Subspace Neural Physics: Fast Data-Driven Interactive Simulation

24/07/2019

This year at SCA I will be presenting some research based around accelerating specific physics simulations to the point where they can be used in interactive applications like video games. The core idea of the approach is to combine neural networks with subspace simulation to produce a simulation step which can run entirely in the simulation subspace, while still being able to interact with external objects. To do this we train a neural network which effectively learns to model the computation of forces inside the subspace and demonstrate it on a number of different scenarios including cloth simulation and deformable objects where it produces performance gains ranging from 300 to 4000 times.

Webpage • Paper • Video • Article • GDC Talk

Abstract: Data-driven methods for physical simulation are an attractive option for interactive applications due to their ability to trade precomputation and memory footprint in exchange for improved runtime performance. Yet, existing data-driven methods fall short of the extreme memory and performance constraints imposed by modern interactive applications like AAA games and virtual reality. Here, performance budgets for physics simulation range from tens to hundreds of micro-seconds per frame, per object. We present a data-driven physical simulation method that meets these constraints. Our method combines subspace simulation techniques with machine learning which, when coupled, enables a very efficient subspace-only physics simulation that supports interactions with external objects - a longstanding challenge for existing subspace techniques. We also present an interpretation of our method as a special case of subspace Verlet integration, where we apply machine learning to efficiently approximate the physical forces of the system directly in the subspace. We propose several practical solutions required to make effective use of such a model, including a novel training methodology required for prediction stability, and a GPU-friendly subspace decompression algorithm to accelerate rendering.