Abstract
Dynamic simulation of elastic bodies is a longstanding task in engineering and computer graphics. In graphics, numerical integrators like implicit Euler and BDF2 are preferred due to their stability at large time steps, but they tend to dissipate energy uncontrollably. In contrast, symplectic methods like implicit midpoint can conserve energy but are not unconditionally stable and fail on moderately stiff problems. To address these limitations, we propose a general class of numerical integrators for Hamiltonian problems which are symplectic on linear problems, yet have superior stability on nonlinear problems. With this, we derive a novel energy-controllable time integrator, A-search, a simple modification of implicit Euler that can follow user-specified energy targets, enabling flexible control over energy dissipation or conservation while maintaining stability and physical fidelity. Our method integrates seamlessly with barrier-type energies and allows for inversion-free and penetration-free guarantees, making it well-suited for handling large deformations and complex collisions. Extensive evaluations over a wide range of material parameters and scenes demonstrate that A-search has biases to keep energy in low frequency motion rather than dissipation, and A-search outperforms traditional methods such as BDF2 at similar total running times by maintaining energy and leading to more visually desirable simulations.
The Stability-Energy Trade-off
In computer graphics, simulating squishy, elastic objects—from bouncing rubber toys to complex character tissues—requires a delicate balance between numerical stability and physical energy. Standard industry tools like Implicit Euler are favored for their robustness; they allow for large time steps without the simulation "exploding." However, they suffer from significant numerical dissipation, effectively draining energy from the system and making objects appear "dead" or overly dampened.
Conversely, symplectic methods (such as the Implicit Midpoint rule) are designed to conserve energy. While they produce more dynamic results, they are notoriously unstable when encountering stiff materials or complex contact scenarios, often leading to catastrophic numerical failure. The challenge lies in creating a "best of both worlds" integrator: one that is as robust as Implicit Euler but allows the user to decide exactly how much energy should be preserved or lost over time.
The A-search Integrator
To bridge this gap, we introduce A-search, a new class of numerical integrators that provides animators with precise control over energy dissipation. A-search is built upon a "linearly symplectic" base method (A-1) that perfectly conserves energy for simple systems while remaining unconditionally stable.
The A-search algorithm operates in two primary phases:
- Implicit Update: It performs a standard Implicit Euler step to determine new positions. This ensures the simulation inherits the stability and geometric guarantees of Incremental Potential Contact (IPC), remaining intersection-free and inversion-free even under extreme stress.
- Velocity Correction: It then performs a "search" for a blending parameter, $\alpha$, to interpolate between the Implicit Euler result and an adjoint method. This correction allows the simulation to hit a specific user-defined energy target. If a user wants perfect conservation, the algorithm finds the $\alpha$ that keeps total energy constant; if they want a natural decay, they can specify an energy half-life.
Results and Visual Fidelity
A critical advantage of A-search is its "low-frequency bias." The method naturally preserves energy in large-scale, visible motions (like the primary bounce of a character) while allowing high-frequency numerical "noise" to dissipate. This behavior aligns with human visual intuition, where we expect to see the macro-level dynamics of an object without being distracted by microscopic vibrations.
In comparative tests like the "Dice Roll" and "Bouncing Ball," A-search maintained motion long after BDF2 (a common industry standard) caused objects to come to a premature rest. While finding the optimal $\alpha$ adds a small computational cost per step, the method often requires fewer Newton iterations than higher-order dissipative methods. At similar total runtimes, A-search produces much more dynamic and visually interesting results.
Conclusion
A-search offers a robust solution for artists and engineers needing both the stability of implicit methods and the vibrancy of energy-conserving ones. By decoupling the position update from the momentum correction, it provides a flexible framework for energy-controllable physics that integrates seamlessly into modern collision-handling pipelines.
Summary updated