Author Topic: Valkyrie (humanoid robot)  (Read 22607 times)

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #20 on: 12/21/2013 11:11 PM »
I've never understood why BD was messing around with hydraulics and pneumatics. Having worked with both on industrial robots I've always considered it a dead end.
They probably had military project requirements early on, I hope Google acquiring them changes course on that.

Japanese team with HRP project heritage winning is not really surprising. Kudos for them and CMU coming in ahead with elegant and robust platforms

I'm impressed that IHMC and MIT managed to make the clumsy atlas platform work well enough.
Orion - the first and only manned not-too-deep-space craft

Online Robotbeat

  • Senior Member
  • *****
  • Posts: 28112
  • Minnesota
  • Liked: 7922
  • Likes Given: 5276
Re: Valkyrie (humanoid robot)
« Reply #21 on: 12/21/2013 11:25 PM »
Because hydraulics can put a lot of power (and force) in a small space. Electric linear actuators suck compared to hydraulic linear actuators, compared pound for pound, liter for liter, if power density and force density matter to you. You can put the actual motive source (the hydraulic pump) at the center of gravity, which helps tremendously by reducing the mass of the appendages.

That isn't to say you could never develop an electric actuator which would compete with a hydraulic one in power and force, but it'd cost you more.

And, of course, electric actuators have other advantages.

But if you were to go the route of electric actuators, you'd have to put a LOT of effort into designing very high force and high power actuators. Hydraulic actuators are fairly simple in that regard. Provided you're using an external power source, hydraulic actuators let you focus on developing the kinematics and control algorithms instead of actuator design. I think this explains why Boston Dynamics has gotten further (I'd say MUCH further) than anyone else in building nimble biomimetic robots.

The advantage seems less pronounced once they have to use their own on-board motive power source, but because the motive power source can be decoupled from the kinematic issue, they've been able to make a lot of progress nonetheless. Better to focus on what you're good at (kinematics), not on something that many other people are probably a lot better at (building high power density motive power sources).
« Last Edit: 12/21/2013 11:35 PM by Robotbeat »
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Online Robotbeat

  • Senior Member
  • *****
  • Posts: 28112
  • Minnesota
  • Liked: 7922
  • Likes Given: 5276
Re: Valkyrie (humanoid robot)
« Reply #22 on: 12/22/2013 12:26 AM »
If they used a micro turbine instead of a conventional ICE for hydraulic power...
http://www.turbinemarine.com/images/range/RangeExtender.pdf

11kW for 10kg (and that includes the electric dynamo), that's pretty good! Might need to gear it down, but if you connected it to a hydraulic pump (vane type?), you'd have one heck of a powerful, lightweight hydraulic power source. It'd screech like a witch, though...
« Last Edit: 12/22/2013 12:36 AM by Robotbeat »
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #23 on: 12/22/2013 02:45 AM »
Um, electromechanical actuator power density has not really been a problem for a while now for similar types of robots. You can squeeze a lot of power into a very small motor, and then trade speed and torque as needed with gearing. If losses become an issue, add active cooling - and nobody really has bothered so far. Also, electromechanical stuff lends itself quite well to miniaturization, so you can get actuators in any size.
If you want more efficient motors you pretty much have them precision manufactured, use higher quality materials and go to more complicated control schemes as from BLDC to PMAC or even ASIM. And if you are really out for it, you go and license harmonic drive "gearing" instead of planetary. So to get better performing actuators, you just pay more money for it.
Smart servo manufacturers like Robotis and Dongbu will give you a full spectrum of performance/price to choose from.

Pneumatics and hydraulics in comparison have serious control bandwidth and other issues. Maintenance being one of them. They are a dead end as far as advancing the state of art in robotics.

However, like i said, none of these robots are held back from power density issues - there are a million other much more important problems to tackle.

