Virtual Rheological Objects for Visual and Haptic Presentation

本研究为流变物体的视觉和触觉呈现建立了基础技术。生物组织、食品等流变性物的变形既具有粘弹性，又具有塑性。在这项研究中，我们首先建立了一个基于颗粒的流变物体模型。引入拓扑保持效应和体积效应，成功实现了基于颗粒模型的流变变形的稳定计算。其次，利用FPGA实现了计算流变变形的逻辑电路。在FPGA（现场可编程门阵列）上实现流变变形的计算，实现了比PC计算快近100倍的计算速度。第三，建立了一种基于粒子的模型参数标定方法。构建了一个实时测量流变性物变形和压力分布的系统，根据测量结果识别模型参数。接下来，我们研究了另一种基于连续力学的基于粒子的建模。我们使用约束稳定方法稳定地计算了泊松比等于0.5的非压缩物体的变形。主要研究内容包括：1)基于粒子建模的大流变变形计算；2)基于现场可编程门阵列（FPGA）的实时变形计算；3)基于激光测距仪和触觉传感器的实时变形测量；4)基于随机算法的模型参数识别。在此基础上，对压缩比为10%时的变形进行了仿真，对超过1000个粒子的基于粒子的模型进行了视频帧率计算，并对模型进行了误差在10%以内的识别。

This research establishes fundamental technologies for visual and haptic presentation of rheological objects. Rheological objects such as biological tissues and foods show both viscoelastic and plastic nature in their deformation. In this research, we first established a particle-based modeling of rheological objects. Introducing topology maintaining and volumatic effects, we have succeeded a stable computation of rheological deformation in the particle-based modeling. Second, we realized a logical circuit to compute the rheological deformation using an FPGA. Implementing the computation of rheological deformation on an FPGA(Field Programmable Gate Array), we realized almost 100 times faster computation than PC computation. Thirdly, we established a calibration method to identify model parameters in a particle-based model. We constructed a system to measure realtime deformation of a rheological object and pressure distribution on it to identify the model parameters based on the measurements. Next, we investigated another particle-based modeling based on continua mechanics. We stably computed the deformation of uncompressive objects where Poisson's ratio is equal to 0.5 using the CSM (Constraint Stabilization Method).This research is summarized as 1) computation of large rheological deformation based on particle-based modeling, 2) realtime deformation computation using FPGA (Field Programmable Gate Array), 3) realtime deformation measurement using a laser range finder and a tactile sensor, and 4) identification of model parameters by randomized algorithm. Based on these results, simulation of deformation where compression ratio is 10%, video-frame rate computation of a particle-based model with over 1000 particles, and model identification with error within 10%.