Learning lessons from working with modular robots with Dave Rollinson

Dave Rollinson from Hebi Robotics talks about learning from watching his robots being used as everyday tools. We also get an initiated overview of the modular robotics systems that Hebi develops.

author avatar

03 Oct, 2019

Dave Rollinson from Hebi Robotics

Dave Rollinson from Hebi Robotics

Dave Rollinson from Hebi Robotics talks about learning from watching his robots being used as everyday tools. We also get an initiated overview of the modular robotics systems that Hebi develops. 

Dave shares how he had the opportunity to work with sewer inspection robots early on. He learned a lot from constantly bringing the robots out in the field. Seeing how the robots were used, as regular tools, and how they succeeded and failed helped him iterate the design and take the steps needed to create a well-functioning product. 

Dave also talks about how he wanted to continue to build robots as he continued his education. He chose to go to CMU and professor Howie Choset’s lab because they build their own hardware. 

We get the opportunity to see one of the modular robotics system Hebi is working on live and a few examples of how it is being used. At the end of the interview, Dave and the host Per, who is a major modular robotics fan, discuss the significance of modularity in robotics and how it can change how we develop and use robots and machines.

Interview:  Learning lessons from working with modular robots.

Per Sjöborg, host of the Robots In Depth podcast, interviews Dave Rollinson from HEBI Robotics to learn more about the modular robotics systems that they develop.  Below is a transcript of the interview. 



Per Sjöborg: Welcome to podcast version of Robots in Depth and this launch episode with David Rollinson in cooperation with Wevolver. Today, I’m really happy to have a modular robotics company here. We have Dave Rollinson from Hebi Robotics. We are even going to see his robot a little bit later. I want to start the way I always start. How did you get into robotics? Why did you choose that field?

Dave: I guess in some ways it was a bit of an accident. When I was in high school I liked tinkering mostly with mechanical things. I had just a ton of different interests. When I we to college I think I was just interested in engineering because I like to build stuff. I chose mechanical engineering I think in part because it’s so broad. It’s kind of a catch all for anything that doesn’t fit in kind of another field. The school I got into is Carnegie Mellon University in Pittsburgh. I started being exposed to robotics there. While there I got into some internships. I got really into robotics on a practical level. I interned and then later worked full time at a company called Red Zone Robotics and they got to make sewer inspection robots. I’ve got to get hands on and use robots that were both cutting edge and then just extremely vital in terms of what they’re doing, in terms of keeping infrastructure that nobody really thinks about. It’s really addressing a needed for that. In the robotics end it’s so cutting edge but where you’re at with sewer is so simple and so fundamental that you realize that you’re working on the next generation of tools to really just kind of keep bringing humanity forward.

Per: Also that we probably neglected both because it’s very hard and also it’s very expensive to maintain this infrastructure properly so we have big technical depth. It’s actually failing on those and we have on real plan on how to fix it fast enough.

Dave: It’s a massive challenge for the United States and it’s actually a massive challenge to just try to start to learn from that and in other places in the world where we’re still trying to build that infrastructure and do it. 

Per: Yes because we really want to get it right the first time there. You got into robotics that way. You studied at CMU. HEBI is a company that actually comes from Carnegie Mellon.

Dave: Correct. HEBI Robotics came out of the love that I kind of gravitated toward when I came back to CMU as a masters and then PhD student. I came from RedZone Robotics. I had this kind of practical, mechanical experience from making sewer robots.

Per: Very challenging environment. Those robot has to be as tough as nails.

Dave: Tough as nails and really kind of a fascinating lesson on the way that people use tools too. You spend weeks and weeks designing parts. You view it as this robot and you take it out to the field and these guys just abuse it because to them and correctly so it’s just a tool. You realize that the way you put things out in the world they’re going to get used in the way that other people see fit. It’s a real lesson of watching how things get used in the real world and so part of it is the people. Part of it is the situation and then the onus is on you as an engineer to just make it all work.

Per: It has to be really world ready, PhD ready. They’re not going to baby it. It’s also hard to exactly estimate the world it’s going to be so you have to over design it.

Dave: You over design it and just have to plan ahead of getting out into the real world, watching everything go wrong and then iterate, just constantly be learning. The toolset that I needed went far beyond just the mechanicals on. I need to know more about the controls, more about the sensing and kind of the fast trends that are evolving there that I wanted to bring towards robotics. I went to the snake robot lab with Professor Howie Choset at Carnegie Mellon in large part because they’re one of the labs at CMU that made their own hardware. As a mechanical designer I knew that was something where I can continue to develop those tools and as a focus coming back into PhD I really wanted and Howie advice that I really have a focus to target things. I said snake robots. In a lot of ways pipes are a simple world to work in. They have a lot of challenges from an environment standpoint but not a lot of ways. You can think of it as a one dimensional world. You can only go forward or back, turn around. It simplifies a lot of the assumptions and it makes it a great starting point for thinking about how you control a complex robot.

