Something that's always impressed me about Japanese robotics efforts is the level of polish in the efforts. Complete cases (skins) on the robots makes them feel much more advanced than the non-Japanese efforts I've seen. It's this thinking about industrial design, along with the awesome technology that makes something like ASIMO feel like an almost commercial off-the-shelf technology.
I've like to know much more about the software functions of the robot as well, which sadly, they don't go into much during the demos.
until this year ASIMO, japanese robots felt naive compared to what's been coming out of boston dynamics | big dog team (which are on the opposite of the 'off the shelf look' spectrum). The latter being awfully uncanny in the depth of balancing behavior combined with the human scale proportions.
What's the state of the art in humanoid or multi-pedal robots in terms of robustness and flexibility of their movement and balance? I assume ASIMO isn't as stable and stress resistant as the robots of Boston Dynamics, for example http://www.bostondynamics.com/robot_petman.html
Asimo seems quite a bit more advanced overall than those robots. They are designed for quite specific purposes and domain - it's hard to tell but they also don't seem autonomous. With Asimo, Handa is going for a general purpose robot that can emulate a human while Boston Dynamic appears to be going for more specific military needs with things like BigDog.
From what I learned from a BD engineer, Japanese legged robotics use mostly static equilibrium in contrast with the (more advanced?) dynamic balancing of US and European robots. Asimo might look cute, but if you compare the gait of any of the popular MIT, Boston Dynamics or CMU robots to the Japanese, you soon notice that the latter don't look very natural or adequate to unpredictable terrain.
That's true enough, but only the military and EMS need robots that can instantly handle unfamiliar terrain. Meanwhile, the Japanese are most of the way towards a consumer product, since the interior of a building is highly predictable.
Good point, but if you don't need to handle unstructured terrain, you might just as well put the robot on wheels. The humanoids robot Justin from DLR [1] for instance has an omnidirectional wheeled platform that even allows for adaptive foot print. The only current advantage for ASIMO that I see at the moment, is that it doesn't have to take the elevator but can use stairs.
Running, jumping on one foot while turning 90 degrees. I think that is just remarkable stable.
Take a look at this slowmo-footage: http://www.youtube.com/watch?v=Ifl55QShWfE
It looks like the legs are constantly adapting. But ofcourse this could be programmed.
Although the Boston Dynamics robots seem more fluid and may have more flexible and natural overall movement, the ASIMO seems more autonomous. I haven't seen any of the Boston Dynamics robots actually performing a task adaptively (other than adapting to terrain and pushes/kicks). PETMAN looks really interesting but until it is turned loose untethered, ASIMO wins the best android contest.
Boston Dynamics robots like Big Dog use hydraulic actuators [1] which act somewhat like a spring and are hence very robust towards perturbations. They also have more power (= force*velocity). ASIMO in contrast uses an electric motor with a high ratio gear drive [2] that makes the robot very stiff and slow while allowing for high torques. This setup doesn't allow for backdrive and hence can't handle impacts well. I don't know about the new ASIMO legs but in the video you can still hear the screaming of the gears.
The way it's knees are overly bent is kinda creepy to me, especially when it runs.
But standing still it seems pleasant enough.
Also they need to code some imperfection into it - the absolutely perfect hand movements, unscrewing the jar and pouring are too precise - make it warble a tiny bit like a human would, make it shake the container at the end to get the remainder out.
The reason for this, is their control algorithm. It uses a inverse kinematic to compute the desired motor commands in the joints. When the legs are stretched out, it is in a singular position where the resulting motor gains would be infinite high.
The way it manipulates stuff with the hands is very impressive, not to mention the 'understanding' required to pour liquid like that. Wonder how much of that was scripted.
Excuse me, but "mincing gate?" Mispelling of "gait" aside, is that really an appropriate comment to make in this day and age? I'm getting a bit bored of feeling like the HN inappropriateness police. If it's not sexism, it's homophobia... are we going to have a fun time of racism next?
So, I'm sexist and homophobic? I guess I have to be the hyperbole police now. And racist? Rickshaw are common across half the planet, and a dozen 'races'.
The highest criticism I would consider fair is, Rickshaw-ist. I cop to that. I think they are emblematic of a class system, similar to sedan chairs.
If 'coolie' is considered rude, I apologize. I meant it in the sense
coolie definition: an unskilled laborer or porter usually in or from the Far East hired for low or subsistence wages.
If some urban slang definition exists that is derogatory, be assured I did not mean it that way.
I've like to know much more about the software functions of the robot as well, which sadly, they don't go into much during the demos.