English subtitles

← My seven species of robot -- and how we created them

At TEDxNASA, Dennis Hong introduces seven award-winning, all-terrain robots -- like the humanoid, soccer-playing DARwIn and the cliff-gripping CLIMBeR -- all built by his team at RoMeLa, Virginia Tech. Watch to the end to hear the five creative secrets to his lab's incredible technical success.

Get Embed Code
26 Languages

Showing Revision 6 created 06/21/2017 by Krystian Aparta.

  1. So the first robot
    to talk about is called STriDER.
  2. It stands for Self-excited
    Tripedal Dynamic Experimental Robot.
  3. It's a robot that has three legs,
    which is inspired by nature.
  4. But have you seen anything in nature,
    an animal that has three legs?
  5. Probably not. So why do I call this
    a biologically inspired robot?
  6. How would it work?
  7. But before that,
    let's look at pop culture.
  8. So, you know H.G. Wells's
    "War of the Worlds," novel and movie.
  9. And what you see over here
    is a very popular video game,
  10. and in this fiction, they describe
    these alien creatures and robots
  11. that have three legs that terrorize Earth.
  12. But my robot, STriDER,
    does not move like this.
  13. This is an actual dynamic
    simulation animation.

  14. I'm going to show you how the robot works.
  15. It flips its body 180 degrees
  16. and it swings its leg between the two legs
  17. and catches the fall.
  18. So that's how it walks.
  19. But when you look at us
    human beings, bipedal walking,
  20. what you're doing is,
  21. you're not really using muscle
    to lift your leg and walk like a robot.
  22. What you're doing is,
    you swing your leg and catch the fall,
  23. stand up again, swing your leg
    and catch the fall.
  24. You're using your built-in dynamics,
    the physics of your body,
  25. just like a pendulum.
  26. We call that the concept
    of passive dynamic locomotion.
  27. What you're doing is, when you stand up,
  28. potential energy to kinetic energy,
  29. potential energy to kinetic energy.
  30. It's a constantly falling process.
  31. So even though there is nothing
    in nature that looks like this,
  32. really, we're inspired by biology
    and applying the principles of walking
  33. to this robot.
  34. Thus, it's a biologically inspired robot.
  35. What you see here,
    this is what we want to do next.

  36. We want to fold up the legs
    and shoot it up for long-range motion.
  37. And it deploys legs --
    it looks almost like "Star Wars" --
  38. so when it lands, it absorbs
    the shock and starts walking.
  39. What you see over here, this yellow thing,
    this is not a death ray.
  40. (Laughter)

  41. This is just to show you

  42. that if you have cameras
    or different types of sensors,
  43. because it's 1.8 meters tall,
  44. you can see over obstacles like bushes
    and those kinds of things.
  45. So we have two prototypes.

  46. The first version,
    in the back, that's STriDER I.
  47. The one in front,
    the smaller, is STriDER II.
  48. The problem we had with STriDER I is,
    it was just too heavy in the body.
  49. We had so many motors aligning the joints
  50. and those kinds of things.
  51. So we decided to synthesize
    a mechanical mechanism
  52. so we could get rid of all the motors,
    and with a single motor,
  53. we can coordinate all the motions.
  54. It's a mechanical solution to a problem,
    instead of using mechatronics.
  55. So with this, now the top body
    is lighted up; it's walking in our lab.
  56. This was the very first successful step.
  57. It's still not perfected,
    its coffee falls down,
  58. so we still have a lot of work to do.
  59. The second robot I want
    to talk about is called IMPASS.

  60. It stands for Intelligent Mobility
    Platform with Actuated Spoke System.
  61. It's a wheel-leg hybrid robot.
  62. So think of a rimless wheel
    or a spoke wheel,
  63. but the spokes individually
    move in and out of the hub;
  64. so, it's a wheel-leg hybrid.
  65. We're literally reinventing
    the wheel here.
  66. Let me demonstrate how it works.
  67. So in this video we're using an approach
    called the reactive approach.
  68. Just simply using
    the tactile sensors on the feet,
  69. it's trying to walk
    over a changing terrain,
  70. a soft terrain where it pushes
    down and changes.
  71. And just by the tactile information,
  72. it successfully crosses
    over these types of terrains.
  73. But, when it encounters
    a very extreme terrain --

  74. in this case, this obstacle
    is more than three times the height
  75. of the robot --
  76. then it switches to a deliberate mode,
  77. where it uses a laser range finder
    and camera systems
  78. to identify the obstacle and the size.
  79. And it carefully plans
    the motion of the spokes
  80. and coordinates it so it can show
    this very impressive mobility.
  81. You probably haven't seen
    anything like this out there.
  82. This is a very high-mobility robot
    that we developed called IMPASS.
  83. Ah, isn't that cool?
  84. When you drive your car,

