together and we’ve had some as I’d say curves cross, right. The ability to process data has surpassed our ability to collect data, right. You had video cameras and the ability to digitize them you know 20 years ago, but you couldn’t probably do it in real time for a low cost. You could probably do it for a lot of money in real time, but then you couldn’t process those in any way and today you can get you know like an intel real sense for you know $100 and you can process it with a cell phone if you want to and process the data in your cell phone camera in real time, right. And so its all-in-one package for a few hundred bucks and that’s like completely game changing thing that you may not realize unless you knew what it was like before.
Dave Kruse: Definitely. And so let’s get into Relay a little bit and it sounds like you almost had that idea a long time ago, but I was going to ask, how did you have idea and technology come about for Relay?
Steve Cousins: So you know Relay, it’s kind of a direct percent in the thinking from what we did at Willow Garage. So we started with you know a two armed mobile robot and it was able to be around people. We could navigate around people. It had arms that were safe to be around and so we had these robots for somewhere around 2008 to I guess 2013. So for about five years we had these robots running around in the lab and we were living with them. I mean literally moving around, they are moving in our space, until we said, okay, this is cool. We kind of got navigation end of it solved. We can go from Point A to Point B reliably, but the kind of tasks that we wanted to do with them, you know we had at one point a really cool demo where the robot would bring you a beer and it actually opened the refrigerator, reached in, picked out the right beer and then brought it to you. But one of the problems that we had to solve for this was how are you going to hold the beer, right and if you think about – and I mean this is a fun technical thing. When the robot is holding the beer, like its arms are in front of it, holding the beer like you would as you walk with it, the robot won’t go. Like why won’t it go? It’s because there’s something floating in front of its sensors, it’s the beer, right. Because the beer is in its hand and you naturally recognize that oh, that’s something I picked up and I’m going to ignore it, but the robot initially says hey, there’s something in front of me. I better not go, because it’s not safe to bump into things. It doesn’t realize that it’s holding the beer, right. So it’s all very condimental common sense stuff. So for that project we ended up putting a cup holder on the robot, so the robot can put the beer in the cup holder and then drive and then pick it out of the cup holder and give it back. But that kind of is the seed of the idea that you know, it’s really important to have a compass sometimes on the robot. Some way to – some place to put things that are designed for that, so that the robot can do its job which is navigating and then the stuff that your delivering is kind of out of the way of the sensors and secure the way. And as we moved from PR2 through a series of robot prototypes, we went from one that could basically pick up a box from a table or on the floor and carry it someplace else and put it down onto a table, onto the floor, onto a shelf and that prototype was pretty cool, because you say Oh, you can automate logistics. Kind of you can move things around, isn’t that cool. We simplified the problem by not trying to pick up every arbitrary shape, but instead just pick up these boxes and that was a good idea. But then we realized as we started at Savioke, you don’t actually need to pick up boxes and put them down. Most of the time or at least half the time we saw, there is somebody at both ends of the delivery, which means that all you need to do is really let somebody put something in, do the work of you know riding the elevator and walking down the hall and then provide it to the person at the other end. So that’s kind of the transition from you know two arms which is very human like to something which simplifies the problem a little bit, which we could actuate a little bit more to something that simplifies the problem even more and that we can actually deploy and put out into the world.
Dave Kruse: So you had this interesting tech. How did you decide to focus on hotels? Did somebody come up with the idea or the hotel approach you?
Steve Cousins: So we had a ha-ha moment at Willow Garage, where we were arguing about should we be in factories or should we be in homes and back and forth, factories homes, factories homes and there was this ah-ah moment where I said you know what, there is something in the middle, right. There’s places that – you know a hotel is not exactly a factory, but it’s got a lot more repeating tasks than the home does, right. So in the home you wash your dishes. In the hotel you have a restaurant and you do lots of dishes, right. In the home you make a bed; in the hotel you have you know hundreds of beds and so it has some of the repeating characteristics of tasks in the factory, but it’s almost untouched by automation, at least certainly by robotics and so we said you know, the service industry, hotel has been one part of, is the really untapped area for automation and if we can really move basically around people, we are going to distinguish from what happens in the factories. In the factories you always put the robots behind a cage, so that they can move fast and they don’t have to worry about safety. But let’s put safety and say, you know what, we think we can make robots move safely around people, which means that we open up the whole world of businesses, the whole you know selection of businesses that could use automation and we looked at hotels and said hey, there’s 55,000 hotels in the U.S. They tend to have a kind of regular structure inside and you know we think we’d be able to navigate really well in them if we can solve this elevator problem and so we went to work and got the robots who could ride the elevator and added that to what we already knew how to do well, which was navigate indoors and put together the solution.
Dave Kruse: Yes, and that was my next question. What was one of the bigger challenges moving to hotels, was it the elevator, trying to figure that out?
