A multi-GPU accelerated virtual-reality interaction simulation framework.

Area of Science:

• Computer Graphics
• Virtual Reality
• Scientific Simulation

Background:

• Real-time simulation of deformable objects is crucial for immersive virtual reality experiences.
• Integrating real-world objects into virtual environments presents significant computational challenges.
• Existing methods often struggle with the balance between simulation fidelity and computational efficiency.

Purpose of the Study:

• To develop a real-time virtual reality interaction simulation framework.
• To enable seamless interaction between reconstructed real objects and virtual deformable objects.
• To enhance computational performance using a multi-GPU approach.

Main Methods:

• Image-based 3D reconstruction using an extended voxel-based visual hull (VbVH) algorithm.
• Hybrid deformation model coupling geometry constrained fast lattice shape matching (FLSM) and total Lagrangian explicit dynamics (TLED).
• Multi-GPU implementation utilizing Compute Unified Device Architecture (CUDA) for parallel processing.

Main Results:

• Successful real-time simulation of one-way virtual-reality interactions, including virtual cutting of soft tissues with bleeding effects.
• Demonstrated real-time performance under moderate calculation scales.
• Significant improvement in computational efficiency per time step.

Conclusions:

• The proposed multi-GPU accelerated framework provides an effective 3D interaction technique for virtual reality.
• The hybrid deformation model ensures efficient and stable simulation of virtual object elasticity.
• The system achieves real-time performance, advancing the capabilities of VR applications.