Why BD started with hydraulics was probably  energy density and quick refueling requirements by military - they pretty much had to have a gas engine in the original BigDog. And if you have a gas engine and a very small number ( 8 or 10 apparently ) of high power joints - hydraulics sort of made sense. Extending the same approach to a humanoid form was a bad call IMHO.
« Last Edit: 12/22/2013 02:45 AM by savuporo »
Orion - the first and only manned not-too-deep-space craft

Offline ChrisWilson68

  • Senior Member
  • *****
  • Posts: 3629
  • Sunnyvale, CA
  • Liked: 2215
  • Likes Given: 2808
Re: Valkyrie (humanoid robot)
« Reply #24 on: 12/22/2013 05:00 AM »
Where do I go to vote for my tax dollars to be taken away from a NASA team spending millions on a third-rate robot and give it to a team from the University of Michigan trying to develop a water-based ion thruster to send cubesats into interplanetary space?

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #25 on: 12/22/2013 05:07 AM »
Reuters still thing that losing team is cool.

http://www.reuters.com/resources_v2/flash/video_embed.swf?videoId=276193303&edition=UK

ChrisWilson - maybe make your voice heard ? Citizen journalism works, to a certain extent. Crowdfunding works, too, assuming you have a good PR team.
« Last Edit: 12/22/2013 05:07 AM by savuporo »
Orion - the first and only manned not-too-deep-space craft

Online Robotbeat

  • Senior Member
  • *****
  • Posts: 28112
  • Minnesota
  • Liked: 7922
  • Likes Given: 5276
Re: Valkyrie (humanoid robot)
« Reply #26 on: 12/22/2013 05:56 AM »
Savuporo:
A bad choice for humanoid robots? Well, I haven't seen any other humanoid robots this robust:

Everything else seems to be very careful, tip-toeing everywhere. Even when they do things like careful little jumps. But this robot seems capable of a real sense of balance, able to stumble blindly over rocks without falling over. I'm just not convinced that this would be as easy to do with electric actuators.

And I challenge you to show me electric linear actuators that have a similar power and strength density as hydraulic ones operating at high pressure. The gearing you mention adds a lot of weight.

Everything I have read suggests that linear actuator strength density (and to some degree power density) is significantly higher with hydraulics than electrics.
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #27 on: 12/22/2013 06:24 PM »
A bad choice for humanoid robots? Well, I haven't seen any other humanoid robots this robust:
You should go through RoboCup archives, and then look at HRP-3 and its close relatives were doing years ago.

When Fukushima hit, there was quite a bit of hand wringing in Japan about "why are we building all these robots if they are useless when disaster strikes". The robotics researchers were caught off guard as they had been focussing on what was perceived a lot more important application at the time - elderly care.

Well, they certainly bounced back, didnt they ?

Robots robustness doesnt come from actuators. It comes down to control algorithms and sufficient control and sensory bandwidth.

And I challenge you to show me electric linear actuators that have a similar power and strength density as hydraulic ones operating at high pressure.
I'm not sure why you are stuck on linear actuators - nobody uses them in dexterous robotics, apart from some limited applications for the newer SMA actuators.
We used linear motors in the past on industrial robot cells as positioning tables, where high speed was required, but they really really have very limited uses.
Dexterous robots use exclusively rotary actuators as things are put together of rotary joints ..
« Last Edit: 12/22/2013 06:32 PM by savuporo »
Orion - the first and only manned not-too-deep-space craft

Online Robotbeat

  • Senior Member
  • *****
  • Posts: 28112
  • Minnesota
  • Liked: 7922
  • Likes Given: 5276
Re: Valkyrie (humanoid robot)
« Reply #28 on: 12/22/2013 08:07 PM »
A bad choice for humanoid robots? Well, I haven't seen any other humanoid robots this robust:
You should go through RoboCup archives, and then look at HRP-3 and its close relatives were doing years ago.

When Fukushima hit, there was quite a bit of hand wringing in Japan about "why are we building all these robots if they are useless when disaster strikes". The robotics researchers were caught off guard as they had been focussing on what was perceived a lot more important application at the time - elderly care.

Well, they certainly bounced back, didnt they ?

Robots robustness doesnt come from actuators. It comes down to control algorithms and sufficient control and sensory bandwidth.

