Slashdot Mirror
Honda Creates Walking Robot
ilyah writes: "The ever-crafty Honda engineers have apparently created a robot that can accurately walk on two legs (i.e. just like humans). Take a look at The Honda Japan site -- a Japanese site that has some pictures and videos worth looking at (hint: videos are under linked marked "04"). It's pretty awesome watching that thing walk around." If I remember correctly, they've been working on this for quite some time - like 15 - 20 years, but this is apparently one of the breakthrough steps, pun intended.
This was posted a while ago, but with a different site. Try this site: http://www.honda-p3.com/ for english text and movies and such.
Stupid people will be persecuted to the fullest extent allowed by law.
- Greatly improved user-robot interface. Last I saw, the robot required a fairly intricate computer interface and some fairly serious developer guidance to work properly. Now, they seem to have the bulk of it hooked up to a wearable PC and game pad. Additionally, the robot is now completely wireless, which I believe wasn't the case earlier this summer. Quite the improvement over the old method.
- Greatly improved motor control software. This model actually moves in a fluid, almost natural way. The turning is vastly improved (the robot used to be able to only execute spot turns) and they seem to pretty much have the basics of the counter-balancing swing down.
- Improved robot design? It appears that they've tweaked the design of the robot somewhat, most noticably in the robot's backpack, which used to tower over it's shoulders like an astronaut's spacesuit. The current model looks much more "approachable" (for lack of a better way of quantifying it) and seems to have a few other cosmetic changes added (the hands, I believe, were less hand-like in the previous model, for example.) I'm not seeing the tech specs readily available in English, so I'll assume that the prototype is pretty much the same physically as the previous one, save for (quite non-trivial) space optimizations and otherwise more modern components. The fundamental design appears the same.
- Overall Humanness. This robot actually moves in a way that I could call human-like. Yes, there are still the herky-jerks, and the arms don't quite seem to swing right, but the movement flows right. It can walk pretty much in whatever fashion it chooses, can vary it's walking pace and step length, can walk sideways and backwards, can "swing" from a side step into a front step, and does it all with stunning grace for a machine. What's more, it's moving fast now--we're talking a brisk walking pace, no more of this ponderous mechanized trudging. The old videos from this summer were impressive because the robot could successfully walk in a straight like, lean to compensate for tilting floors, and navigate up and down staircases without falling over and with minimal programmer intervention. Today, this thing's strutting around like it owns the place.
This is some serious progress from the good folks at Honda. I'm really excited to see what the next version brings.$ man reality
Obliteracy: Words with explosions
The more interesting thing that I got from the special was the robot that *learned* to walk like an infant does. At power up, it is wobly and nervous. As the instructor helps it maintain its balance and catch it falls, the robot's AI (neural networked?) learns from its mistakes and improves. After a few hours, the bi-pedal contraption is able to walk on its own. That is, until it is turned off and its memory gone.
Having discussed extensively design possibilities with the big walker operator and worked a bit on a Walking robot repairs with him while with SRL (www.srl.org), I can attest that the Honda robot is definately an impressive achievement. Stable walking is difficult enough on tripedal and quadropedal robots, and the hacking done at SRL only succeeded in 2 and 3 legged robots, with not much hill-climbing ability, and only moderate speeds. The feedback circuitry and balancing techniques needed for a biped like the Honda robot is, sadly, not yet within reach of even the sophisticated hackers without a ton of money.
The Honda robot qualifies as an "impressive start"... its 2.0km/h speed of the Honda robot is not impressive for a lightweight tri or quad walker, but it is for a bipedal robot of human size.
The fact that it can climb stairs is especially cool, given the extensive rebalancing done every moment in a step, and the feedback sensors needed to read these motions properly.
Certainly there are lifting robots which can hoist many tons, so the 5.0kg/hand weight limit seems skimpy - but not when considering that this machine can allegedly walk and perform complex arm manipulations while holding this weight. Sadly, its continuous runtime before recharge is only 30 minutes, but I suspect later versions will take advantage of increasing innovation in charge/weight ratios in batteries, and perhaps solar panels for space use (an obvious application of these robots would be EVAs for the ISS or other craft).
