Thermodynamics-informed neural networks for physically realistic mixed reality

Quercus Hernández (1) , Alberto Badías (2) , Francisco Chinesta (3, 4) , Elías Cueto (1)

1 I3A - Aragón Institute of Engineering Research [Zaragoza]
2 UPM - Universidad Politécnica de Madrid
3 PIMM - Laboratoire Procédés et Ingénierie en Mécanique et Matériaux
4 ESI Group - ESI Group

The imminent impact of immersive technologies in society urges for active research in real-time and interactive physics simulation for virtual worlds to be realistic. In this context, realistic means to be compliant to the laws of physics. In this paper we present a method for computing the dynamic response of (possibly non-linear and dissipative) deformable objects induced by real-time user interactions in mixed reality using deep learning. The graph-based architecture of the method ensures the thermodynamic consistency of the predictions, whereas the visualization pipeline allows a natural and realistic user experience. Two examples of virtual solids interacting with virtual or physical solids in mixed reality scenarios are provided to prove the performance of the method.

Mots clés: Deep learning; Augmented reality; Thermodynamics; Structure preserving; Real-time simulation

Domaines: Mécanique des solides [physics.class-ph]; Synthèse d'image et réalité virtuelle [cs.GR]; Apprentissage [cs.LG]

Quercus Hernández, Alberto Badías, Francisco Chinesta, Elías Cueto. Thermodynamics-informed neural networks for physically realistic mixed reality. Computer Methods in Applied Mechanics and Engineering, 2023, 407, pp.115912. ⟨10.1016/j.cma.2023.115912⟩. ⟨hal-04073075⟩