Subject:
|
Re: Optimising piece use (Was: LDraw.org MOTM/SOTM voting for March is open)
|
Newsgroups:
|
lugnet.cad, lugnet.cad.dev
|
Date:
|
Fri, 29 Mar 2002 06:57:59 GMT
|
Viewed:
|
924 times
|
| |
|
|
Cool. I like the parallel-paths-to-ground idea that you'd get
with conductance and studs as resistors. I can see large 1x16
"ground plane beams being used, likewise maybe wall pieces.
Another interesting aspect is surface gradients, and the use
of sloped or curved bricks. Any human could pick off 1x1 plates
until the surface matched what was desired, but a good computer
tool could dither a bunch of plates/slopes/widgets to better
approximate a surface or texture, similar to how those picture
mosaics-made-from-other-pictures were so fascinating.
In lugnet.cad, Jacob Sparre Andersen writes:
> I wrote:
>
> > Paul Gyugyi wrote:
> > > I've often thought of doing a tool that starts with a big
> > > cube of 1x1 plates and removes all but the ones inside
> > > a closed mesh. It would be really cool if you've done that.
> >
> > That's more or less what I have done. The problem is to
> > automate the replacement of 1x1 plates with larger pieces in
> > a sensible way.
>
> I spent some time thinking about this yesterday. If we look at
> the problem of substituting the 1×1 plates with larger pieces as
> an optimisation problem, then we have to define a good "energy
> function" to describe the quality of various piece combinations.
>
> One element of the energy function should of course be the
> available pieces. Assume that we know the optimal distribution
> of pieces in a LEGO collection. If we were creating a model that
> LEGO should sell, then we would use the deviation of the piece
> distribution in the model from the optimal piece distribution as
> something that subtracts from the quality of the model. If we on
> the other hand are creating a model based on our own collection,
> it will of course be the deviation of the distribution of the
> remaining parts from the optimal distribution that we look at
> (plus a barrier when we run out of a piece).
>
> Another element of the energy function should be how strong the
> model is. A simple definition of the strength of a model can be
> made using conductivity as an analogy to strength. If we imagine
> that for every stud-hole connection (or some other kind of
> connection) there is between two pieces, they are connected with
> a resistor. The more resistors the higher the conductance. If
> we then measure the average conductance between randomly chosen
> pieces in the model, we will have a measure of how many parallel
> connections there are on the average in the model, and thus a
> measure of how strong the model is. - My first idea was to use
> the minimum and not the average conductance, but this would give
> many optisation methods problems, because they would be unable to
> distinguish "good" from "better", but only distinguish "better"
> from "best". But maybe the minimum conductance should count as a
> separate element, and not only as a part of the average
> conductivity.
>
> Does it make sense?
>
> Play well,
>
> Jacob
> --
> http://jacob.sparre.dk/LEGO/Ydre_rum/Skibe/Complexity/
|
|
Message is in Reply To:
6 Messages in This Thread:
  
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
