MIT’s new simulation brings physically accurate bouncy, squishy characters to life

Simulating soft reality

Simulating the bounce of a rubber ball or the stretch of a cartoon character may seem simple, but achieving realistic elastic motion in animation is a longstanding challenge. Many existing methods rely on fast solvers that sacrifice physical accuracy for speed, often leading to energy loss or unstable simulations that break down entirely.

The goal of more precise methods, such as variational integrators, is to maintain important physical characteristics like momentum and energy. However, the usage of these techniques in animation workflows is limited since they entail intricate equations that are challenging to solve well and consistently.

Researchers at MIT have created a novel technique that gets around this obstacle by exposing a hidden convex structure in the variational integrator equations. After separating the stretch and rotation components of elastic material deformation, they found that the stretch component creates a convex optimization problem. Convex issues are easier to handle and can be resolved more consistently with algorithms that provide convergence guarantees.

According to MIT News, this method helps animators avoid frequent problems like energy loss or simulation failure while producing stable, physically realistic simulations of elastic materials. The team’s discovery demonstrated that hidden convexity might also apply to dynamic systems, opening the door for more realistic animation of elastic and bouncy materials. Previous research had examined hidden convexity in static circumstances.

Reliable motion modeling

In experiments, the MIT team’s solver successfully simulated various elastic behaviors—from bouncing objects to squishy, deformable characters—while maintaining key physical properties and long-term stability.

In contrast, many existing methods struggled under similar conditions: some became unstable, producing erratic motion, while others exhibited noticeable damping that made animations look unrealistic. “Because our method demonstrates more stability, it can give animators more reliability and confidence when simulating anything elastic, whether from the real world or even something completely imaginary,” Silva told MIT News.

Although the MIT solver isn’t as fast as simulation tools prioritizing speed over accuracy, it offers a key advantage: reliability without the usual trade-offs. Unlike many physics-based methods that rely on complex, nonlinear solvers prone to instability and failure, this approach maintains accuracy while remaining more robust and easier to manage.

Looking ahead, the researchers plan to focus on improving computational efficiency to make the method faster without sacrificing stability. They’re also interested in extending their use beyond animation.

The team claims that the technique has potential for practical engineering and design applications, including producing toys, clothing, and other flexible materials where correct deformation modeling is essential, due to its capacity to precisely and consistently reproduce elastic behavior.

“We were able to revive an old class of integrators in our work. My guess is there are other examples where researchers can revisit a problem to find a hidden convexity structure that could offer a lot of advantages,” Silva told MIT News.