Virtual object deformations using attributed boundary model

Abstract

Virtual reality may provide reliable simulation for object interactions, when embedding virtual objects with artificial attributes to emulated physical and mechanical properties. This paper presents an Attributed Boundary Model (ABM) that emulates realtime and realistic deformations for VR applications. In the ABM, the physical, mechanical, topological and geometrical properties of the solid are extracted and packed into a set of attribute matrices that generate equivalent displacements for its boundary model upon contact with other virtual objects. The attribute matrices are created by training the back propagation artificial neural network with displacement data provided by finite element method (FEM). These matrices are then encoded into the deformation behavioral module for the VR engine to perform realtime interactions in the virtual space. Implementations are built on Virtools using XEON PC with nVIDIA Quadro4 display device. The result shows that the speed of the animation is better than 30 FPS and the deviation from calculation is less than 1%.

Development and validation methodology of the Nuss procedure surgical planner

“…Despite the increasing computational power of modern computers, solving FEM for realistic animation purposes still poses as a difficult task. Recently, an artificial neural network (ANN) was employed to approximate the force–displacement model 18,19 using a commercially available simulation engine. The authors achieved real-time performance for several thousands of input nodes in the ANN.…”