Paul Birkmeyer, EECS at UC Berkeley, talks about his work in the Biomimetic Millisystems Lab designing and building robots. The Lab seeks to harness features of locomotion, actuation, mechanics, and control strategies to improve millirobot capabilities.
Transcript
Speaker 1: [inaudible] [inaudible]. Welcome to spectrum
Speaker 2: the science and technology show [00:00:30] on k a l x Berkeley, a biweekly 30 minute program with interviews featuring bay area scientists and technologists, a calendar of local events and news. My name is Brad swift and I'm the host of today's show. Today's interview is with Paul Burke Meyer, a phd candidate in the electrical engineering and computer science department known as Ekes. He is working with Professor Ron fearing in his biomimetic millis systems lab building six legged crawling and climbing robots. [00:01:00] The goal of the biomimetic Miller systems lab is to harness features of animal manipulations, locomotion, sensing actuation, mechanics, dynamics and control strategies to radically improve Miller robot capabilities. Miller robots are small robots. For instance, the robot Paul Burke Meyer has built named dash is 10 centimeters long, five centimeters wide and weighs 15 grams. This interview [00:01:30] is prerecorded and edited. Welcome to spectrum Paul Burke. Myer, thanks for coming.
Speaker 3: Yeah. Thank you for having me. It's a pleasure to be here. Where are you situated at cal? What's your current status there? I am pursuing my phd here. I'm entering into my fifth year actually. Uh, and I'm studying Ekes specifically electrical engineering and I'm working on robotics in the w department. So
Speaker 2: are you in a specific group with any x or is [00:02:00] it just a general study thing? No, it's gotta be something more specific for a Ph d
Speaker 3: it is. So, uh, I've been working with Professor Ron fearing since I arrived and he runs the biomimetic Milly systems lab. And within that he has a few different projects, but specifically I'm working on a sort of six legged crawling and climbing robots. Describe for us the robots you're building that my goal for my phd when I first came and still true is to make [00:02:30] a robot that can dynamically climb up a any sort of surface that it's presented with. So the contribution I'm trying to make is how do you make a robot that's minimally actuated? So class uses only a single actuator right now, single motor to drive all the legs. How do you create something that is passively stable? So the structure itself makes it stable when it's climbing. So you don't actually have to spend extra computation and have extra motors on there to keep you from either [00:03:00] falling off the wall or turning and things like that.
Speaker 3: Um, how can you climb dynamically, not this sort of slow plodding climbing. How can you climb dynamically, rapidly up a surface and do it stable and do it with very little effort. And what does the foot look like that allows you to make a robot like that. So what does your foot need to do in order to be able to engage and disengage rapidly and without any actuation? So that's [00:03:30] sort of what my phd will say in the end, hopefully. And maybe a year and a half or two years. How did you go about building that kind of a robot?
Speaker 3: So the design was long and hard. Um, so when I first came to the biomimetic Mullin systems lab, they were already using what they're calling the smart composite manufacturing process, if you want to describe it. Yeah. So the original process was taking [00:04:00] two pieces of carbon fiber and cutting mirrored slits in both. You cut a bunch of slits on the one piece and you mere it across to the other, and then you take a piece of thin Palmer thin plastic sheet and then you take those two mirrored pieces and put them together and make a sandwich structure. And so you have carbon fiber with one pattern polymer, and then the other piece of carbon fiber with the same pattern that now aligns with the other one, it [00:04:30] bends. Now it's flexible at those polymer hinges at those where those slits were originally. So if each slit is a joint, it doesn't cost you anything to cut more joints out.
Speaker 3: Whereas if you're making sort of traditional machined robot out of say aluminum and ball-bearings and things, each new joint does a new bearing, which has some costs, has extra weight. So you can add many, many joints. For example, Dash I think has 75 or more joints in [00:05:00] the robot. Um, many of them are fixed, so they're used just to fold up the final structure and then you glue them in place. Each hip has six moving joints. So each hip has six moving joints. They're six hips. So Justin, the hips alone, they're already 36 moving joints. Um, whereas if you were to do this with ball-bearings, you quickly get something very big and very heavy. So this actually started off as a prototyping process. [00:05:30] Before they would use the carbon fiber process to make their robots. At the time they were making very small robotic flies and you have to assemble these flies under microscopes and it's very tedious.
