Physically based Simulation for Soft Body

Yaqi Zhao, Purdue University

Abstract

This thesis introduces a general framework for soft body dynamics, focusing on real-time deformation for complex objects with physically-based methods. Physically-based simulation has a wide range of applications such as virtual reality, visualization, training, education, and the game industry. The most challenging task for complex soft body simulation is to choose an appropriate modeling method that can achieve accuracy and computational efficiency. Unfortunately, there is no one method that meets all the needs for soft body simulation. For instance, one modeling method is the mass spring system which is computationally efficient, and visually plausible. However, it is unrealistic for the simulation of complex objects. The finite element method, on the other hand, is physically accurate, but it significantly increases the computational complexity making it unsuitable for real-time applications. In order to increase the accuracy within the simulation while maintaining a fast rate of calculation during the simulation process, we developed a novel method that enhances the performance of the mass spring system. With our method, complex objects are broken into primitives, which enables deformation to be processed at the primitive level. By using dynamic constraints, these primitives could be reconstructed into the original complex object. To simulate the soft body deformation process, collisions need to be detected and processed in real time. Since inter penetration is not allowed in soft body deformation, none of the current collision detection methods can be used in the system. To overcome this problem, we developed a new collision detection algorithm. In this algorithm, a bounding box, which is defined by the largest extension of the xyz coordinate, will be updated in the preprocessing stage. In the simulation step, the bounding box will detect collisions first, followed by soft body / soft body collision detection.

Degree

M.S.E.

Advisors

Yang, Purdue University.

Subject Area

Computer Engineering|Computer science

COinS