"You over design it and just have to plan ahead of getting out into the real world, watching everything go wrong and then iterate, just constantly be learning"

Per: They had worked on the snake module for a long time.

Dave: Yes, the snake robot project had already been going for over 10 years, probably close to 15.

Per: It’s quite a mature project. What number of iterations have they gone through? Is that hard to specific?

Dave: It is hard to specify. We have a bunch of them up on our wall. What is really interesting is it kind of started off with hobby servos being chained together and then eventually they developed enough to where we had a controller on the tail to control a set of 15 servos? By the time that I got to the lab in 2009 they had really done an iteration where they had started to distribute the control into each module. The folks that now started to go into more modularity really just in the name of being to maintain these things. In addition to being able to experiment with making robots shorter, longer. For years and years it was mostly an undergraduate project and then every three months, every six months. They would make a robot mostly because they would just get destroyed in the process of using. They just weren’t very reliable. When I came we were taking the step of moving from kind of hobby servo technology to getting a better funded project where we can design a really solid module from the ground up so metal housing, nice Maxon motors, really better performance. Instead of going to an iteration every three months also that robot about three and a half, four years. It’s still used today.

Per: Very big steps, we don’t even know how long it’s going to last because it’s still used.

Dave: It’s still cranking. There was that big step and then towards the end of my research around 2000, in the middle of my research around 2011 or so DARPA project started out and we did the next big iteration of a snake robot with moving to the latest technology that is coming from the cell phone industry so being able to put things like Ethernet, much higher powered processing, many more embedded sensors into the module and then as well as moving to being able to do torque control not just position control.

Per: We’ve seen this six legged spider thing that you have built with it. It looks very competent and very robust. This is a powerful machine.

Dave: It’s a powerful set of tools that lets us build the machine.

Per: I presume that when you have done the module doing the actual spider is reasonably easy achievable task because so much is in the module.

Dave: It is. Basically what we’re trying ot do is kind of abstract an actuator down to the point where you’re really thinking in terms of positions, velocities and forces or torques. The math that we think of modelling a robot, all the low level details kind of get balled up and then you can just kind of work at a higher level.

Per: Can you talk a little bit about that robot? How strong is it? What kind of payload can you put on it? What kind of speeds can it achieve?

Dave: That snake monster, we call it snake monster because we kind of have a vision as snakes like hey, we could turn this into arms. People call the spider robot much more like a spider but in terms of the speeds each joint has about four Newton meters continuous torque, peak of 70 Nm that lets us build legs in that configuration that could support about a three kilo…

Per: Per leg?

Dave: Total. It depends on the stance but kind of from a practical turnaround. In terms of speed it’s about maybe a half meter per second if you kind of go low and go forward. The research both on the snake and in that robot is once you have it together you can create some basic controllers but it’s really kind of exploiting the design space of how you make it move. It’s kind of where the games are to be made.

Per: This is built with the research version of the module that you have then but you also market that.

Dave: Yes.

Per: That is available for anyone that wants to work with a very high quality, small and compact module?

Dave: That is exactly right. What we did is we started the company two years ago to basically take the modular technology from the snakes and start using that as a building block for other types of robots. We started off with exactly those snake modules. We brought them out. We said, we use these. We think they’re great. See if anybody else wants to do it. We quite frankly put it out there. We met a year ago at ICRA and we were trying to sell those. Pretty much nobody bought them but we got a lot of feedback saying this is great but it’s not the right form factor for my particular robot or it has not enough torque or not enough speed. I really like continuous rotation. You have several other things. We basically took all that and did an iteration now where it’s a much more general purpose building block, all the things they did like in terms of being able to abstract out all of this control so that they can really think at a high level, play with the latest tools like ROS and Matlab and all these other great software ecosystems that are out there. I kept all of that and then made it to kind of a nice fundamental building block.

Per: Took all the experience that you’ve had from this snake module and also probably took a step away from the snake because that puts out requirements on its size and design. Now you could think more freely and use all the experience you have and kind of take a new step.

Dave: That is exactly right.

Per: That’s also what you’re marketing right now? 

Dave: Yes.