  85. when you steer your car, you use
    a method called Ackermann steering.
  86. The front wheels rotate like this.
  87. For most small-wheeled robots,
  88. they use a method
    called differential steering
  89. where the left and right wheel
    turn the opposite direction.
  90. For IMPASS, we can do many,
    many different types of motion.
  91. For example, in this case,
  92. even though the left and right
    wheels are connected
  93. with a single axle rotating
    at the same angle of velocity,
  94. we simply change the length
    of the spoke, it affects the diameter,
  95. then can turn to the left
    and to the right.
  96. These are just some examples
    of the neat things we can do with IMPASS.
  97. This robot is called CLIMBeR:

  98. Cable-suspended Limbed Intelligent
    Matching Behavior Robot.
  99. I've been talking
    to a lot of NASA JPL scientists --
  100. at JPL, they are famous
    for the Mars rovers --
  101. and the scientists,
    geologists always tell me
  102. that the real interesting science,
    the science-rich sites,
  103. are always at the cliffs.
  104. But the current rovers cannot get there.
  105. So, inspired by that,
    we wanted to build a robot
  106. that can climb
    a structured cliff environment.
  107. So this is CLIMBeR.

  108. It has three legs.
  109. It's probably difficult to see, but it has
    a winch and a cable at the top.
  110. It tries to figure out
    the best place to put its foot.
  111. And then once it figures that out,
  112. in real time, it calculates
    the force distribution:
  113. how much force it needs
    to exert to the surface
  114. so it doesn't tip and doesn't slip.
  115. Once it stabilizes that, it lifts a foot,
  116. and then with the winch,
    it can climb up these kinds of cliffs.
  117. Also for search and rescue
    applications as well.
  118. Five years ago,
    I actually worked at NASA JPL

  119. during the summer as a faculty fellow.
  120. And they already had
    a six-legged robot called LEMUR.
  121. So this is actually based on that.
  122. This robot is called MARS:
  123. Multi-Appendage Robotic System.
  124. It's a hexapod robot.
  125. We developed our adaptive gait planner.
  126. We actually have a very interesting
    payload on there.
  127. The students like to have fun.
  128. And here you can see that it's walking
    over unstructured terrain.
  129. (Motor sound)

  130. It's trying to walk
    on the coastal terrain, a sandy area,

  131. but depending on the moisture content
    or the grain size of the sand,
  132. the foot's soil sinkage model changes,
    so it tries to adapt its gait
  133. to successfully cross
    over these kind of things.
  134. It also does some fun stuff.
  135. As you can imagine,
    we get so many visitors visiting our lab.
  136. So when the visitors come,
    MARS walks up to the computer,
  137. starts typing, "Hello, my name is MARS.
  138. Welcome to RoMeLa,
  139. the Robotics Mechanisms
    Laboratory at Virginia Tech."
  140. (Laughter)

  141. This robot is an amoeba robot.

  142. Now, we don't have enough time
    to go into technical details,
  143. I'll just show you
    some of the experiments.
  144. These are some of the early
    feasibility experiments.
  145. We store potential energy
    to the elastic skin to make it move,
  146. or use active tension cords
    to make it move forward and backward.
  147. It's called ChIMERA.
  148. We also have been working
    with some scientists and engineers
  149. from UPenn
  150. to come up with a chemically actuated
    version of this amoeba robot.
  151. We do something to something,
  152. and just like magic, it moves.
  153. "The Blob."
  154. This robot is a very recent project.

  155. It's called RAPHaEL:
  156. Robotic Air-Powered Hand
    with Elastic Ligaments.
  157. There are a lot of really neat,
    very good robotic hands
  158. out there on the market.
  159. The problem is,
    they're just too expensive --
  160. tens of thousands of dollars.
  161. So for prosthesis applications
    it's probably not too practical,
  162. because it's not affordable.
  163. We wanted to tackle this problem
    in a very different direction.
  164. Instead of using electrical motors,
    electromechanical actuators,
  165. we're using compressed air.
  166. We developed these novel actuators
    for the joints, so it's compliant.
  167. You can actually change the force,
  168. simply just changing the air pressure.
  169. And it can actually crush
    an empty soda can.
  170. It can pick up very delicate
    objects like a raw egg,
  171. or in this case, a lightbulb.
  172. The best part: it took only 200 dollars
    to make the first prototype.
  173. This robot is actually
    a family of snake robots

