Subject:
|
Re: show-me training
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Wed, 8 Sep 1999 17:55:32 GMT
|
Viewed:
|
496 times
|
| |
|
|
In lugnet.robotics, lego-robotics@crynwr.com (Pete Hardie) writes:
> [...]
> The real problem is that it is difficult to generalize the behaviors
> does the robot move forward until Button A is pressed, or just for
> 5 seconds? What does it do at the end of the recorded behavior? Etc.
Yes, I agree completely.
This wasn't meant to be a replacement for real programming -- just a good idea
for a kid's toy. It would probably make a great replacement for the Scout's
current interface. I think it would be a lot better, in fact, because it would
open the Scout to a younger set of children and thereby expand the market. It
would also easily replace the Micro Scout's current interface.
As for "does it move forward until button A is pressed or just for 5 seconds",
that would obviously depend on what type of mathematical model is embodied in
the software. If you check my web page:
http://www.mrob.com/lego/showme.html
you'll see that I mention statistical analysis, linear regression,
pseudo-random functions, matrix transforms, etc. Depending on what methods you
use, different types of behavior result.
It definitely would not ever replace the RCX Code or NQC approach. I've been
working on different formulas to try to figure out if you could even do a
single-sensor line follower. It seems likely it would require a nonlinear
correlation between the sensor and the motors. A two-sensor line follower, on
the other hand, looks pretty easy with straight two-variable linear regression.
> It's great for cyclic behaviors, or sporadic actions, but a
> line-following maze runner might be very hard to produce with
> this method.
Exactly. I'm not even sure how a line follower would run a maze (is it
following the walls or the line?) but I agree that trying to use "show-me"
training for that is inappropriate.
- Robert Munafo http://www.mrob.com/
LEGO: TC+++(8480) SW++ #+ S-- LS++ Hsp M+ A@ LM++ YB64m IC13
|
|
Message has 1 Reply:
 | | Re: show-me training
|
| (...) Ah, yes. And it would be a nice feature for behavioral 'subroutines', like collision reaction (on BUMP, steer left and backup) (25 years ago, 8-Sep-99, to lugnet.robotics)
|
Message is in Reply To:
| | Re: show-me training
|
| (...) This is in fact used for some industrial robots (welders, etc) on assembly lines. It also seems a lot like the motion control camera systems pioneered by ILM for Star Wars. The real problem is that it is difficult to generalize the behaviors - (...) (25 years ago, 8-Sep-99, to lugnet.robotics)
|
17 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
|
|
|
|