Per: When I saw this one, you know I like modular robotics. I love the ability to kind of have a smart Lego for adults, really robust. You can really build some stuff with this.

Dave: It’s cliché but Lego really is the inspiration. What is great about Legos is you look at them you know how they go together. Even like the youngest kid sees them and the first time you press them together like boom, I get how it works. You make anything. You make spaceships, you make cars, you make castles and then your imagination fills in the rest and it all just works. The grownup part of Legos is okay, that imagination part we actually have to put something in there. That is where the massive amount of work that we put in to the software comes in. 

"What is great about Legos is you look at them you know how they go together. Even like the youngest kid sees them and the first time you press them together like boom, I get how it works"

Per: This unit is really solid, really strong, you can really build some practical machines with this. That is why I like it so much. As you say still abstracting a lot of the hard stuff doing that. We’re going to see it later. We’re going to have a look at it. You’ll have to talk us through why you designed it like that and what trade-offs you did. I love the modular robotics stuff. I want it out there in market. You mentioned that when you started to sell the first type of unit. It was a tough sell. With the new unit do you see that people are getting the modularity power, the power of the modular concept?

Dave: Yes, people are getting the power of the modularity. They’re getting the fact that you can create anything out of these. Not just arms, not just leg robots, you could have disembodied degrees of freedom all around the room and just kind of automate a room in pieces if you want. They’re getting the fact that because it has Ethernet connectivity you have easy access to both controlling it and then logging all the rich amounts of data now that machine learning is becoming much more of kind of a widespread tool that people apply. People want as much data as they can get.

Per: Have you heard anything about what people would like to build with when you talk to them?

Dave: We’ve heard walking robots, humanoids, some snake robot people still out there. An interesting one that terrified me at first was somebody looking to create a dentistry robot. I was just whew, worried about that but it was actually to get a good position around somebody’s face because dentists have a lot of neck strain from reaching around people. It’s one that embraces another part of it not just the modularity but the safety aspect, being able to control torque not just position so that as you interact with a very uncertain world you can be in a very compliant way.

Per: This is then co-robot safe. Humans and robots built with this system could operate together rather than have the robot in a cage.

Dave: That is exactly right. 

Per: How do you see the future for the company? Where are you taking it?

Dave: Basically we’re trying to just accelerate the pace of robotics innovations. We want to get these modules, keep them as robust as possible and give the tools to as many hands as possible and then branch out from kind of research R and D into some industrial applications. We think there’s a whole world out there that we kind of slip in between the giant kind of general purpose arms which are becoming more and more capable but still very general purpose and then the world of fixed automation if you’re going ot make a million of some widget you can make kind of a precise fixed machine for that kind of there’s this whole space in between both in the factory and I think even increasingly at home.

Per: But even the very specialized machine doing paper clips and doing that for a hundred years still benefits from modularity because when you want to replace something you don’t sit there with something special that nobody knows how to maintain. It’s a module. It’s the same everywhere.

Dave: Or a big part of it is instead of producing extremely high quantities in one place you can create smaller quantities in a much more distributed fashion.

Per: That was very interesting. We’re going to bring in the robot now. You’re going to have to walk us through the different aspects to it. We have the robot in front of us here. I hope you see it perfectly on the image. If you hold the robot around here, the single unit we’ll see it. This is the new unit then.

Dave: This is it.

Per: It has a different shape.

Dave: It does.

Per: Is there a reason for the different shape?

Dave: The different shape is because it’s designed for essentially everything except a snake robot which is probably the larger domain. The snake robot is a great forcing function in terms of driving a lot of requirements, in terms of being compact in our motor control, being very high power to weight so exploiting every last ounce of performance out of an electric motor, making very unique designs for the way that you do your torque sensing inside a module but then what it then does is it gives you a chance to then take a step back and say okay, with those core ideas how do I repackage this into something that is more affordable, more manufacturable and then just kind of much more versatile for robotics community in general. The snake modules that you saw last year were basically 1 degree of freedom. This is also 1 degree of freedom but they connect on the front and back with this custom interface to make things nice and sleek and sealed and as lightweight as possible.

Per: A snake basically.

Dave: A snake and only plus or minus 90 degrees of rotation. This keeps roughly the same size and power class but we have continuous rotation at the output. We have moved to using (0:15:56.9) aluminium and flatter machine parts to make the manufacture more easy. We have moved to using standard connections, just RJ45s and Molex mini fit junior for power. We have given a lot of thought ot just the flat form factor to keep things out of the way, big bolt pattern to mount it nice and securely and a through bore so that all the wire as you chain these things together can pass through the center rotation. Instead of having big service loops on the outside, telemetry rotation you can turn more than 360 degrees if you need. You can run pneumatic tubes through the middle if you need to put some sort of vacuum or grip around the front of an arm, all sorts of things.

