![]()
Visiting guest. Why not sign in?
|
|
Real 'Robot' Wars Learning From Experience | |
Real 'Robot' Wars
There's an extremely interesting article over on BBC News named Another especially relevant thing to point out are the solar panels which power the bots. This allows experiments to run for days (in theory, but this is england ;) That can be a major advantage over standard robot experiments, where you only have 6 hours before the batteries run down! |
|
Bid to Breed Free-Thinking Robots
This is London is running a story on the exact same event: Bid to Breed Free-Thinking Robots. The article describes the experiments a bit more: the prey have to find the right balance between charging their batteries in the white light, or fleeing from the predators. |
|
Thinking Robots Go to War in Fight for Survival
The Guardian also has a feature on the same event: Thinking Robots Go to War in Fight for Survival. In this article, the differences in structure of the robots is highlighted as a crucial element. I'm still wondering why they didn't use reinforcement learning! |
|
Bots Battle, Breed in A.I. Test
Yet another article on the same exhibition. This one's over on Wired and seems by far the best so far: Bots Battle, Breed in A.I. Test. It goes into much more details about the actual settings of the experiments, and gives the little details I'm very fond about ;) On the technical side, the article hints that the Neural Networks may learn during their life-time. Now I'm not sure if that's claiming that NN are capable of learning within their lifespan, or if they actually do in this case. Typical GA based on Darwin's theories tend to not allow life-time learning, but Baldwinian and Lamarkian schemes do... It would be interesting to find out more about that. |
|