Steve Cousins: Yes, I mean I think that the biggest challenge for any – for robotics today in general is getting the use case right. So how do you find the use case where you can do it from end to end and you can handle all the things that come up, right. So I mean you see I think there’s been a bunch of robots lately that are designed to go down the side walk, right and that’s fine as long as you can deal with all the things that happen on the sidewalk, right, and there’s a wide range from weather to cracks to you know somebody parked their bicycle there to you know whatever. There’s lots of special cases that come up and we ran into the same thing in hotels. You know we tried to navigate through the hotel, you run into housekeeping carts. Not literally run into them, you encounter them and go around them or you get blocked by them right and you run into, you know somebody left the vacuum cleaner in the hall or there is a tray on the floor with breakfast or you know there’s any number of things that you have to deal with. But it’s not an incident number, right. There is a – or at least there is a very, very long tail. So you can handle 99% of the cases with the sufficient amount of engineering and then there’s always something else that comes up and you have to be able to handle that or at least recognize that you can handle it and ask for help and that’s kind of what we do.
Dave Kruse: And how do you guys deal with kind of the more of the social aspect, the social engineering, so that people – you designed a beautiful robot, but how do you make it so people weren’t freaked out by a robot delivering it. I mean like I would love it, but yeah.
Steve Cousins: When we do our best job at designs, one of the things we argued about was whether we’re going to have eyes on the screen or not, whether we’re going to humanize it a little bit. We don’t want to pretend like it’s more intelligent than it is. On the other hand you know it’s nice to make it cute and friendly, but you know it’s a combination of design, but also one of the key elements of good design is to go off and test things. So you know when we were at Willow Garage and we were talking about putting robots in elder care, one of the big questions was, are older adults going to be comfortable having robots or are they going to be freaked out and we didn’t know the answer. So what we did is we brought robots to the elder care facilities and we saw how people reacted to them and we continue to do that with Relay, all throughout its development in order to get the kind of feedback over what people like and what people don’t like. We added a motor system that isn’t strictly necessary, but the bin you know opens by itself and that was purely a usability trade off thing, right. If we said, if we do that, if we can make basically 100% of the time people know how to work it, right, because if you open the lid for them, everybody figures out that you should reach in, look inside and see what’s there. It’s very – it’s like so obvious. If you don’t automate the bin, you have to somehow communicate to people that they have to open the bin, right and it turns out we tried that, because it was easier, but it was – there was a certain failure rate. Certain percentage of people just wouldn’t get it and so you don’t do that, right, because you have to do – you have to build for your use case. Our use case was there are first time to first approximation. Everybody is a first time robot user when they get a hotel delivery and so you know we have to deal with lots and basically everybody is a first time user, so how do you make it as easy as possible and so that’s the design. It turned out that the combination of all the little design elements that we did in testing, we got it so people really liked it and now we see – what was amazing to us is that the guest experience in the hotel and hotels always measure the guest experience indicators and we’re seeing measurable bumps in all the guest experience because of the robot, which is really cool. You know we designed it to help out the staff, but it turns out that even more than that, we are delighting the guests.
Dave Kruse: Yes, well I can see and tech is becoming so much part of the culture and you were saying that it’s almost expected and when you got something unique and fun like that, yes I would stay at a hotel just to experience it. So I can see what you do.
Steve Cousins: Yes, me too.
Dave Kruse: Well, we’re almost out of time here and I’ve got many questions, but I am curious, what are some – I guess a couple, just a couple of questions. But what additional use cases do you want to focus on or tell us for a while and what features do you want to add to it and are sensors and cameras already a part of the tool set. Just curious what else you might want to add in the future?
Steve Cousins: Well I think the big opportunity that we have before us now is that, you know there’s a lot of hotels and we’ve really done a good job of nailing the hotel guest room delivery use case, right. But the robot that we’ve built and the technology has always been something that you know we feel like its broader than hotels and inapplicability and ever since we’ve been talking about this publicly and putting robots out in hotels, we’ve seen people – have people calling us and saying, hey, I can use it for this and I can use it for that. So you know the next big step for us is this becomes a platform that you can use for delivery in any service industry application, basically in any indoor facility as long as you don’t have to deal with stairs, which obviously this robot doesn’t do. But other than that we’ve got a solution that can take things from Point A to Point B in a building and whether you’re in a lab or you need to move samples around, whether in a hospital, whether in elder care, whether in hotels obviously or whether you’re in some other indoor facility, if you need stuff to move on a regular basis from Point A to Point B, we’ve got a great solution that works around people and that’s kind of what we’re focused on now.
Dave Kruse: Interesting. That’s exciting. That sounds like a big feature and that’s probably a good way to end this podcast. But I definitely appreciate your time Steve and it’s really interesting what you guys are doing and well, I’ll keep a close eye to see how things progress and where else do you take Relay into other industries. So I definitely appreciate it and good luck with things in the future.
Steve Cousins: Thanks very much Dave.
Dave Kruse: And thanks everyone for listening to another episode of Flyover Labs. We’ll see you next time.