September 28th, 2013

Transcript – Origami Robot

ROBOTS (Matthew Schroyer): Hi Nick, welcome to Robots. Introduce yourself to our listeners, and tell us a little bit about your background…

Nick Kohut: Sure, my name’s Nick Kohut. I recently received my PhD in Mechanical Engineering from the University of California at Berkeley, working in the Biomimetic Millisystems Lab on small robots. I also recently founded, along with three cofounders, Dash Robotics Inc., which aims to provide super fun and educational toy robots to kids and hobbyists. We also have some ideas for what small mobile robots can do in the future, in a broader range beyond the toy space.

ROBOTS: Excellent, so Dash Robotics specializes in small robots that are inspired by insect movements. Is that correct?

Nick Kohut: Yes. We’ve been working for several years now with biologists at UC Berkeley … most notably Robert Full, who has been studying animal motion for decades, and who is the preeminent authority on this topic. We found some interesting traits that all animals possess, and one is the springiness of their legs… their leg stiffness is about the same when you normalize it for their weight and size. If you get that right, you’ll have a robot that runs very fast and very efficiently, and if you get it wrong, you won’t. The other common trait is the pattern in which their legs move – their gait: it’s also shared whether the animals have eight legs, six, four or even two. That’s very interesting and something we’re able to emulate in the robot, which gives us a very fast machine, but also a very efficient machine.

ROBOTS: Could you tell us a little bit about the latest creation from Dash Robotics, the Dash Robot, which uses some of that technology?

Nick Kohut: This is our first product, and [it’s based on] a research version that was invented back in 2009 by my co-founder Paul Birkmeyer. Its construction method is very interesting: it’s made of a sandwich of run-of-the-mill poster board and 1/1000”-thick plastic, and it’s a flat fabrication method where you fold up the robot into a 3D shape – somewhat similar to origami – and this has a lot of advantages.

Its light weight is one big advantage, and you can also make things very small that you couldn’t with traditional methods. For example, a traditional joint with a ball bearing can only get so small before the friction overwhelms it, and our structure joints don’t have that problem. So we’re able to produce these 15-gram robots that run at 5 feet per second (15 body lengths per second) but give incredible battery life… our prototypes have around 45 minutes of battery life. We are not sure what the final spec will be as we add features, but that’s about almost two miles of running on one small battery charge.

ROBOTS: That’s quite a distance out of such a small robot, but the thing that struck me about this robot was its agility. I believe there is a video that shows the Dash Robot outmaneuvering a wheeled toy car. There are also videos of children playing with the robot, and it was really surprising that the robot is so agile despite having feet and legs. We have typically associated walking robots with being very slow. Could you speak a little bit about what the magic is, why is it able to run so fast in such a small package with legs?

Nick Kohut: That’s a fantastic question. The reason that you see most legged robots moving so slowly is because they have a lot of joints, a lot of degrees of freedom and a lot of sensing. To anthropomorphize, they’re thinking really hard about “Where do I put my foot next?” because they don’t want to fall. That requires a lot of computational power, requires all these motors and degrees of freedom, and makes everything really heavy and really slow.

We threw all that out and looked to nature. What animals do when they run very fast is that they’re essentially running open loop. What I mean by that is: they’re not thinking about where their foot goes next, they’re just thinking, “Run!” The dynamics of the system are so good – their legs are so well tuned to the surface – that they don’t need to think about it, so this makes them much faster.

Dash has a high power-to-weight ratio, for sure, but it’s the dynamic tuning that we do, refined over years in our PhD work, that allows a robot to run so quickly and so robustly over a variety of surfaces. We call it mechanical intelligence.

ROBOTS: That makes it very interesting and fun for children. Is this also an educational product?

Nick Kohut: Yes, this was a nice side effect of designing it to run like an animal. In the research lab we were interested in performance, but it turns out that people have a strong reaction when they see it run. Some people are delighted, some people are creeped out, but we think it’s great either way that people have this strong reaction.

Because it’s so small and low cost, it’s a great platform for educational efforts. It’s equipped with an Arduino board that we’re designing ourselves that has an integrated Bluetooth module and motor drivers. For a very low cost – it’s a $65 robot – you can learn to program, you can learn to build these robots, you can learn a lot about mechanical engineering and linkages, gears and drive trains, you can learn about electronics, and you can learn a little bit biology too. That’s a large educational space at a price that was previously impossible. A Lego Mindstorms kit is missing a lot of that, and it costs $300, so we’re a fraction of that price.

