The marvellous limp
Steve told me a story about Honda's ASIMO robot, I don't know if it's true or perhaps applies to another humanoid robot, but it sounds entirely plausible. The robot was being shown off at some expo or another; in particular, its ability to climb stairs. Halfway up, one of the servos in a leg froze, and it fell over.
Then I thought: wouldn't it be amazing if someone designed a robot that could limp?
This isn't as stupid as it sounds at first blush. Limping, when you think about it, is an amazing behaviour. It's the ability to circumvent almost any non-critical musculoskeletal damage or defect in the legs and pelvis (chassis?) and keep going.
It is not something we think about consciously, which is amazing. If you get a rock in your shoe and don't have time to take it out, you will automatically alter your gait to put less pressure on that part of the foot. You can sprain your ankle and thereby limit its range of motion and the load which that leg can bear, or you can even fracture an important structural element like the foot or the tibia and if the situation is important enough, you will keep walking.
Even something as small and stupid as an ant, if you break off one of its legs, will go where it wants to go instead of wobbling around in a circle as a six-legged robot would.
OK...I just Googled it...robot limping has been looked into. We really need to know more about this if robots are ever to function usefully "in the wild". ASIMO will not be useful for taking care of housebound old folks if itself is also prone to falling down the stairs and lying there helpless.
Labels: science, technology
posted by xenobiologista @ 23:31
2 Comments:
Machine learning rocks. I found out recently that they're beginning to use neural networks to study the debris from particle collisions. It's still in the early stages, but I think the whole aesthetic of having the machines figure out on their own appropriate algorithms for how to move and interact with the world is particularly attractive. If you know John Searle's 'Chinese Room Argument' against intelligence in digital computers, I consider this a strong rebuttal.
BTW, did you ever see the video of a DARPA project - a robotic mule? It traverses tough terrain pretty well, and can recover from a well-placed kick to the rear.
It occurred to me... rudiments of intelligent behaviour, i.e. figuring out based on physical characteristics just how a creature should move, are going to be featured in the game Spore. Computer games - driving the development of computing!
I do know of the Chinese room scenario. It's sort of like the Turing test backwards in a way. The problem with it is, the people standing outside the room have NO WAY of determining that the dude inside the room does definitely not know Chinese.
Re computer games - if you're interested in artificial intelligence, go and download Creatures: Docking Station NOW. The Creatures series (especially 2 and 3) are some of the most amazing and under-reported games ever made. DS is a free demo, smaller version of C3 that runs on the same engine.
On the face of it, Creatures looks like a cute virtual pet game. Actually the creatures in the game have simulated brains and biochemistry. The company that made the games also released some development tools (gene editor, realtime brain activity viewer, etc.). Back before C3 came out, some players used the tools to create a breed of creature that was actually smarter than the stock Norns in C2.
I seriously think it's amazing that these people produced a neural network creation engine and released it as a kids' game. C2 and C3 came out in And it's practically criminal that physiology and psychology professors don't use it as an undergrad teaching tool.
Post a Comment
<< Home