  174. that we call HyDRAS,
  175. Hyper Degrees-of-freedom Robotic
    Articulated Serpentine.
  176. This is a robot that can climb structures.
  177. This is a HyDRAS's arm.
  178. It's a 12-degrees-of-freedom robotic arm.
  179. But the cool part is the user interface.
  180. The cable over there,
    that's an optical fiber.
  181. This student, it's probably
    her first time using it,
  182. but she can articulate it
    in many different ways.
  183. So, for example, in Iraq, the war zone,
    there are roadside bombs.
  184. Currently, you send these remotely
    controlled vehicles that are armed.
  185. It takes really a lot of time
    and it's expensive to train the operator
  186. to operate this complex arm.
  187. In this case, it's very intuitive;
  188. this student, probably
    his first time using it,
  189. is doing very complex manipulation tasks,
  190. picking up objects and doing
    manipulation, just like that.
  191. Very intuitive.
  192. Now, this robot
    is currently our star robot.

  193. We actually have a fan club
    for the robot, DARwIn:
  194. Dynamic Anthropomorphic
    Robot with Intelligence.
  195. As you know, we're very interested
    in human walking,
  196. so we decided to build
    a small humanoid robot.
  197. This was in 2004; at that time,
  198. this was something really,
    really revolutionary.
  199. This was more of a feasibility study:
  200. What kind of motors should we use?
    Is it even possible?
  201. What kinds of controls should we do?
  202. This does not have any sensors,
    so it's an open-loop control.
  203. For those who probably know,
    if you don't have any sensors
  204. and there's any disturbances,
    you know what happens.
  205. (Laughter)

  206. Based on that success, the following year
    we did the proper mechanical design,

  207. starting from kinematics.
  208. And thus, DARwIn I was born in 2005.
  209. It stands up, it walks -- very impressive.
  210. However, still, as you can see,
    it has a cord, an umbilical cord.
  211. So we're still using
    an external power source
  212. and external computation.
  213. So in 2006, now it's really
    time to have fun.

  214. Let's give it intelligence.
  215. We give it all the computing
    power it needs:
  216. a 1.5 gigahertz Pentium M chip,
    two FireWire cameras,
  217. rate gyros, accelerometers,
    four forced sensors on the foot,
  218. lithium polymer batteries --
  219. and now DARwIn II
    is completely autonomous.
  220. It is not remote controlled.
    There's no tethers.
  221. It looks around, searches for the ball ...
    looks around, searches for the ball,
  222. and it tries to play a game of soccer
    autonomously -- artificial intelligence.
  223. Let's see how it does.
  224. This was our very first trial, and ...
  225. (Video) Spectators: Goal!

  226. Dennis Hong: There is actually
    a competition called RoboCup.

  227. I don't know how many of you
    have heard about RoboCup.
  228. It's an international autonomous
    robot soccer competition.
  229. And the actual goal of RoboCup is,
  230. by the year 2050,
  231. we want to have full-size,
    autonomous humanoid robots
  232. play soccer against the human
    World Cup champions
  233. and win.
  234. (Laughter)

  235. It's a true, actual goal.

  236. It's a very ambitious goal,
    but we truly believe we can do it.
  237. This is last year in China.

  238. We were the very first team
    in the United States that qualified
  239. in the humanoid RoboCup competition.
  240. This is this year in Austria.
  241. You're going to see the action
    is three against three,
  242. completely autonomous.
  243. (Video) (Crowd groans)

  244. DH: There you go. Yes!

  245. The robots track and they team-play
    amongst themselves.
  246. It's very impressive.
  247. It's really a research event,
  248. packaged in a more exciting
    competition event.
  249. What you see here is the beautiful
    Louis Vuitton Cup trophy.
  250. This is for the best humanoid.
  251. We'd like to bring this, for the first
    time, to the United States next year,
  252. so wish us luck.
  253. (Applause)

  254. Thank you.

  255. (Applause)

  256. DARwIn also has a lot of other talents.

  257. Last year, it actually conducted
    the Roanoke Symphony Orchestra
  258. for the holiday concert.
  259. This is the next generation
    robot, DARwIn IV,
  260. much smarter, faster, stronger.
  261. And it's trying to show off its ability:
  262. "I'm macho, I'm strong."
  263. (Laughter)

  264. "I can also do some Jackie Chan-motion,
    martial art movements."

  265. (Laughter)

  266. And it walks away. So this is DARwIn IV.

  267. Again, you'll be able
    to see it in the lobby.
  268. We truly believe this will be
    the very first running humanoid robot
  269. in the United States.
  270. So stay tuned.
  271. All right. So I showed you
    some of our exciting robots at work.