Speaker 3: And if you, if you mess up, so in your design process, you didn't account for something or something doesn't quite align. You've lost a couple of days just working under a microscope, your back hurts, your eyes are tired and it's very frustrating. They realize, hey, this is just a geometric [00:06:00] pattern. So if we make it very small, little fold up the exact same way as if we make it very big, the pattern is the same, the folds are the same. So they take cardboard and make the pattern just bigger and then assemble it by hand without a microscope within a few hours. And exactly, they can tell it's gonna move in the way I want. So this started off as a prototyping process designed by, uh, Aaron Hoover, who's now a professor at Olin and he just graduated. So I actually took this process and started to make [00:06:30] robot designs and realized, Hey, these are actually very functional.
Speaker 3: They don't have to be prototypes necessarily. They're actually functional robots at the end. And uh, the cardboard was used, it's cheap cuts very quickly on a laser and you can go through designs very quickly. So instead of having one design that takes two days to build, you can build one in an hour or two. And so you can sort of explore that design space very quickly. So coming into the lab, they were using this manufacturing process where you design everything flat and you cut it out with the laser and you have to fold [00:07:00] it up into something that is functional and moves in the way that you want. And at the time, and still true, we don't have any good way of mapping what a 2d pattern is in the laser cutter, what that map looks like. And what you'll get out when you fold it up into three dimensions.
Speaker 3: Keeping in mind that these joints can't spin 360 degrees like a ball bearing. They're limited to at most 180 [00:07:30] degrees before they hit the link on the other side. So you have to in your in your head or on paper draw these structures. Say I started with hips, how can I get a nice leg motion out? And so I designed the hips and then like extrapolated that to six hips and sort of as you go you have to sort of mentally unfold these hips and figure out what that pattern looks like and then you put six hips and then you have to make sure that it can all fit on a flat piece and that when you unfolded [00:08:00] they don't have pieces that are unfolded on top of each other. As you go. Say you'll make a pattern and the first one you make, you fold it up and you realize that some part has to go through another part because the way you designed it actually you didn't realize this part was going to fold into the other cause you have to go back and redesign it.
Speaker 3: A lot of trial and error, a lot of trial and error and it took more than 50, maybe, maybe less than a hundred different design iterations for the dash that is [00:08:30] published now from where I started. And even then there were some designs I did with just a single hip just to see what a good hip design was. And it took a lot of time just to get familiar with this folding and unfolding process and laying out parts in two dimensions. And that took me six months just to get familiar with that when I first came. So, so dash is made out of this paper composite. Um, but I've made Balsa wood versions, [00:09:00] I've made fiberglass versions. I actually have not made carbon fiber just because our laser that we use to cut carbon fiber, the bed is not quite big enough so you can't cut pieces quite big enough to make dash. But now we have actually a new laser that I, I will probably pursue carbon fiber if only for the novelty. Um, so it was a, it was a long process.
Speaker 4: [inaudible] you are listening to spectrum [00:09:30] line a l x Berkeley. You're talking with Paul Burke Meyer about designing and building small six legged crawling and climbing robots.
Speaker 2: The robot that you've built and published a paper about is called Dash. What does that stand for? Dash stands for the dynamic autonomous sprawled hexapod. Once you'd spent a lot of time with Dash, you then wanted [00:10:00] to create an x generation. What was it out of dash that you wanted to explore with clash?
Speaker 3: So the things I liked about dash were the fact that it was still fairly small, 10 centimeters long, only 15 grams and very powerful. So if I kept it attached to a wall so it couldn't fall backwards off the wall, it had a lot of power. Could accelerate to full speed within a few hundred milliseconds. I mean it was very, very powerful. So that was nice. But its failure [00:10:30] was in the fact that in order to run it has these two plates basically that move up and down and forward and back relative to each other to drive the legs. That's basically the body is the transmission and it's true, the transmission is moving up and down. And so that's actually the problem is that it's pushing itself off the wall and it does this. So that was the, the main thing I wanted to address, but I liked the way the legs moved.
Speaker 3: They call it alternating tripod gait where you have three legs in contact of any one time, so you have this [00:11:00] sort of tripod of support. So I knew what I had generally that worked and I knew sort of what didn't work. And so with clash it was how do I get rid of this up and down motion? And I'd spent enough years doing this smart composite manufacturing that the transition from dash to an entirely new design was only a couple iterations before I got something that actually climbed rather than multiple 50 or so iterations. So that was a lot smoother. The hips are essentially the same, but though the way that they're driven is a little bit different. [00:11:30] And now instead of moving up and down, it's sort of moving side to side and forward and back. So it's not pushing itself off the wall.
