4 ms·

Can anyone in robotics weigh in on whether new approaches like pneubotics / otherlab [0, 1] make Boston Dynamics' super heavy/noisy/expensive robots irrelevant

Can anyone in robotics weigh in on whether new approaches like pneubotics / otherlab [0, 1] make Boston Dynamics' super heavy/noisy/expensive robots irrelevant or not? That's what Saul Griffith has been saying at least [2], I'd like to hear an impartial take on it [0] http://www.pneubotics.com/ http://www.pneubotics.com/ [1] https://otherlab.com/projects https://otherlab.com/projects [2] "[...] we're the only game in town." 21:00 onwards https://youtu.be/gyMowPAJwqo?t=1263 https://youtu.be/gyMowPAJwqo?t=1263

pj_mukh 11y ago
Boston Dynamics' primary innovation is in the software/control design. The reason their robots are noisy is because they really haven't innovated a whole lot in the basic actuator design. That's not to say their mechanical design was easy, not by a long shot. It's super impressive stuff! Any new inventions that would make their mechanical design steer away from loud hydraulics would no doubt be amazing. However, without Boston dynamics' delightfully ground-breaking controls software design, it'll be dead in the water! For a little bit of proof, see who they are hiring! http://www.bostondynamics.com/bd_jobs.html http://www.bostondynamics.com/bd_jobs.html
- jarmitage 11y ago
  Thanks for your reply. Is there much publicly available information about their software/control design innovations? Hmm, maybe one might expect they would be hiring more mechatronics engineers to do just that? Or maybe there's still not much point for them because as you mentioned their focus is slightly different currently.
  - pj_mukh 11y ago
    A person with a core Mechatronics focus will probably be a great systems integrator. BD's core workforce are these PhD's in controls (usually from ECE departments). Source: Am a Mechatronics Engineer with a Controls focus. check out some of the work of BD's CEO (Marc Reibert). This (http://web.unair.ac.id/admin/file/f_7773_bigdog.pdf http://web.unair.ac.id/admin/file/f_7773_bigdog.pdf) is the summary paper on Big Dog, but they cite a bunch of the older Reibert papers on Controls designs for legs.
- PhasmaFelis 11y ago
  A lot (most?) of the noise comes from the engine, not the hydraulics, and the engine can't be replaced because sufficiently high-capacity batteries do not exist yet. That's not a problem that can be solved by innovative actuators.
  - pj_mukh 11y ago
    That too. Also engine design was contracted out.
gene-h 11y ago
In my opinion, I don't think inflatable robots are very practical.(except for niche applications) One of the most important things in robotics is reliability, the more work your robot does before breaking down or needing maintenance the more money it makes you. In industrial robot arms the gold standard for this is a Mean Time Between Failures of more than 100,000 hours. That's more than 10 years of continuous operation! I am skeptical that inflatable robots will be able to last this long. The fabric/elastomer combo is certainly not going to last through 10 years of continuous operation. Sure this part may be cheap, but you still need expensive valving and pneumatics to control said robot. Not to mention that maintenance costs money too. The payback period for robots is also shortening, The other problem that pneumatic robots suffer from is that air is compressible. This means that moving it around to drive pneumatics is gonna be inefficient and that pneumatic structures aren't that rigid. Higher rigidity means higher resonant frequency which means your robot can operate faster without wobbling around. The inefficiency might be compensated for because the robot is so light, but I have yet to see any hard numbers on this. This lack of rigidity is touted as a feature by the people who make inflatable robots. Because they are so light and aren't rigid they aren't going to hurt people if they fuck up. There are other ways to solve this problem that are currently used in the robotics industry. One of them is to put a spring on every link in the robot, which is what the Baxter robot does. Another is to make the robot as light as possible and limit speed which is what one of Kuka's human safe robots does[0]. Better control also fixes this problem, if you don't hit the human then you don't have any problem. But there might be niche applications. Maybe they will find use in the medical field where having anything rigid touch a human is unacceptable or where you need a weird shape to grab a human on a bed. Entertainment might be another, a while back a japanese company made giant inflatable robots for parades. [0] http://www.phriends.eu/URAI_08.pdf http://www.phriends.eu/URAI_08.pdf
- jarmitage 11y ago
  Hey thanks for responding! If the parts are cheap enough, might it not matter that they don't last the regular MTTF? One of the arguments I've seen them using is that their robots can actually move faster than regular ones due to their weight advantage - currently I would guess accuracy is not as good though. Another point I've been thinking about is that biological systems tend to combine the rigid and non-rigid per the needs of the organism, right? It seems that combining these techniques could also be interesting.

