Cloth simulation based on neural network regression
Heng Zhu, Zhan Gao, Lei Xu · 2023
The article proposes a fabric simulation method that integrates a BP neural network regression model, which improves collision detection efficiency and reduces computational overhead. Based on data computed using traditional implicit integration algorithms, this approach utilizes a BP neural network to establish a model for predicting the coordinates of particles in the next time step. Subsequently, an interpolation algorithm is used to increase the quantity of fabric particles, repeating this interpolation process multiple times until achieving a finer fabric simulation. Ultimately, the study demonstrates that the integration of the BP neural network-predicted fabric simulation model not only conserves significant computational resources but also optimizes operational efficiency. This integration yields stable animations, establishing itself as a reliable and efficient fabric simulation approach.