Speaker 2: Can you describe the control systems you use for your robots? So the, the
Speaker 3: interesting thing with the robots that we're making in our lab is that we're trying to reduce the amount of controls necessary as much as possible. Traditional robots, heavy computational power, um, so that they can control each limb and very precisely so in, in, or wants, they don't fall over. [00:12:00] Basically the biggest problem is not falling over for, for legged robots and maintaining stability at least traditionally. So what we're trying to do is to minimize the amount of overhead you have to have, just to be functional. So we've worked with biologists here at Berkeley. They've sort of found these really interesting properties and cockroaches where if they're running over smooth terrain, if you measure their, uh, leg muscle activity, it follows some very repeatable pattern [00:12:30] over smooth terrain, meaning that they're, they're activating the legs the same and then they give them this very rough, varied terrain with bumps, maybe two or three times the height of the cockroach.
Speaker 3: They're very significant and they measure the leg activity and it looks almost exactly the same as when it's running on flat terrain. So what that that said to them was the roach is basically saying run and it doesn't care what the terrain is. They've decided that there's this [00:13:00] mechanical complexity and compliance. So the legs basically act as shock absorbers. They're just running and the legs sort of compensate for any roughness in the terrain. What we're trying to do is basically have a robot that does that where you just tell the robot to run and it doesn't care what it hits or what it's running over. It just basically runs and the legs are soft enough and bend enough to sort of compensate forever variation. There isn't the terrain. So the first design of dash that actually [00:13:30] put a motor in the motor actually came from a radio shack toy and I just took the electronics from that toy because it was remote controlled.
Speaker 3: Since then, the electronics have been swapped for custom electronics. A couple other students in our lab have designed really small lightweight electronics with an accelerometer and a gyroscope, even a port for uh, integrating a cell phone camera and there students who are using that cell phone camera to sort of [00:14:00] guide the robot from my end. I'm basically doing the robot design and I put these electronics on and I have two commands, three really run. And I tell it how fast and turn left or turn right. And that's it. The nice thing is you don't have to do anything more than that because it, it, it runs well and it can go over a different terrain. It can climb obstacles and dash climb obstacles as tall as itself and it doesn't really care. And so that was what that lets you do is get really [00:14:30] small CPS, really small computers that basically you put on these robots and they take very little power. But now for control, all of all they have to say is go or turn when they can use the rest of their computational time to say, read information from the camera and decide which way do I want to go? What's my objective? So from a stability controls point, it's couldn't be easier. Um, and now we're using these whatever extra [00:15:00] CPU cycles in our small board to do sort of more complicated behavior, but that's sort of another person's project.
Speaker 2: What sort of applications do you see this robot having? I know that you would want to use it as a vehicle, right? To have payloads on it. Right? And it also then goes into these strange places or if it can climb walls that's astounding. Right. On its own. Right. And then how do you then utilize it?
Speaker 3: The original goal was to have a robot that you could deploy [00:15:30] in search and rescue operations. So, um, say in an earthquake where you have claps buildings or claps minds, um, you can send in very small robots, uh, through the cracks, through the crevices down to find survivors. And you can have thousands of these really cheap and small robots and you don't care if 99% of the robots fail to find anyone or fail to even make it down as long as some small fraction finds a survivor, then you have, [00:16:00] technically you've succeeded. So the goal is to make lots of these small, inexpensive robots that can climb through the cracks, have sensors on them that can detect if someone's alive and then little radios to communicate with each other and communicate with the outside world to say, this is where someone is. Even if it's with some high probability that there's someone here, you know, it's worth spending your time digging in this exact location rather than having to uncover the entire building.