And I challenge you to show me electric linear actuators that have a similar power and strength density as hydraulic ones operating at high pressure.
I'm not sure why you are stuck on linear actuators - nobody uses them in dexterous robotics, apart from some limited applications for the newer SMA actuators.
We used linear motors in the past on industrial robot cells as positioning tables, where high speed was required, but they really really have very limited uses.
Dexterous robots use exclusively rotary actuators as things are put together of rotary joints ..

Well Boston Dynamics seems to rely pretty heavily on linear actuators. I have seen the Japanese robots like HRP-3 and Asimo (which can jump a little bit, a cool trick) and while they are neat, they still aren't nearly as agile. There's a reason you don't see them walking on rocks and being pushed, walking through mud, etc. It may well be that the biomimetic approach of using linear actuators vs rotary ones may offer some non-obvious advantages. That hydraulics can develop very high forces is probably related to this.
« Last Edit: 12/22/2013 08:17 PM by Robotbeat »
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Online Robotbeat

  • Senior Member
  • *****
  • Posts: 28112
  • Minnesota
  • Liked: 7922
  • Likes Given: 5276
Re: Valkyrie (humanoid robot)
« Reply #29 on: 12/22/2013 08:18 PM »
Which isn't to say the Japanese don't know what they're doing or something... Clearly they do, given their rankings in this contest.
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #30 on: 12/22/2013 08:28 PM »
Well Boston Dynamics seems to rely pretty heavily on linear actuators.
They are pretty much the only ones in the field and like said, driven by their initial requirements. All of automation industry has been gradually moving away from pneumatics and hydraulics as the applications get more sophisticated.  You only really see hydraulics in very heavy automation applications, i.e. we are talking about tons of force. Pneumatics is still left on factory floors for simplest on/off type actuators.

For example, see, feel free to jump to last page
http://www.thomsonlinear.com/downloads/articles/Choosing_Between_Electromechanical_and_Fluid_Power_taen.pdf

Quote
I have seen the Japanese robots like HRP-3 and Asimo (which can jump a little bit, a cool trick) and while they are neat, they still aren't nearly as agile.

You really are not paying attention, even the simplest small humanoid bots driven with hobby servos can be made to do backflips. Again, very little to do with actuators.

Quote
There's a reason you don't see them walking on rocks and being pushed, walking through mud, etc.
Yeah and its got nothing to do with hydraulics or types of actuators being used.

Quote
It may well be that the biomimetic approach of using linear actuators vs rotary ones may offer some non-obvious advantages.
There is nothing more "biomimetic" about attaching a linear actuator vs a servo horn to the bracket or a "bone" in the body. Torque is torque.

Quote
That hydraulics can develop very high forces is probably related to this.
No.
Orion - the first and only manned not-too-deep-space craft

Online Robotbeat

  • Senior Member
  • *****
  • Posts: 28112
  • Minnesota
  • Liked: 7922
  • Likes Given: 5276
Re: Valkyrie (humanoid robot)
« Reply #31 on: 12/22/2013 08:40 PM »
If you say so. Show me a large legged robot walking through snow or mud. Hydraulics allows you to very cheaply have enormous amounts of torque (using linear actuators and joints) without requiring a huge joint motor.
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Online Robotbeat

  • Senior Member
  • *****
  • Posts: 28112
  • Minnesota
  • Liked: 7922
  • Likes Given: 5276
Re: Valkyrie (humanoid robot)
« Reply #32 on: 12/22/2013 08:42 PM »
No reason you can't run an electrical robot using a gasoline motor. It's basically the same situation mass wise as a hydraulic power unit. So that can't be the reason.
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #33 on: 12/22/2013 09:03 PM »
If you say so. Show me a large legged robot walking through snow or mud. Hydraulics allows you to very cheaply have enormous amounts of torque (using linear actuators and joints) without requiring a huge joint motor.
The real reason is that not many others in the world have spent energy or money on military applications and are trying to resolve the real hard problems in robotics first.

