Subject:
|
Re: An Idea for new Mindstorms - event potential?
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Wed, 29 Jun 2005 02:59:36 GMT
|
Viewed:
|
1587 times
|
| |
|
|
> Overall, a common housefly probably has more available compute power than
> a modern desktop PC.
I'm not knocking the fly, I'm saying that landing on the ceiling is a very
difficult problem to solve by realtime calculated piloting (even using a desktop
PC). But it needs very little calculation at all if the problem is almost
entirely offloaded into the body design. The research I saw indicated that the
computational power the fly was applying to _this_particular_ task was very low.
The reason is that the task is better solved by body than by brains, so the
brain is barely involved. And the importance of this is easy to underplay when
building robotic.
I imagine the vast bulk of a fly's neurons would be related to processing sight,
yet from what I recall, you can blind a fly, effectively removing all those
neurons from the system, and it will still land on the ceiling without
difficulty. (While a desktop PC realtime piloting solution would be greatly
dependant on the instrumentation, if it even worked at all).
I'm saying "This stuff is really cool!", rather than "Flies are lame and only
saved by this other thing being well done" :-)
|
|
Message is in Reply To:
 | | Re: An Idea for new Mindstorms - event potential?
|
| Justin wrote: > A fly can land on the ceiling not (...) Don't be too hard on the fly's brain. A housefly has about a third of a million neurons. If you think of a neuron as being about the power of a transistor - then there is a computer that's more (...) (19 years ago, 29-Jun-05, to lugnet.robotics)
|
32 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
