【作者】 刘金鹏；

【导师】 陈定方；

【作者基本信息】 武汉理工大学 ， 机械设计及理论， 2004， 博士

【摘要】 虚拟环境是一种先进的人机界面，能同时提供诸如视觉、听觉、触觉、力反馈等各种直观、自然的手段，与计算机中的模型进行实时交互，从而生成一种具有真实世界效果的计算机环境。随着计算机技术的迅猛发展，虚拟现实技术的应用领域也越来越广泛，同时对虚拟环境的真实感要求也越来越高，这里所指的真实感不仅仅是指图形，还包括模型的物理属性和行为属性。而且，在满足真实感的前提下，还应当在图形渲染输出以及对用户交互的响应方面能达到较好的实时性。其中，真实感要求是主要的，尤其是对于模型的物理属性和行为属性的研究，已成为虚拟现实领域中亟待解决的瓶颈问题。 在虚拟现实系统中，模型必须能够反映质量、重量、惯性这些基本属性，同时对某些特殊的应用而言，干涉、碰撞以及碰撞后的响应、关联物体的运动等的表现也必不可少，所有这些都会严重影响虚拟环境的真实程度，将直接影响操作者的沉浸感，最终关系到整个系统设计的效果。 本论文工作旨在解决虚拟环境中模型的物理属性和行为属性的表现问题，同时对传统的LOD算法进行改进，使其能满足模型的物理属性和行为属性的表现要求，以期能达到虚拟场景的真实感和对实时性的良好结合。 在现实世界中，任何物体都有质量，并由此而产生惯性并受到其它物体的引力作用，在受到外力作用时将满足牛顿三个定律。两个物体的运动轨迹相交时，将发生碰撞、干涉，在碰撞前后满足动量守恒和能量守恒定律。此时，其运动状态将发生改变，表现为速度、加速度、运动方向的改变。传统的计算机图形学方法对这些问题的解决方案是非常笨拙的，基本上是通过设计者手工设定运动的关键帧，在关键帧处由设计者根据现实生活经验或根据科学计算来设定模型的速度、运动方向等属性，然后由计算机来根据这些关键帧进行插值，从而完成整个动画过程。虽然目前优秀的商业应用软件已经具备了一定的智能，对这些关键帧之间的插值算法进行某些设定，如使物体运动的速度在两个关键帧之间呈抛物线分布，达到缓慢启动和停止的效果，使之具备一定的真实感，然而这些工作仍然远远不够，尤其是用户想在运动过程中想表达交互愿望时，这些软件的表现更是难以满足。 论文以虚拟现实系统中模型的物理属性和行为属性表现为主要内容，着武汉理工大学博士学位论文重对虚拟环境中模型的经典力学物理属性进行研究，并在经典力学范畴内对模型由于这些物理属性而产生的行为进行研究。由于传统计算机图形学的学科基础根本不支持对模型的物理属性建模，因此，在对模型的基本物理量进行建模之前，论文在这两个学科之间建立了一个桥梁，希望通过对这些物理量进行抽象，从而能在有效利用计算机图形学的现有成果的基础上来满足物理属性表现的要求。在此基础上，对经典力学中的基本物理量如时间、质量、受力、速度、加速度等进行计算机图形学建模，以期建立一个完整的理论体系，然后在这个体系基础之上研究模型的行为属性，如干涉、碰撞以及碰撞后的响应、.关联物体的运动等。为了使上述研究成果能更好地应用于虚拟设计、虚拟制造以及游戏、娱乐领域，论文最后提供了一整套方便、灵活的开发接口。通过使用该接口，可以在传统图形学的基础上，方便地开发出带有物理属性和行为属性的虚拟环境。最后，论文在上述理论体系的基础上，对传统LOD算法进行了改进。传统的LOD算法在增、删模型的节点和面时，并没有考虑物理表现方面的问题。论文对模型在受集中力、分布力作用下，产生的变形进行分析，对传统的LOD算法的节点删除算法进行改进，使其能更好地满足物理属性和行为属性的表现，以满足虚拟环境的要求，实现对虚拟环境的真实感生成及对其浏览操作的实时性的较好结合。 论文完成的主要工作包括:第二章介绍了相关的研究背景，然后对本论文的研究内容进行了界定，并对在传统计算机图形学成果上如何表现物理属性和行为属性进行了哲学上的讨论。第三章首先对经典力学范畴内的物理量建模进行讨论，企图建立一个完整的理论体系。第四章在上述体系基础上对行为属性的表现进行研究，分别研究了刚体干涉、碰撞以及碰撞后的响应、关联物体的运动，并根据机械设计、制造领域的特殊要求对上述模型算法进行了优化。第五章对传统的LOD算法进行改进，使其满足物理属性和行为属性表现的要求。第六章介绍PBMEVR开发包的设计思路和使用方法，使用该开发包将可以在传统计算机图形学基础上非常容易地开发出带有物理属性和行为属性的应用。最后，第七章对论文所做的工作进行了总结和展望，指出论文以后的研究思路。更多还原

