Subject:
|
RE: A Generic Idea
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Mon, 6 Jan 2003 05:39:06 GMT
|
Original-From:
|
Andrew J. Huang <{ajhuang@velocitus}StopSpam{.net}>
|
Viewed:
|
1029 times
|
| |
|
|
Steve Baker summarized neatly:
The big design issue is whether to have the brain
near or far. If far, you need more communications
bandwidth, but you can have a mongo brain. If
near, you can minimize your investment in
communications overhead, but you have to live
within the constraints of CPU that you can support
on board. Both designs have their advantages.
The problem you're solving will make the decision
for you. Neither solution is generic because
different problems will need different solutions.
If the problem is to build a football goalkeeper,
the ideal solution is local - the computation is
reasonable (detect and track black & white moving
object) and reaction speed is important. A remote
brain is feasible but the required investment in
communications bandwidth probably exceeds the cost
of a local brain.
On the other hand, if the problem is to build a
striker for the same game, you would need both the
local brain as well as the communications channel
to synchronise plays with the teammates.
On the third hand, if the problem is a bomb
handling robot, then you would really want purely
remote control by a human capable of
differentiating between an innocent plastic RCX
brick and a plastique brick.
A few other comments on this thread:
It looks like we are reducing the RCX to simply a
smart I/O controller which is slaved to a more
capable CPU. If that's what it boils down to, we
can simply buy a cheap notebook ($400-$900) and
then use it's parallel port to drive an I/O
controller (I saw such an I/O controller many
years ago, for $300?). This would replace the
"brain" and the RCX, and move the sensor/output
limitations up to 8 or 16. Technically, it's
trivial. Finding the pieces available
commercially is not so difficult. If you really
want to be generic and do some fancy robotics,
throw out the RCX and get real.
Regarding PDAs and cell phones: even those with
cameras do not have the CPU power to do image
analysis. They are only really capable of storing
the data after the hardware-based compression
algorithms have reduced the pixels to a manageable
lump. The analysis needed to detect motion or
recognize objects is way beyond their
capabilities. PDAs have roughly the CPU power of
a 1990s PC - think back to 1990 and recall what
kind of graphical manipulation you could do
then - it wasn't much and it entirely consumed
your CPU while it took 30s to think about it.
-andy
|
|
Message has 2 Replies:
 | | Re: A Generic Idea
|
| (...) I summarized neatly too, while he got into the intricacies of DTMF and Phreaking in the 70s... :D (...) It wasn't a generic solution, it was a generic idea. The generic cel phone would provide a generic link to any generic PC in the world. You (...) (22 years ago, 6-Jan-03, to lugnet.robotics)
| | | Re: A Generic Idea
|
| (...) My aging Agenda PDA does integer work at about the speed of a 100MHz Pentium. Floating point performance is terrifyingly slow though. Since the Agenda is a couple of years old now, I'd expect PDA's to be two or three times faster by now - so (...) (22 years ago, 6-Jan-03, to lugnet.robotics)
|
Message is in Reply To:
| | Re: A Generic Idea
|
| (...) Yes - but when they install ISDN or DSL on your line, they switch out some electronics at the telephone exchange - remove some line filtering, pull out the digital part of the stream from the analog part, etc, etc. If you just have the audio (...) (22 years ago, 5-Jan-03, to lugnet.robotics)
|
38 Messages in This Thread:
 
  
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
Active threads in Robotics
|
|
|
|
|
|