Again, its got nothing to do with hydraulics, but you can keep believing whatever you want. But if you want to learn about what the people in the field are actually working on you can browse through the topics and papers at IROS 2013 workshop for example
http://www.iros2013.org/tutorials-workshops.html


Orion - the first and only manned not-too-deep-space craft

Online Robotbeat

  • Senior Member
  • *****
  • Posts: 28112
  • Minnesota
  • Liked: 7922
  • Likes Given: 5276
Re: Valkyrie (humanoid robot)
« Reply #34 on: 12/23/2013 04:20 AM »
What the heck does them being military have to do with picking hydraulics? A gasoline engine can just as well turn a dynamo as a hydraulic pump.
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #35 on: 12/23/2013 07:40 AM »
I think we are at the point where it's not really worth explaining anymore.

Again, you dont see legged robots walking in the snow or mud because nobody else needed this application. People in field of robotics have been solving much more challenging problems.

BD picked hydraulics because it matched system design criteria in that application, where they could waste power for example. But again for dexterous robotics its a dead end due to its inherent limitations like control bandwidth and precision, and abysmal efficiency, just as most of automation, automotive industry etc is all moving to fully electromechanically actuated systems.

Associating "robustness" with hydraulics shows a fundamental lack of understanding of control theory.

Orion - the first and only manned not-too-deep-space craft

Offline ChrisWilson68

  • Senior Member
  • *****
  • Posts: 3629
  • Sunnyvale, CA
  • Liked: 2215
  • Likes Given: 2808
Re: Valkyrie (humanoid robot)
« Reply #36 on: 12/23/2013 08:44 AM »
It seems to me Robotbeat is making a much stronger case.  There's been no evidence presented to support the idea that Boston Dynamics chose hydraulics because of some kind of military requirement that wouldn't apply to non-military robots.  And there's no evidence to support the claim that anyone else can do the kinds of things Boston Dynamics can, or that what Boston Dynamics has achieved is somehow easy or too boring for others to do.

Boston Dynamics built the robots used by the second and fourth place teams at this challenge.  It was their robots that DARPA supplied to the second-tier teams to do software-only development, indicating DARPA considers BD's Atlas the standard to use.

Boston Dynamics really seems to know what they're doing.  And they've continued to use hydraulics.

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #37 on: 12/23/2013 04:23 PM »
Boston Dynamics built the robots used by the second and fourth place teams at this challenge.
Emphasis mine.

Quote
And there's no evidence to support the claim that anyone else can do the kinds of things Boston Dynamics can
::) Did you even watch the trials ?
Do you keep up with the news on robotics at all ? How many robots have you built, or at least did you go to Atlanta this year or Osaka in 2012  ?

If you actually look for your "evidence", you'll find plenty.

Quote
Boston Dynamics really seems to know what they're doing.
Nobody claimed otherwise

Orion - the first and only manned not-too-deep-space craft

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5155
  • Liked: 981
  • Likes Given: 343
Re: Valkyrie (humanoid robot)
« Reply #38 on: 12/23/2013 04:43 PM »
And if you really really think that staying upright when kicked is the pinnacle of robotics you can look at old videos posted by DLR and AIST - two humanoid robots research powerhouses. Or go look at what Willow Garage has built.

And then go look at what else has been built in the same labs meanwhile.  And then go count how many RoboCup finals have been won with hydraulics.



( notice HRP-3 standing quietly in the corner. This is the origin of the robot that just aced the DRC challenge )
« Last Edit: 12/23/2013 04:52 PM by savuporo »
Orion - the first and only manned not-too-deep-space craft

Offline darkbluenine

  • Full Member
  • **
  • Posts: 208
  • Liked: 31
  • Likes Given: 51
Re: Valkyrie (humanoid robot)
« Reply #39 on: 12/23/2013 06:47 PM »
Where do I go to vote for my tax dollars to be taken away from a NASA team spending millions on a third-rate last-place robot and give it to a team from the University of Michigan trying to develop a water-based ion thruster to send cubesats into interplanetary space any of the six university teams that placed ahead of JSC?

Seconded, with some minor changes.  I'm sorry for the team, but from a management point-of-view, this has been a wasteful allocation of resources in today's very constrained budget environment.  Worse, it makes it that much harder to argue for a larger agency budget when so much, especially in comparison to the other competitors, is being squandered on a tangential project with such poor performance.

No wonder NASA doesn't fund any sizable competitions with its prize program. 
« Last Edit: 12/23/2013 08:28 PM by darkbluenine »

Tags: