Real-Time Physics Simulation Forum

Hi,

I've been implementing a physics engine for the research in my niche application, using the impulse-constraint method and this document:
http://danielchappuis.ch/download/Const ... Body3D.pdf

Now, I was wondering about using higher order integrators in order to reduce the number of computations needed to reach physical relevance (rather than visual accuracy and stability). I have a couple of questions in this regard maybe some people on this forum know the answer to?

• In the spring-mass type of algorithms, it is not unusual to go to higher order Runge-Kutta methods for integration to have more stability. In the impulse-constraint method, I cannot seem to find any resources on this topic? Is there a specific reason for this?
• How would you use a Runge-Kutta method on impulse-constraint exactly? The result of the every update step are your new velocities, not the forces/accelerations. To find these new velocities, you need the velocities (and positions) of the previous time step as well. I fail to push this into the Runga Kutta integrator.
• Is there an open-source library with other integrators of which I could study the code?