pj_mukh 11y ago

Boston Dynamics' primary innovation is in the software/control design. The reason their robots are noisy is because they really haven't innovated a whole lot in the basic actuator design. That's not to say their mechanical design was easy, not by a long shot. It's super impressive stuff! Any new inventions that would make their mechanical design steer away from loud hydraulics would no doubt be amazing. However, without Boston dynamics' delightfully ground-breaking controls software design, it'll be dead in the water! For a little bit of proof, see who they are hiring! http://www.bostondynamics.com/bd_jobs.html http://www.bostondynamics.com/bd_jobs.html

jarmitage 11y ago
Thanks for your reply. Is there much publicly available information about their software/control design innovations? Hmm, maybe one might expect they would be hiring more mechatronics engineers to do just that? Or maybe there's still not much point for them because as you mentioned their focus is slightly different currently.
- pj_mukh 11y ago
  A person with a core Mechatronics focus will probably be a great systems integrator. BD's core workforce are these PhD's in controls (usually from ECE departments). Source: Am a Mechatronics Engineer with a Controls focus. check out some of the work of BD's CEO (Marc Reibert). This (http://web.unair.ac.id/admin/file/f_7773_bigdog.pdf http://web.unair.ac.id/admin/file/f_7773_bigdog.pdf) is the summary paper on Big Dog, but they cite a bunch of the older Reibert papers on Controls designs for legs.
PhasmaFelis 11y ago
A lot (most?) of the noise comes from the engine, not the hydraulics, and the engine can't be replaced because sufficiently high-capacity batteries do not exist yet. That's not a problem that can be solved by innovative actuators.
- pj_mukh 11y ago
  That too. Also engine design was contracted out.

gene-h 11y ago

In my opinion, I don't think inflatable robots are very practical.(except for niche applications) One of the most important things in robotics is reliability, the more work your robot does before breaking down or needing maintenance the more money it makes you. In industrial robot arms the gold standard for this is a Mean Time Between Failures of more than 100,000 hours. That's more than 10 years of continuous operation! I am skeptical that inflatable robots will be able to last this long. The fabric/elastomer combo is certainly not going to last through 10 years of continuous operation. Sure this part may be cheap, but you still need expensive valving and pneumatics to control said robot. Not to mention that maintenance costs money too. The payback period for robots is also shortening, The other problem that pneumatic robots suffer from is that air is compressible. This means that moving it around to drive pneumatics is gonna be inefficient and that pneumatic structures aren't that rigid. Higher rigidity means higher resonant frequency which means your robot can operate faster without wobbling around. The inefficiency might be compensated for because the robot is so light, but I have yet to see any hard numbers on this. This lack of rigidity is touted as a feature by the people who make inflatable robots. Because they are so light and aren't rigid they aren't going to hurt people if they fuck up. There are other ways to solve this problem that are currently used in the robotics industry. One of them is to put a spring on every link in the robot, which is what the Baxter robot does. Another is to make the robot as light as possible and limit speed which is what one of Kuka's human safe robots does[0]. Better control also fixes this problem, if you don't hit the human then you don't have any problem. But there might be niche applications. Maybe they will find use in the medical field where having anything rigid touch a human is unacceptable or where you need a weird shape to grab a human on a bed. Entertainment might be another, a while back a japanese company made giant inflatable robots for parades. [0] http://www.phriends.eu/URAI_08.pdf http://www.phriends.eu/URAI_08.pdf

jarmitage 11y ago
Hey thanks for responding! If the parts are cheap enough, might it not matter that they don't last the regular MTTF? One of the arguments I've seen them using is that their robots can actually move faster than regular ones due to their weight advantage - currently I would guess accuracy is not as good though. Another point I've been thinking about is that biological systems tend to combine the rigid and non-rigid per the needs of the organism, right? It seems that combining these techniques could also be interesting.