Optimization Integrator for Large Time Steps.

Area of Science:

• Computer Graphics
• Computational Physics
• Numerical Simulation

Background:

• State-of-the-art simulators use Newton's method for nonlinear equations.
• Newton's method struggles with large time steps and stiff simulations.
• Current methods are unreliable at high frame rates.

Purpose of the Study:

• To stabilize Newton's method for large time step simulations.
• To improve the reliability and speed of physics simulators.
• To integrate collision handling efficiently within the solver.

Main Methods:

• Recasting the backward Euler method as a minimization problem.
• Applying standard and novel optimization techniques to stabilize Newton's method.
• Incorporating collisions via constrained minimization and novel penalty formulations.

Main Results:

• The developed solver handles difficult simulations at high frame rates ([Formula: see text] and beyond).
• Collision detection and response are efficiently integrated.
• Material Point Method (MPM) simulations show improved behavior.

Conclusions:

• Recasting backward Euler as a minimization problem enhances solver stability and performance.
• The method allows for efficient, high-frame-rate simulation of complex physical phenomena.
• This optimization-based approach offers a robust solution for challenging simulation scenarios.