【Abstract】 Virtual Environment is an advanced interface, which provides some intuitional and natural perceptive means, such as vision, hearing, feeling, and so on. Users construct a set of realistic models with computer and then create virtual environment with 3D characteristic. With the rapid development of computer technology, virtual reality technology is used in more and more areas. In the meantime, it needs to generate realistic images in real-time. The third dimension here refers to not only graphics but also models’ physical and behavior properties. So, under the precondition of satisfying realistic images, we also need a good real time property in graphics rendering and rapid response. But the realistic images generating is the primary problem. Especially for research on physical and behavior properties, has been the bottleneck and key point in virtual reality application areas.In the virtual reality system, the model must have some basic properties such as mass, weight and inertia. For some special applications, flexibility , friction and collision must also be described. All these can affect realistic of the virtual environment and the users’ immersion seriously, and affect the whole design impact consequently.This thesis solves the description of models’ physical and behavior properties in virtual environment; and improves on the traditional LOD algorithms so that it can satisfy the description of models’ physical and behavior properties, thus we can get a good combination of realistic scenes generating and real time controlling.Any object in the real world has its’ mass, it generates inertia, gets gravitation from other objects, and meets Newtonian three laws. When the motion tracks of two objects intersect, they may collide and have interference. Before and after their collision, they meet law of conservation of momentum and law of conservation of energy. Now, their motion states will change, such as their motion velocities, accelerations and motion directions. It is very clumsy to solve these problems with the traditional computer graphics method. It needs thedesigner to set some key frames manually, and at every key frame, the designer should also set the object’s motion property according to his experience or scientific compute results. Then computer gives some interpolatory values and realizes the whole movie. Though a lot of commercial soft wares have some intelligence and can set the interpolatory values to achieve users’ requirement in some degree, these soft wares can’t satisfy users’ requirement well, especially when users want to express their interactive desire.This thesis solves the description of models’ physical and behavior properties in virtual environment; the thesis also takes emphasize on objects’ physical properties from classical mechanics point of view, and researches the objects’ behavior. The base of traditional computer graphics does not support objects’ physical properties modeling, so, before modeling for the objects’ basic physical properties, we build a bridge between these two knowledge, traditional computer graphics and physical properties modeling. We hope to abstract these physical variables, and express their physical properties satisfactorily by taking good use of the existing computer graphics productions. On the bases of these, we give computer graphics models of some basic physical variables, such as time, mass, force, velocity and acceleration, etc.. Thus we can build an integrate system. Then we can do some research work about the models’ behavior properties, such as collision , collision response and invert kinematics, and so on. For taking good uses of these research productions in virtual design, virtual manufacture , games and entertainment, we give a whole set of develop tools, which are very convenient and flexible. By using the develop tools, users can build a virtual environment with physical and behavior properties on the bases of traditional computer graphics. At last, this thesis also gives out an improved LOD method. The traditional LOD algo更多还原