Research On Parametric Space Techniques In Physically-based Modeling

Recent years, with the development of movie industry, architectural expression and game production, 3D animation is more and more widely used, but at the same time, the requirement is higher and higher.The traditional keyframe animation technique is widely used due to its simplicity, low computation cost and editing convenience. However, to produce realistic results, an animator must have enough experience. On the other side, the physics-based animation can produce realistic results, but at the cost of high computation and editing complexity.According to this situation, improvements on the FEM based animation technique mainly focus on enhancing efficiency and editing convenience. Typical One is a parametric space technique. This technique maps physical quantities to a low-dimensional parametric space and performs time integration in it. As a result, after simulation, an animator can bake results into animation curves and modify them by adjusting each parameter trajectory using traditional keyframe animation tools. The latter implies that this technique can be integrated seamlessly into the traditional animation pipeline. Also, with such low-dimension results, the storage space is significantly reduced.The paper improves the parametric space simulation technique by introducing proportional-derivative control mechanism, which allows users to specify target parameter trajectories and strength of control forces. During simulation, a parameter may be affected by physical laws or external forces, and deviate target value temporarily, but with our control mechanism, it can keep tracing the target trajectory as a whole. Compared to current parametric space simulation technique, an animator can offer desired trajectories based on his or her requirement, and finally get realistic and desired result.On the other hand, for non-uniform materials, we improve the algorithm of fitting stiffness of its each part. The algorithm requires a short simulation provided by user, and then computes the relative stiffness of each finite element, which makes internal forces’ projection in parametric space close to a user-specified value. Compared to current algorithm, our algorithm can learn more data and get more accurate results.Finally, we implement a Maya-based FEM simulation module using C++. With this module, one can perform parametric space FEM simulation. Our module can interact with Maya’s animation and dynamic modules during simulation and save results as Maya’s animation curve for further editing, so it can be integrated into popular animation pipeline seamlessly.