-----Original Message-----
From: zero <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED] <[EMAIL PROTECTED]>
Date: Saturday, May 19, 2001 6:41 AM
Subject: Re: [JAVA3D] Contact Point
>
>summing up, the i have no problem using the WakeUpOnCollisionXXX.. cause
>with USE_GEOMETRY it works quite fine and if there is an intersection i
move
>back to the sate before - or an interpolated between - so intersecting
isn't
>noticeable for the viewer...
>
>but there is still my problem implementing a realistic physik-system
>(including torque,...), where the contact point is need for computing the
>new angularMomentum,...
>
>so i would be very glad, if someone has a solution.
Doing a high fidelity dynamics simulation of collisions between bodies with
arbitrary geometry is a very difficult problem. As stated above, Java 3D
will help with pre-screening by bounds or simple bounding geometry. It will
support whatever object model you build to solve the physics on the scene
graph, and maybe you can use its geometry data structures. Other than that,
you're on your own.
I've avoided looking at it seriously, and I've developed some
plausible-sounding rationales for my avoidance.
First, the problem you pose is famously chaotic, and simulated solutions to
problems in chaos are practically impossible to validate except by running
them a very long time and comparing their summary statistics to those of
reality . How can you speak of a high-fidelity solution when you can't
answer the next question "Fidelity to what?" For a chaotic system, it's
hard to claim that your hi-fi geometry-based collision system is 'better'
than an approximate solution, especially when the fast
good-enough-for-gaming approach gives up plausible summary statistics.
Second, there's the more interesting super-problem of structuring the scene
graph to do physics at all.
Third, Java 3D's primitive approach to time, clocks, and Behaviors doesn't
help much for running combined discrete-continuous models. A simulation
clock probably has to be designed and grafted on too. Because it has to be
grafted on, it figures to be slow.
I'll be in the home long before I get around to hi-fi collisions in Java 3D.
Beyond all that encouragement, good luck with it, and please report your
progress here.
Cheers,
Fred Klingener
Brock Engineering
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
