Episode 44: Jerry Pratt discusses the evolution and future of humanoid robots and bipedal walking

// Aug 15, 2017

Today’s podcast features Ken Ford and Dawn Kernagis interviewing their colleague, Dr. Jerry Pratt, a senior research scientist at IHMC who heads up the institute’s robotics group. In 2015, Jerry led an IHMC team that placed second out of 23 teams from around the world in the first-ever DARPA Robotics Challenge. IHMC also placed first in the competition which featured humanoid robots that primarily walked bipedally and first among all U.S. teams.

Jerry is a graduate of MIT, where he earned a doctorate in electrical engineering and computer science in 2000. As a graduate student at MIT, Jerry built his first robot which was also one of the first bipedal robots that could compliantly walk over rough terrain.

As you will learn in today’s interview, it was called “Spring Turkey” and is on display in MIT’s Boston museum. The second robot he built as a graduate student was called “Spring Flamingo,” and is on display in the lobby of IHMC’s Fred Levin Center in Pensacola.

After graduation, Jerry and some MIT colleagues founded a small company called Yobotics, which specialized in powered prosthetics, biomimetic robots, simulation software and robotic consulting.

He joined IHMC in 2002 and has become a well-known expert in bipedal walking. His algorithms are used in various robots around the world. Recent work on fast-running robots has resulted in ostrich-inspired running models and robot prototypes that are currently believed to be the fastest running robots in the world.

Jerry has six U.S. patents and was inducted into the Florida Inventors Hall of Fame in 2015. He lives in Pensacola with his wife Megan and their two children. He and he wife founded a science museum called the Pensacola MESS Hall, which stands for math, engineering, science, and stuff. The MESS Hall is a hands-on science museum for all ages that just celebrated it’s five-year anniversary.

Show notes:

4:37: Ken and Dawn welcome Jerry to the show

4:54: Dawn asks Jerry to talk about the time he once stole a science book from school.

5:45: Dawn asks Jerry to discuss his first invention, the knockout keyless door lock, that he came up with for his tree fort when he was a teen.

6:21: Dawn asks Jerry if he recalls his first computer program he wrote on the Commodore 64.

6:47: Ken comments on how in addition to writing computer programs, Jerry had an interest in electronics, particularly Heathkits.

7:08: Dawn discusses how Jerry played a lot of sports as a kid, going on to run varsity track and cross country at MIT.

7:46: Dawn asks Jerry if it was as an undergrad or a graduate student that he first became interested in robotics.

8:20: Ken discusses the first two robots Jerry put together: Spring Turkey then Spring Flamingo. He then asks Jerry to talk about the machines and how he came up with the names.

9:16: Dawn comments on how a few of Jerry’s colleagues have mentioned that much of our understanding of dynamic walking is still based on some of the original work Jerry did at MIT, and she then asks Jerry to talk about that work.

10:03: Ken asks Jerry to talk about how he and his wife, Megan Benson, met.

10:54: Ken asks Jerry to discuss the experience of co-founding Yobotics, which specialized in powered prosthetics, biomimetic robots, simulation software, and robot consulting, with his colleagues at MIT.

11:36: Dawn discusses the growth of robotics at IHMC since Jerry joined the team. She then asks Jerry to give a summary on the types of robots that he and his colleagues have been working on over the last 14 years at IHMC.

13:55: Dawn asks Jerry to talk about the books he often reads on organizational culture and teambuilding.

15:08: Dawn comments on how she has heard that Jerry is one of the worst motivational speakers ever and asks if it is true.

15:28: Ken comments on all of the work that Jerry and the IHMC team put into the DARPA Robotics Challenge, where they placed second in the world and first among the United States teams. He then asks Jerry to describe the experience.

16:24: Ken asks Jerry what it would look like if he were designing a new challenge focused on robotic mobility.

17:03: Dawn comments on a story about Jerry’s daughter Annie telling her kindergarten teacher that her daddy builds robots that fall down.

17:42: Dawn discusses how Jerry has spent most of his career thinking about how humans balance themselves to keep from falling, and how we use these strategies to help balance robots. She then asks Jerry to walk through that process.

18:35: Dawn asks Jerry what happens when a robot recovers its balance.

19:13: Dawn comments on how Jerry’s focus has been on bipedal walking. She then goes on to ask why engineers design robots with legs, and even more specifically, two legs.

