Real-Time Physics Simulation Forum

Even for ray cast wheels, a nonholonomic constraint is usually used, except that the lateral constraint force (friction) has a limited magnitude to allow for sliding. These constraints don't need special treatment if the engine can handle lo-hi limits on the Lagrange multipliers. I'm guessing Doom3 ...

It is odd that they are using biconjugate gradient when conjugate gradient would do. They appear to be restarting the BCG whenever the active set changes. This might be inefficient. A good PCG algorithm should work on the whole system, not just the active set. Then the PCG does not restart and needs...

I have implemented this algorithm. I have to say that it is a nice idea. You have to be careful to implement it correctly. A tricky case is when the line segment originates on the side of a box. Shorter segments are faster than longer ones. Long segments have O(N) complexity where N is the number of...

The Novodex guys describe how to find an approximate convex hull for the contact manifold in their Game Programming Gems 4 article. Basically you find the min-max contact points along two orthogonal axes in the contact plane. I use a simlar technique so that a table with four boxes for legs only get...

I think tesselation is not a bad thing. You automatically get rolling resistance. This was done with barrels in Half Life 2.

Perhaps there are some differences in our implementations so that we are seeing different results. In your dissertation, you describe a method for applying velocity constraints to contact points that are not actually touching. Perhaps this would eliminate some of the jitter I was seeing from the ove...

Here's what I do for box-box collision: 1. Find the separating axis with minimum overlap. In this determination, I bias against edge-edge contact. 2. If the minimum axis is edge-edge, I just generate one point. 3. If the minimum axis is face-(vert | edge | face), I determine the reference face accor...

Hi Kenny, there are two aspects to weight feeling. One is to have a seesaw that works properly. I think this is handled well by a dual stage integration with regular dynamics followed by shock propagation. The second is multilevel stability. Consider a stack of three boxes arranged as: http://www.gp...

Here's what I'm doing in our game (keep in mind this has to run on a PS2): 1. Projected Gauss-Seidel (PGS) with 5 iterations and Baumgarte stabilization. 2. Two-way anisotropic friction with a constant bound. 3. 30Hz simulation (can dip to 15Hz) 4. To prevent tunnelling I use line probes and swept s...