This Robot Can Ski...Sorta
The robot invasion continues: snowy slopes are no longer safe from robots. Bojan Nemec from the Jozef Stefan Institute in Slovenia recently presented his skiing robot at the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). The robot won’t be winning skiing records anytime soon and its usefulness as a ski instructor or in any other capacity seems quite a ways off. Nonetheless, the videos of this robot taking a ride down the ski slopes are well worth watching.
As the robot slides down the snowy mountain we can see an impressive ability to self balance, change direction, and autonomously sense and navigate around race gates. A gyroscope, four sensors mounted between the skis, and motor position sensors support the lower control of stability and joints. A simple USB camera and a GPS system allow the robot to gather the necessary input to determine its position and calculate what action it should take next. My favorite feature of the ski bot is that it uses normal off the shelf human skis – the robot does not need to cheat by using any sort of custom made ski. Check out the introduction video below:
Impressive as the skiing robot may be in many ways, clearly this robot is a far cry from gaining human respect on the slopes. The robot is unable to conquer anything more complex than a four gate course with an easy gradient and no major pitfalls in the way. Even in this simple scenario the robot was prone to crashing or even running away, as we can see comically in the following videos:
Jokes aside though, you have to start somewhere, and I gotta hand it to Nemec for creating an interesting robot. Even in a field as random as robotic skiing there is apparently competition in the field. Check out the extensive documentation from Shiro Shimizu (University of Fukui) where another skiing robot has been developed. As the DARPA grand challenge has shown us, with the right motivation and incentive, humans are capable of propelling robots to master incredible tasks such as driving cars autonomously. Luckily for us though, incentives are being directed elsewhere, and we now have wonders such as the Da Vinci surgical robot instead of world class robotic skiers.
Tags: ski, skibot, skiing robot
Share
Everything was version 1.0 at some point, and 0.1 before that.
I’ve recently been reading about computer scientists enthusing about how GPUs have seen 100x or more improvements in very short amount of time (sub 1-year), blowing, as they said “moore’s law* out the window.” Technology has a way of going along steadily increasing, and then some crazy thing comes along and makes the old change of pace look sedate. When GPU (or whatever similar) technology is feasible to stick into a small bot like this and run on batteries, their capacity to deal with stimuli should jump quite a bit.
*moore’s law technically refers to transistor size/price (or performance?) doubling on a 18 month basis, but roughly equates to doubling of computing power in the same period of time. GPU technology just laughs at these ideas. It went from toy to supercomputer almost overnight. My video card, which I use to play games with, basically a toy, is more powerful than all but three of 1999’s most powerful supercomputers. And it’s not even the best you can get. It wasn’t even that expensive.
“GPU”
You keep using that word. I do not think it means what you think it means.
GPU (dedicated purpose) technology has always been dramatically faster than CPU (general purpose) technology due to the ability to optimize performance to a very restricted set of tasks. Moore’s law addresses transistor count and price (not that it’s very scientific even in that sense) where the GPU and CPU relationship with Moore still holds up under a more strict interpretation of it.
The dramatic advancement in the GPU area has come thanks to some novel new programming that blurs the line between GPU and CPU, but the GPU will always be limited to specific tasks otherwise it will have to take the performance hit that comes with being a truly general purpose CPU.
What were we talking about again?
Don’t minimize these single-task solution robotics projects; the robot will enter the mainstream when a Honda or similar company decides to go “all in” and unify these disparate projects into on 40 to 50k domestic multipurpose robot.
it’s sensor limitations too, not just onboard processing power that are keeping robots from being very good…
[...] http://singularityhub.com/2009/10/21/skiing-robot-not-very-useful-but-totally-fun-to-watch/ a few seconds ago from web [...]
Should it have a reference to a human? If so then its center of gravity is WAY to low and if not then why this design for a robot? Does not make any sense.
[...] interesting article, with videos, about a skiing robot. Pretty cool stuff. As the title says, it’s not terribly [...]
At least the thing looking cool as it swayed from side to side. Now we just need to let it loose down a ski slope alongside humans. LOL.
A robot that can TOTALLY out ski humans !!! That is compared to MOST human skiers.
In recent years I have learned the EASIEST way to turn. The researchers have accidentally made this robot to do the same thing.
The easiest way to turn is to push back just a bit, 2 inches, more and you turn too much, on the INSIDE ski, the inside of the arch of the turn that you wish to make.
It works so well and is sooo easy.
This robot does this when it bends the inside leg more then the outside leg to lean and turn. When it does that the inside ski gets pushed back just a little bit in relation to the outside ski and that is why the robot turns with soo much power and so little effort.
By the way, LEANING of the body DOES NOT MATTER, although leaning the ski DOES, its is pushing the ski back that makes it work.
Equilibrium of down hill momentum interrupted in an intelligent way sets the inside edge of the downhill ski to turning.
Although, if a skier or robot was airborne starting out with skies straight but then pushed back on the inside ski and tilted the skies would still turn with the inside ski closer to the radius albeit slowly.
Thus, the equilibrium of drag and relative positions of each ski is the MOST important factor.
Although, part of the MAGIC of this method is that pushing back the inside ski gives the skier’s legs greater flexibility and that is also WHY the robot moves this way, it is a mechanical compatibility between the necessity of the physics of turning and the mechanism of legs and hips of the skier.
Just because you fail to use your imagination it doesn’t mean the robot is not useful. What about rescuing people from an avalanche?
Now I’m just waiting for the day when I can tell my robot to make me some cheese omlets. The robot will see we our out of eggs and it has permission to go to the store (or contact the stores robots) and retrieve the eggs. and the procedes to do everything else to make my omelets nice and tasty. He will learn everything he needs to know about omlets and serving a human omlets from the internet and will also be able to download knowledge from other robots. or maybe just be remote controlled by the robot overmind.
I for one will welcome our robot overlords! Hail to the new world order!
Cool videos. I want one for a pet.
Wouldn’t that get a little monotonous, just Akron, cold beer and ‘poor, poor thing’ for two weeks?
[...] Shared Skiing Robot Not Very Useful But Totally Fun To Watch. [...]
[...] Skiing Robot Not Very Useful But Totally Fun To Watch (video) – Damn robots, stealing our freshies [...]
Skiing is a very demanding pastime. I enjoy it every time.
[...] original Website I found was this one, entitled Skiing Robot. But the originator of this project is Bojan Nemec from the Jozef Stefan Institute in Slovenia. [...]