Speaker 2: I would imagine there are lots of uses in that realm of, of sensing [00:16:30] environments just in general, whether it's a collapse, building, a search and rescue, but you're just a hazardous place to monitor. And to have these things patrolling. So there's the, the whole idea is numbers and inexpensive, right? Manufacturer,
Speaker 3: right. So, so there are also proposals for environmental sensing. So deploying these robots, especially these nice mobile robots and say agricultural areas where you want to track how a crop dusters pesticides [00:17:00] travel across the countryside. You could have robots that sort of move and they can respond to say changing winds so that it can sort of get into the line of you know, the the path of these plumes of pesticides and sort of track how they're progressing across the country if they're affecting, you know, downwind communities. Also we have visions of putting these on bridges to do, checking for signs of stress on bridges and or say the nuclear power plants [00:17:30] in Japan. You could deploy these and have them run around and find you know, leaks or just have a nice mesh sort of deployed sensor network and sort of get readings from lots of different spaces and sort of try to understand how the radiation is moving. Oh
Speaker 4: you are listening to spectrum line k a l x Berkeley. We are talking with Paul Burke Meyer about designing and building small six legged crawling [00:18:00] and climbing robots.
Speaker 2: So Paul, how did you become interested in engineering?
Speaker 3: For me it was pretty clear from the beginning. So when I was younger, um, I was really interested in, well like most people in engineering right now. I built a lot of things out of Legos and connects and things and was really interested in electronics. I actually had [00:18:30] an elderly neighbor next door to me who I would go over and visit and uh, he would give me all of his popular mechanics magazines and popular science magazines when he was done reading them. And I think that was really the hook that got me because I was reading these magazines, seeing all these cool things and thinking like, how can I end up in this magazine? What can I do to be in this magazine because these are all really, really neat things. I think that was the, the original hook. Then, uh, it sort of blossomed [00:19:00] in high school.
Speaker 3: We had, uh, an advanced physics class. It was the first time it was offered and it was really sort of undefined. The curriculum wasn't really well formed and uh, as a result we had some freedom that you might not normally have in a high school course to do different projects that we wanted. Uh, the teacher at the time approached me maybe two thirds of the way into the year and said, hey, I have this, uh, this little programming board that they use at MIT for basic robotics things and I just have one of them and [00:19:30] you're doing well in the class. You want to see if you can maybe make a something and we can try to define a project for you using this board. The project ended up being making a robot that could drive through a maze and pop a balloon at the end. And he actually let me pick a partner to work with me. And I actually chose my girlfriend at the time who is now my wife. Um, and so we worked on this project for a long time and had a lot of fun. We made the whole, like the car system programmed the robot [00:20:00] and it was a spectacular failure, but it really was a lot of fun. And I think that was sort of what really cemented engineering for me.
Speaker 2: So you mentioned in, in talking about getting started in robotics and engineering, the the aspect of having a lot of fun with it and are you able to maintain that sense of fun and play in your work? For me
Speaker 3: this is, it's all fun. It's, I feel like I'm making toys all day [00:20:30] and I don't have to work at it to keep it fun because I love making these things and I think it's really exciting to come up with new structures and sort of understand why things aren't working, what you can do to change them. So for me it's, I mean adjust the, the project itself is so I think, I think it can be fun for other people when you have a like I can make this project fun for other people by actually making something that works and [00:21:00] sharing it with people and having this cool little robot that they can play with that can run up walls and things like that. But I think, I think it's true for lots of people in their careers. I think if you find the one you like, it's fun no matter what you do as long as, as long as you get to do it. So
Speaker 2: well thanks very much Paul for coming in and talking.
Speaker 3: Came with us was great. You're welcome. There was a lot of fun.
Speaker 4: The [00:21:30] video of dash on Youtube, search for dash resilient, high speed 16 gram x and pedal robot regular feature of spectrum is to mention a few of the science and technology events happening locally over the next few weeks. [inaudible].
Speaker 2: The Science at Cau lecture series for July will be presented by professor Romanian Kezar Rooney [00:22:00] and will be entitled Exoskeleton Systems for medical applications. Dr Casa Rooney is a professor in the Mechanical Engineering Department at the University of California, Berkeley and director of the Berkeley Robotics and human engineering laboratory is one of the world's leading experts in robotic human augmentation. The date of the lecture is Saturday, July 16th at 11:00 AM in the genetics and plant biology building room 100 which is on the northwest corner of the UC Berkeley campus. [00:22:30] The East Bay Science cafe is held the first Wednesday of every month that the cafe of Valparaiso at La Pena Cultural Center, 31 oh five Shattuck avenue in Berkeley from 7:00 PM to 9:00 PM the cost of admittance is the purchase of a beverage or food item of your choice. Wednesday, July 6th our crystal Cha graduate student and National Science Foundation Graduate Research Fellow in the Department of integrative biology at UC Berkeley will present. [00:23:00] Her topic is titled Spiders, Crustaceans, and sells omi. A story of how animals use cells to put themselves together.