Per: Very cool and I also think that this module is easier to produce so it’s quite a lot cheaper.

Dave: It is still I’ll admit not cheap so we’re at $3000 per degree of freedom right now but it’s significantly easier to manufacturer and less expensive than the snake modules for that.

Per: I also guess it has advantages, all of the snake model but it’s the same strength and speed.

Dave: It’s a little bit stronger and it’s available in different torques and speed profiles. One of the things about the versatility of the design is we’re using a spur gear train and on some of the stages we can change them out so that we can have a high torque low speed module as well as a low torque high speed.

Per: You can choose the one you need.

Dave: Example if you’re putting a manipulator you can use the torque modules that the shoulder. You also don’t need to move as fast but the wrist can be nice and dexterous.

Per: You can hook it up to Ethernet and power and you’re ready to go.

Dave: Exactly right. An idea that came from the snacks but implemented another way was we wanted to be able to daisy chain our communications. You plug in the power. It’s easier to split power outside of a module and your requirements on that different by application but the data connection just daisy chains on through as much as you want and just runs on a standard network hardware.

Per: To build the solution we have in front of us here it’s a robotic arm. You just connect this together. Do you also manufacture and sell the brackets or stuff like that?

Dave: People can connect this however they want but we do make these brackets that just bolt on to standard tubing. You can have it any orientation you want, any length you want and then part of the software tools we have lets you configure this in a flexible way.

Per: What kind of applications have you used to test this on? Have you got a pet application that you use as a test?

Dave: We have a couple. We have made six degree of freedom arms to do like pipe assembly, stuff like that. This is driven by a potential industrial application where we wanted to do some assisted assembly of screw. Because we’re at the right scale and we’re really interested in kind of dogfooding our developments on a mechanical level we’re basically creating a demonstration where we can assist in the assembly of the screws of our own modules because we can cooperate at roughly that scale and that precision.

Per: I’ve seen where you demonstrate the assembly of one of these arms in about 25 minutes. Most structures would probably be built in a few hours or a few days and of course they could also be disassembled in a few days. You could have students come in, propose a project, build a prototype, test it, realize that I probably have to go back to the math. Disassemble the robot, somebody else comes in and uses the hardware while this student is away working on the code comes in and reserves time like we did on the old mainframe, reserves time, gets to build the second version of the robot, learns even more and this speed with a threshold entry and a speed that is unheard of before with any other system.

Dave: That is exactly right.

Per: Do you have any samples of stuff like that within your lab and with your customers?

Dave: Absolutely. Our lab, the modules that were built for the snack and both these, they’re being used to just make other robots just kind of as a fun experiment. We have had cases where people come in, take modules that were meant for the snake robot, build a biped, test balancing, take them right apart, put them right back on the shelf. That all happens in a space of a day. You’re not starting from scratch. What we really want is if you go to most research universities including ourselves you can go into some sort of basement and you can see a pile of robots no longer being used.

Per: That is just a pile of money wasting.

Dave: We really want to be able to make these go back to the shelf as opposed to the basement.

Per: Also students could work together. I create my group of modules doing a leg for instance and then somebody else does an arm and somebody does a waist and suddenly you have an ant.

Dave: It goes back to that community thing. You want to get people working together on different projects, same projects, different projects that through a wonderful accident because the same project.

Per: That’s very nice. Thank you very much for taking the time to do an interview. It’s been a pleasure to talk to you. I know this is a foundation for a new world.

Dave: This has been fantastic. Thank you.

Per: Thank you very much. I hoped you like 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 helps 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 you host Per Sjöbor.

END OF TRANSCRIPT


Follow the podcast on Apple, and Spotify.


The Wevolver Robots in Depth podcast is published multiple times per week. If you, or your company is interested in supporting this podcast and reach an audience of professional engineers and engineering students, please contact us at richard [at] wevolver.com

03 Oct, 2019

I interview CEOs, founders and CTOs from successful and up and coming robotics companies, as well as researchers, investors and thought leaders within robotics and AI. You'll learn about the latest technologies that are being developed at innovative robotics companies, the lessons learned by incredi... learn more

Stay Informed, Be Inspired

Wevolver’s free newsletter delivers the highlights of our award winning articles weekly to your inbox.