20:03: Ken states that one of the advantages of a humanoid robot or bipedal robot is that the physical built world was built assuming our human form factor.

21:02: Ken comments on how Jerry mentioned that walking can be thought of as modeled by an inverted pendulum and asks Jerry when this approximation is valid and under what conditions this simplification breaks down.

21:47: Dawn asks Jerry which is a more limiting factor in having a bipedal robot perform as well as a human: the software or the hardware.

22:29: Dawn comments on how bipedal robots are able to walk but asks Jerry what the challenges are in getting them to run.

23:22: Ken discusses how our robots are comprised of lots of actuators and sensors and perform many complicated hundreds and perhaps thousands of calculations per second. He then asks Jerry how practical and robust this approach is in the real world and how their fundamental research into open-loop running robots such as the planar elliptical runner could be leveraged into advances in humanoid robots, or whether they are distinctly separate tracks with no cross talk.

24:54: Jerry talks about  how his group is looking at how to use robots to do other things in the real world than just walking, such as going up to valves, turning valves and flipping switches. And because this needs to be done super reliably, the group is using more mathematically reliable techniques to create systems and controllers that exploit natural dynamics.

25:18: Dawn asks Jerry to explain some of the potentially practical applications of bipedal robots.

25:49: Ken comments that he has noticed that Jerry and others have been working to get a humanoid robot, Atlas, to walk across a plank of wood that seesaws up and down. He then asks Jerry why he is doing that.

26:37: Dawn asks what some of the most efficient gates are as far as different animals go and how bipedal walking and programmability compare to some of those gates in terms of efficiency.


29:16: Ken discusses how Jerry has been spending a lot of time thinking about how robots can help us explore Mars. He goes on to comment that the robots currently on Mars are all wheeled or tracked machines, and asks Jerry if he sees the need for robots with legs on Mars, and why.

30:16: Ken asks Jerry how walking and running differs on Earth versus different gravity environments like Mars.

31:24: Ken asks Jerry if we would walk on Mars or hop, or if it would partially be a function of the spacesuit design.

32:52: Ken comments that as far as sensory input goes, the skin is our largest sensory organ that we have as humans. He then asks Jerry if he sees the ability to integrate things like synthetic skin with robots to improve their interaction with the environment.

34:22: Dawn discusses how in undersea robotics one of the major issues is developing a hand with dexterity that simulates a human is very difficult. She then asks Jerry the challenges in this and where the field stands with respect to developing a good hand.

35:57: Ken comments that understanding robotics and AI doesn’t seem to diminish our appreciation for humanity, but rather seems to elevate it.

36:21: Dawn states that when Jerry and his wife Megan moved to Pensacola 14 years ago they had a dream of creating a science museum. She then asks how he managed to make this happen.

38:36: Ken asks Jerry in what ways mechanical mobility is starting to surpass human mobility, and where Jerry sees the next breakthroughs coming in machine mobility.

40:27: Dawn comments on how about 10 years ago researchers set a goal that we would have a robot soccer team that would beat a human team by 2050, and she then asks Jerry if he thinks that this is still possible.

41:50: Ken discusses how Jerry’s team also works on powered exoskeletons for people with paralysis, but all of the balance is provided by the user. He then asks if there are ways to utilize Jerry’s work on walking and balancing algorithms in powered exoskeleton developments, so that the exoskeleton itself can be a more active partner in balance.

42:51: Dawn asks Jerry what his advice is to a race walker who wants to do well without cheating.

44:56: Dawn asks Jerry about the role of AI techniques in helping to advance mobility.

45:25: Jerry talks about how the deep reinforcement learning community may start looking at bipedal walking as one of its next big challenges.

46:45: Ken comments how deep learning algorithms are an ideal for bipedal walking because one doesn’t have to explain how the capability to walk was arrived at.

48:03: Jerry agrees but talks about how deep learning might someday help us understand walking in a new way and afford new insights.

48:37: Dawn asks Jerry what he sees as the emerging commercial opportunities for robotics.

50:02: Dawn asks Jerry about opportunities for entertainment.

50:43: Ken asks Jerry to give a quick update and sketch on IHMC’s role in the MegaBot adventure.

51:17: Dawn asks Jerry to talk about his pleasure in puzzles and games, and proving the odds of them.

52:00: Ken and Dawn thank Jerry for joining them.