Speaker 2: UC Berkeley. Professor Gordon. Frankie will present a discussion on native bee populations in the bay area at the Peralta community garden. This event is free and open to the public. It will be held Saturday, July 9th at noon in the Peralta community garden. The garden address is 1400 Peralta [00:23:30] AV in Berkeley. Since today's show is at the beginning of the month, let me remind you of the free admittance days for some of the local institutions that normally charge admission. The exploratorium in San Francisco is the first Wednesday of each month. The UC botanical garden in Strawberry Canyon. Berkeley is the first Thursday of each month. The Tech Museum in San Jose is the second Sunday of each month. The Cal Academy of Science in San Francisco is the third Wednesday of each month. [00:24:00] Now several news stories from the UC Berkeley News Center. The story about a new public website providing access to extensive climate change research being conducted at California universities and research centers.
Speaker 2: The website. cal-adapt.org has a variety of features tailored for different types of users, including members of the general public, concerned about their neighborhood or region decision-makers such as city planners and resource managers [00:24:30] and experts who want to examine data. The information on the website comes from peer reviewed climate change research funded by the California Energy Commission's public interest energy research program. The site displays the research data in a variety of climate change related scenarios and in map format modeling various projections such as changes in snowpack, wildfire, danger and temperature throughout the end of the century. The cal dash adapt website was developed by the [00:25:00] geospatial innovation facility at UC Berkeley's College of natural resources.
Speaker 2: The journal Science gives out a monthly prize called spore. Spore stands for science prize for online resources in education. The June award was given to the molecular work bench software developed by the Concord consortium. The molecular workbench is a free open source software tool that helps learners overcome challenges and understanding the science of atoms [00:25:30] and molecules. This software simulates atomic scale phenomenon, permits users to interact with them. It can model electrons, atoms, and molecules, which makes it exceptable across physics, chemistry, biology, and engineering. Students from grades five through college can use the software to experiment with atomic scale systems. The software includes an author ing tool that enables educators to create complete learning activities with simulations, [00:26:00] text, images, graphs, navigation links and embedded assessments. Hundreds of these activities have been created and tested in classrooms. Educators are free to download and use completed activities or simulations or create their own.
Speaker 2: The website is mw.concorde.org/modeler/in an earlier show, we carried a story [00:26:30] about research into toxic flame retardant chemicals in clothing and furniture which pose health hazards for babies and young children. A companion study on the efficacy of the flame retardants was released in June in a peer study presented at the 10th annual symposium on fire safety science at the University of Maryland on June 21st scientists found that California's furniture flammability standard technical bulletin one one seven does not provide measurable fire safety [00:27:00] benefits. The standard has led to the unnecessary use of flame retardant chemicals at high levels and baby products and furniture, widespread human and animal exposure, and the potential to harm human health and the environment. While there are no proven fire safety benefits to technical bulletin one one seven the chemicals used to meet it leak from furniture into house dust, which is ingested by people in pets.
Speaker 2: Humans studies have shown associations [00:27:30] between increased flame retardant body levels and reduced IQ in children reduced fertility and to Krinn and thyroid disruption changes in male hormone levels, adverse birth outcomes and impaired development. Flame retardants have been found in the bodies of nearly all north Americans tested with the highest human levels in young children and Californians. Dogs have retardant [00:28:00] levels up to 10 times higher than humans and cats because of their grooming behavior have levels up to 100 times higher. The California standard established by technical bulletin one one seven has become a de facto national standard legislation to allow an alternative fabric flammability standard that would provide equal or greater fire safety without the use of chemical flame retardants failed last month with strong opposition [00:28:30] from lobbyists for Kim Torah, Alber Marley and Israeli chemicals limited. For more information and the complete study, go to the website, green science policy.org
Speaker 5: [inaudible] [inaudible].
Speaker 4: The abuse occurred during the show is by Listonic Donna David from his album folk and acoustic made [00:29:00] available by a creative Commons attribution only licensed 3.0 editing assistance was provided by Judith White Marceline and Gretchen Sanders. Thank you for listening to spectrum. If you have any comments about the show, please send them to us via email. Our email address is spectrum dot k a l x@yahoo.com join us in two weeks [00:29:30] at the same time.
Speaker 5: [inaudible].
Hosted on Acast. See acast.com/privacy for more information.