ROBOTS: You mentioned Arduino, which is a very important system to makers and people in the DIY movement. Can you explain what Arduino means, and tell us a bit about it?

Nick Kohut: When you talk about Arduino, a lot of people think of the Arduino Uno board, which is a standard Arduino board. It’s a very easy to use micro controller that lets you get things working very quickly … LEDs, servos, motors, things like that … but Arduino is also an open source language and standard, so that anyone can build their own Arduino board and use the programming language to very quickly build up robotics and other hardware related projects.

It’s a great trend, but one thing we think is missing is an integrated motor driver … you have to add a shield, which costs some money and is also a heavy and clunky add-on. So we’re developing an Arduino board that integrates that, which is great for small robotics. Arduino is a big enabler because it just makes everything so much easier.

ROBOTS: Could you explain what the motor driver is, and what it does for these systems?

Nick Kohut: The motor driver is an h-bridge chip … I’m actually not an electrical engineer, that’s the role of one of my partners, so forgive me if this isn’t 100% accurate … We’re going to use dual h-bridge drivers, and it’s able to drive the motor at one amp in either direction. If you want to do robotics and you want motors right off the bat, with an Arduino Uno board or a Digispark (which are both great products) it’s not as easy to drive a motor … you’d have to buy a separate module and snap it on.

This [motor driver] allows us to keep the board very small – we’re aiming for 2 x 1 inches – and to drive the motors in an integrated way. It’s a single board that lets you do most of the things that you would want to do in robotics.

ROBOTS: As part of this Arduino compatibility, there is also a dedicated smart phone app to drive it. Is that correct?

Nick Kohut: Yes, that’s something we’re developing as well. The app we’re planning is quite straightforward. It’s essentially a joystick for the robot: forward, backward, left and right, and hopefully adding some of the sensor functionality as well. A lot of people want a cool toy that they can run around and that’s fun to build, and that’s what the app will be great for.

But one of the big ideas here is that the robot is expandable. So you can write your own app for browsers and you can run it off any computer … if you’re really into it you could write for android. With the Arduino language and Bluetooth 4, it’s pretty easy to interface to.

ROBOTS: So the Arduino language makes it very easy to hack into expanded capabilities, and one of the things that was mentioned on the website, was the idea that you can add on sensors. How would a sensor help the robot be more functional in its environment?

Nick Kohut: There are a few sensors we’re going to include in the robot off the bat. One is a gyroscope, which will help the robot steer because it will tell you the turning rate of the robot. The other is a visible light sensor, so you can use it to follow a light or, if you want to make it even more cockroach-like, you can have it run away from lights. We’re planning on putting IR meters and sensors on, so you can program in behaviors like wall-following, and also some LEDs and some pins for expandability.

One cool idea was to add a touch sensor on the front and the back, and if you have two robots, you can play battle bots, where when you touch you can score points, and your LEDs can act like health meters … going from green to yellow to red.

ROBOTS: What kinds of lessons do you hope kids will take away from [working with] the Dash Robot?

Nick Kohut: That’s a great question. We’re actually looking for educators to help us develop a curriculum around Dash at this time. Some initial ideas include having different transmissions, so you could learn fractions based on gear ratios. Or you could teach physics by learning about the difference between dynamic and static friction: as Dash runs up an ever-increasing incline, he’ll stick to the incline when he’s not moving, but when he starts to run he’ll eventually run downhill, and that’s when you know that he’s in the slip range and there is dynamic friction. That’s probably more of a high school-level thing. You could teach biology and learn about how animals run …

There are a lot of things out there I think that you can learn. Not to mention of course, programming with the Arduino language on a fun platform, and all the cool mechanisms of Dash. He’s driven by a lot of four-bars and other classic mechanical engineering linkages that I think would be great for kids to learn about in the K-12 range instead of waiting for college.

ROBOTS: I’d like you to talk a little bit about the company Dash Robotics, which is a new startup. Could you tell us a little bit about how it got started, the kind of talent you’ve got, and where you’re based?

