533DAnimationPhysics.ppt

cs542g-term1-2006 cs533d-winter-2005 Notes Even if you’re not registered (not handing in assignment 1) send me an email to be added to a class list Time Integration Recall Taylor series Linearization in higher dimensions Reduction to first order systems Well-posed problems: problem stability Linear analysis Forward Euler Accuracy: truncation error vs. global error Algorithm stability Test Equation Stability region More accurate explicit methods Leapfrog Multistep/multivalue methods Runge-Kutta methods Heun’s method (RK2) Standard RK4 Error control Adaptive time stepping Integrator pairs Stiffness A problem is stiff when an explicit integrator needs to take small steps for stability, not accuracy Solution space includes very rapidly vanishing transients Note: not stiff just because you want to take large time steps If the real solution has variations on a small timescale, there’s no avoiding small time steps Usually handled with implicit integrators Backwards Euler * *