Modeling and Calibration of Contact, Friction and Deformation between a Rigid Object and another Rigid or Rheology Object

本研究的目的是建立刚体之间的接触（包括摩擦）模型，并设计推力流变（复杂粘弹性）物体的变形模型。这些模型不仅适用于现实世界中机器人对各种材料的操纵，也适用于虚拟世界中触觉对各种材料的操纵。针对前一个问题，我们提出了一种基于脉冲动量的方法来人为地产生任意脉冲（力序列）。这是米尔蒂奇著名方法的改良版。我们的新方法有足够的参数来表达几种物理现象。例如，遇到的两个物体之间存在静摩擦系数和动摩擦系数，泊松接触模型中的恢复系数等等。在大量实验数据下，采用随机化算法和遗传算法两种近最优概率搜索算法对未知参数进行了有效的标定。为此，我们需要用高速摄像机精确测量碰撞前后的速度。最后，在空中曲棍球和桌球运动中验证了该方法的可行性。针对后一个问题，我们提出了智能MSD （mass - spring -阻尼器）模型来表示变形物体的粘度、弹性和残余位移。一般来说，在我们的生活生活中有很多种类的流变物体。因此，对于每种材料，我们用遗传算法校准了MSD模型的许多不确定参数。每个MSD元件依次由Voigt和阻尼器部件组成。因此，需要标定的参数有两个阻尼器系数、一个弹簧系数、Voigh与阻尼器的长度比等。为了校准所有参数，我们尽量减少真实和虚拟流变物体之间的形状差异。通过两个或多个立体视觉相机系统Digiclops及其软件开发工具包（SDK） Triclops（由Point Grey Research Inc .， Canada提供）精确捕捉真实流变物体的形状变形。每一个都可以实时捕获大约3000个或更多的点作为真实流变物体的形状。少

The purpose of this research is to construct contact (including friction) model between rigid bodies and also design deformation model of pushed rhelogy (complex viscous - elastic) object. These models are quite useful for manipulating many kinds of materials by robotic manipulator in a real world and also for manipulating them by haptics in a virtual world.Concerning to the former problem, we propose an impulse-momentum-based approach to generate an arbitrary impulse (sequence of forces) artificially. This is a modified version of Mirtich's famous approach. Our new approach has enough parameters to express several physical phenomena. For example, there are static and kinetic friction coefficients between two encountered objects, restitution coefficient in the Poisson's contact model, and so on. The unknown parameters are efficiently calibrated by two kinds of near-optimal probabilistic search algorithms such as the randomized algorithm and genetic algorithm under many experimental dat … More a. The calibration is based on the conservative law of momentum variation. For this purpose, we should exactly measure velocities before and after collision by high-speed camera. Finally, we check the feasibility of our new approach in the air hokey and billiard games.Concerning to the latter problem, we propose smart MSD (Mass-Spring-Damper) model to represent the viscosity, elasticity, and residual displacement of deformed object. In general, there are many kinds of rheology objects in our living life. Therefore for each material, we calibrate many uncertain parameters of MSD model by the genetic algorithm. Each MSD element consists of Voigt and damper parts serially. Therefore, parameters to calibrate are two coefficients of dampers, one coefficient of spring, length ratio of Voigh and damper and so on. In order to calibrate all the parameters, we minimize the shape difference between real and virtual rheology objects. The shape deformation of real rheology object is precisely captured by two or more stereo vision camera systems Digiclops and its software development kit (SDK) Triclops (provided by Point Grey Research Inc, Canada). Each captures about three or more thousand points as shape of real rheology object in the real-time manner. Less