Nick Kohut: We’re based in Berkley, California, and there’s four of us: myself, Paul Birkmeyer, Kevin Peterson, and Andrew Gillies. We all worked in the same lab at Berkley on these small robots, and we were getting close to graduation and had started doing a lot of outreach events … going to schools and science museums, and showing them to people. We got a ton of questions like, “Can I buy this?” and “Can I have one?” and as we got closer to graduation we thought, “People really like this. We should see if we could turn this into a company.” We all believe in engineering education, and education in general, and we wanted to start something of our own.

ROBOTS: If a person wanted to obtain a Dash robot, how should they go about doing that?

Nick Kohut: The easiest way is to go to dashrobotics.com where we have a big green button that says, “Check out our campaign.” That’s for a crowd funding campaign – similar to Kickstarter – where you pledge money in exchange for a Dash, and that helps us get off the ground, covers some of our fixed costs, and in return you get a Dash. You get an awesome robot and we get a jumpstart to our company.

ROBOTS: Just looking at the crowd funding website, the initial goal was $64,000 and it’s already blown past that goal with 18 days still left. How long did it take to reach that goal, and what’s your reaction to the kind of support that’s come to the project?

Nick Kohut: The support has been fantastic. The goal took about five and a half days to reach. We’re really happy with that. Our goal from the outset has been to sell 1000 robots, and that’s to get feedback from customers and see what they do with it and what they like about it, without running into a lot of the manufacturing problems that you see in other crowd funding campaigns. A lot of crowd founding campaigns sell way more and have all kinds of delays, and we’re trying to do our best to avoid that by limiting our sales.

We don’t expect to go out and make a million dollars on this, but what we do hope is that people will get a Dash Robot, get involved and give us feedback such as “This is what I like about Dash”, “This is what I want to see on Dash in the future”, “This is how I’m using it”, “It’s good for my kids”, “I love it myself as a hobbyist” or “We could see this in a classroom.”

That’s the real goal of the campaign, and we have currently 78 robots left. They’re going pretty quick and we’ll be offering the standalone Arduino board with the integrated motor drivers as its own item shortly, probably sometime this coming week.

ROBOTS: You mentioned collaboration, and it’s always been a key component of the Arduino ecosystem that you can find solutions online and share them. Do you plan a dedicated community to trade ideas and solutions for the Dash robot?

Nick Kohut: We do. We’re planning on implementing what we’re going to call the “Dash Design Forum” where people can exchange ideas for what they can do with Dash, exchange code, that sort of thing. We hope to build a community behind this that will make Dash better for all users and allow us to support them better in the future.

ROBOTS: Where would you like to see Dash Robotics after the campaign finishes?

Nick Kohut: That’s a great question and it’s something that we’re using the campaign for to figure out. That’s part of that feedback [we’re looking for]: how do you use Dash? and how would you want to use Dash in the future?

We’ve had a lot of people say to us, “This is a really cool toy and I love it for my kids,” or, “I’m a maker and hobbyist and this could be really fun to hack on.” And we’ve had a lot of interest from educators who have told me, “We do robotics education for disadvantaged schools and we’re spending an unbelievable amount of money on Mindstorms kits, and it’s just not sustainable … what you have is a fraction of the price and that would help us immensely.” We think that’s amazing.

There are also other possible applications, such as mobile sensors: you have this robot that’s very light and very small and very maneuverable. Let’s say there’s an earthquake and a building collapses, and you want to find people … your options right now are to send in human rescue workers (which is dangerous and expensive), or to send in a big clunky robot (and like you mentioned before, that’s heavy, and it has a lot of moving parts, and if any of those parts fail, your robot has failed and you’re out a million dollars).

What we see as a great alternative to that is a thousand robots similar to Dash that go into the collapsed building, where they can reach small spaces that people and larger robots can’t … and if a few of them fail, it’s okay. You really just have to have a few succeed out of that 1000, so you decentralize the failure and you lower the cost significantly. If we mounted CF2 sensors on these small robots and we found the people, now we can send the rescue workers to exactly where they need to be, which is safer and cheaper and better for everyone.

ROBOTS: So in a sense they’re disposable in that sort of environment or that kind of strategy…

Nick Kohut: Yes. A big part of the idea is that when you’re deploying 1,000 … 2,000 … 10,000 robots – which is reasonable with this technology – you’re not worried about a few failing. You can have 1000 failures and still do what you want to do.

ROBOTS: Nick, thank you very much for joining us.

Nick Kohut: Thanks Matt.

| More