Starlight Posted January 8, 2006 Posted January 8, 2006 :icon_syda your post Roman. Now i can see the wheels turning. It just goes to show if you think you can or you think you can't you are right. The concept is not much different from the LOD-mesh concept used in most of 3D applications (LockOn already included). The refinement process is also used in numerical analysis and other applications, whenever an error is introduced because of a discretized/numerical representation of a [continuos/symbolic] model (whether physical or mathematical). The basic concept is, if your representation unit isn't satisfactory for your purposes, you split it into smaller units, until you get the desired precision. Me and some my colleagues at University made some months ago an adaptive (dynamic) model for Finite Elements Method, which was applied to 2D and 3D functions. in the 3D case it was something like that you linked at wikipedia, basically a dynamic mesh refinement. Maybe "Subdivision surface" is a bit faster than other methods, but it's often difficult to find the "always best" method when programming. Programming is usually an art to find the most satisfactory tradeoff or balance, that is why is mostly regarded as an art, not a science ;)
Caretaker Posted January 8, 2006 Posted January 8, 2006 Traditionally dynamic mesh algorithms work poorly for object models; it's usually better to provide predefined LOD models and switch at a certain distance. For terrain rendering though it's already a standard (used in Lock On as well). As for reinventing the wheel, this is what Michelin proposes for the future: http://www.motorintro.com/reportajes/_imagesreportajes/michelin_tweel/michelin_tweel07.jpg http://www.geocities.com/tiensoon_law/images/blog_michelin_tweel_02.jpg Seems to have its advantages (higher resistance to damage as no air can be lost); too bad it looks ugly ;) Caretaker ED Beta Test Team
golfsierra2 Posted January 8, 2006 Posted January 8, 2006 I can reinvent the wheel. Then I'd have to reinvent the ground. And so forth. I'd make a 50 pointed shape, and make the ground zigzaged so that each point rests neatly on the ground. As I wrote before: There's something about a wheel which is more simple than any other shape. Its edge is EXACTLY the same distance from the centre on any angle. I try to think outside the box. :rolleyes: Well, what makes a wheel so interesting is it's ability to roll with almost no resistance. But that is relative. Imagine a gear wheel instead, which's tooth exactly fit to the shape of a triangle rows shaped surface... >>Thinking outside the box II<< kind regards, Raven.... [sigpic]http://www.crc-mindreader.de/CRT/images/Birds2011.gif[/sigpic]
Roman G Posted January 9, 2006 Posted January 9, 2006 Wow, that bicycle is really unconventional ! ... but, somehow I feel it would not perform very well if you also need to go other route than perfectly straight ... :)
Roman G Posted January 9, 2006 Posted January 9, 2006 Traditionally dynamic mesh algorithms work poorly for object models ... Traditionally people were fine if wheel had eight angles. Traditionally meshes had less than 100,000 triangles. If you want more triangles then the amount of data going through bus is becoming too big. Traditionally dynamic meshes were (and probably still are) not supported by hardware. We are talking about what future might be ... ... For terrain rendering though it's already a standard ... Not sure why this could not be extended to a few objects which are very close ...
zaGURUinzaSKY Posted January 9, 2006 Posted January 9, 2006 Or, spin the user abou thte monitor real fast! Now THAT's simulation. Not only do you SEE the wheel, you FEEL the wheel! lol Robbie.
Recommended Posts