Cool, in terms of integration with other systems, is the use of wireless ethernet as the comm standard, rather than some proprietary system. This means this robot could be controlled by base-station systems of arbitrary complexity - including a Beowulf cluster running a complex AI system like Webmind. This means that while technology is not quite there yet to put any advanced computational intelligence inside a biped robot, it can be controlled by advanced systems running at fixed-position stations through LAN technology - a good compromise in terms of merging the state-of-the-art in Robotics with AI to try to build towards a better convergence.
Regarding the robustness, it appears that the 25 minutes of runtime is the primary limitation in terms of continuous operation - there is no data I was able to find on failure rates or the fault tolerance of the sensors or computational systems on-board.
As for hacks vs. new general purpose algorithsm... They obviously do not reveal tremendous amounts of details, but suffice it to say that the engineering done to build 3 successive models of bipedal robots that can walk and climb (stairs, hills) represents fundamental work in robot dynamics engineering which, while parameter tweaked for this robot's operations, is certainly applicable (with some tweaks or modifications, as with all engineering techniques) to other bipedal robot applications.
The wireless lan comm technology, improved user interface (over the previous version), and sensor systems are all also certainly reusable in similar robots (indeed, likely also in multi-legged robots).
However, as it is a commercial product AND I do not read Japanese, I was not able to find any papers on specific algorithms to give a more detailed analysis...
Here are some useful resources I did find:
The official site in English
An article about the robot's walking functions
Images of the robot at UIUC
Biped Robots in General
Robodex Robotics Conference
o/~ we are pissed, we are pissed, we have to resist... o/~ - ec8or
But, on a more serious note: What happens if Honda gets to the point where they produce a robot capable of digging wells and planting crops in underdeveloped, potentially hazardous countries? What happens if Honda manages to develop a robot that can act as a lab assistant, routinely, tirelessly and accurately performing a wide range of menial yet necessary tasks in a lab environment? Sure, it's too advanced to be able to accurately predict something like this coming along today, but the very essence of science is the pursuit of being able to do and understand things we currently cannot do or understand. Today, this thing is just a prototype that can saunter across a studio floor. Another fifty years, it may be able to run for weeks on end with minimal human intervention, able to do a huge array of tasks that would require fairly heavy human intervention today.
Just as the pioneers of computing could have no way of accurately predicting the impact computers would have on the world fifty years later, the Honda engineers can have no way of knowing what will become of their robot in another 50 years. Rest assured, though, that they, like the computing pioneers, are serious scientists and serious dreamers...
$ man reality
Obliteracy: Words with explosions
KILROY!
But can it flail its arms around and say "Danger! Danger Will Robinson!" That is the true mark of a great robot.
But can it chew gum at the same time?
The Moore-Murphy Law: The number of things that will go wrong will double every 2 years.
Answers on a postcard please . . . :)
Any sufficiently advanced technology is indistinguishable from a rigged demo
--Andy Finkel (J. Klass?)
Well, you're right, it's seems easy. But where will you get this magical 'balance sensor'?
Aceleration/tilt sensors are used very often, sometimes a gyro, and there has even been talk of using an internal pendulum system for balance feedback. Using a system like this is a fairly good way of getting a walking biped. The problems arise when you don't have a level surface for your biped to walk on.
To perfectly mimic human walking, each leg would need 6 or more degrees of freedom. 2 for the hip, 1 for the knee, 3 for the ankle.
So imagine you have a bipedal robot that has a correct walk including hip movements for the transfer of balance between each leg. Now tilt the surface by a few degrees. Suddenly one leg needs to move less than the other, suddenly the ankle needs to change it's orientation so that the foot strikes the ground correctly, etc.
Now take this scenario to a rapidly changing surface such as a hill or rocky terrain. You can see how bad it could get.
But you would think that if you have a correct walking algorithm, no matter what surface you're on you would be able to adjust it for a good walking gait. Unfortunately, the robot needs some way of knowing what it will be stepping on before it ever does. It also needs to react quickly to feedback from sensors on it's feet and legs to adjust to anything it's previous scan didn't pick up correctly.
Because of the many factors, digital walking still has to get to the point where the computer can do all of this on the fly.
I think right now analog walking is the way to go. Analog creatures already walk quickly and easily on 4 legs, and it's easier to adapt them to two legs. It's just waiting on the right kind of sensors. My 4 legged bots are doing well, but I don't have enough money to work on a biped right now.
Oh well... hope this might help.
____
In an article off yahoo it appears that Sony too has a walking bot.
As x approaches total apathy I couldn't care less.