  272. So, what is the secret of our success?
  273. Where do we come up with these ideas?
  274. How do we develop these kinds of ideas?
  275. We have a fully autonomous vehicle
  276. that can drive into urban environments.
  277. We won a half a million dollars
    in the DARPA Urban Challenge.
  278. We also have the world's very first
    vehicle that can be driven by the blind.
  279. We call it the Blind Driver
    Challenge, very exciting.
  280. And many, many other robotics
    projects I want to talk about.
  281. These are just the awards
    that we won in 2007 fall
  282. from robotics competitions
    and those kinds of things.
  283. So really, we have five secrets.

  284. First is: Where do we get inspiration?
  285. Where do we get this spark of imagination?
  286. This is a true story, my personal story.
  287. At night, when I go to bed,
    at three, four in the morning,
  288. I lie down, close my eyes,
    and I see these lines and circles
  289. and different shapes floating around.
  290. And they assemble, and they form
    these kinds of mechanisms.
  291. And I think, "Ah, this is cool."
  292. So right next to my bed
    I keep a notebook, a journal,
  293. with a special pen
    that has an LED light on it,
  294. because I don't want to turn on the light
    and wake up my wife.
  295. So I see this, scribble everything down,
    draw things, and go to bed.

  296. Every day in the morning,
    the first thing I do,
  297. before my first cup of coffee,
    before I brush my teeth,
  298. I open my notebook.
  299. Many times it's empty;
    sometimes I have something there.
  300. If something's there, sometimes it's junk.
  301. But most of the time,
    I can't read my handwriting.
  302. Four in the morning --
    what do you expect, right?
  303. So I need to decipher what I wrote.
  304. But sometimes I see
    this ingenious idea in there,
  305. and I have this eureka moment.
  306. I directly run to my home office,
    sit at my computer,
  307. I type in the ideas, I sketch things out
  308. and I keep a database of ideas.
  309. So when we have these calls for proposals,
  310. I try to find a match
    between my potential ideas
  311. and the problem.
  312. If there's a match,
    we write a research proposal,
  313. get the research funding in,
  314. and that's how we start
    our research programs.
  315. But just a spark of imagination
    is not good enough.

  316. How do we develop these kinds of ideas?
  317. At our lab RoMeLa, the Robotics
    and Mechanisms Laboratory,
  318. we have these fantastic
    brainstorming sessions.
  319. So we gather around, we discuss problems
    and solutions and talk about it.
  320. But before we start,
    we set this golden rule.
  321. The rule is:
  322. nobody criticizes anybody's ideas.
  323. Nobody criticizes any opinion.
  324. This is important, because many times,
    students fear or feel uncomfortable
  325. about how others might think
    about their opinions and thoughts.
  326. So once you do this, it is amazing
    how the students open up.

  327. They have these wacky, cool,
    crazy, brilliant ideas,
  328. and the whole room is just electrified
    with creative energy.
  329. And this is how we develop our ideas.
  330. Well, we're running out of time.

  331. One more thing I want to talk about is,
  332. you know, just a spark of idea
    and development is not good enough.
  333. There was a great TED moment --
    I think it was Sir Ken Robinson, was it?
  334. He gave a talk about how education
    and school kill creativity.
  335. Well, actually,
    there's two sides to the story.
  336. So there is only so much one can do
    with just ingenious ideas
  337. and creativity
    and good engineering intuition.
  338. If you want to go beyond a tinkering,
  339. if you want to go
    beyond a hobby of robotics
  340. and really tackle
    the grand challenges of robotics
  341. through rigorous research,
  342. we need more than that.
  343. This is where school comes in.
  344. Batman, fighting against the bad guys,

  345. he has his utility belt,
    he has his grappling hook,
  346. he has all different kinds of gadgets.
  347. For us roboticists,
    engineers and scientists,
  348. these tools are the courses
    and classes you take in class.
  349. Math, differential equations.
  350. I have linear algebra, science, physics --
  351. even, nowadays, chemistry
    and biology, as you've seen.
  352. These are all the tools we need.
  353. So the more tools you have, for Batman,
  354. more effective at fighting the bad guys,
  355. for us, more tools to attack
    these kinds of big problems.
  356. So education is very important.
  357. Also -- it's not only about that.

  358. You also have to work really, really hard.
  359. So I always tell my students,
  360. "Work smart, then work hard."
  361. This picture in the back --
    this is three in the morning.
  362. I guarantee if you come
    to our lab at 3, 4am,
  363. we have students working there,
  364. not because I tell them to,
    but because we are having too much fun.
  365. Which leads to the last topic:
  366. do not forget to have fun.
  367. That's really the secret of our success,
    we're having too much fun.
  368. I truly believe that highest productivity
    comes when you're having fun,
  369. and that's what we're doing.
  370. And there you go.
  371. Thank you so much.

  372. (Applause)