In this thesis, the idea of applying VR technology to the design of product line with robotic workcells is proposed. Two problems related with virtual environment computing theory, level of detail representation of virtual environment and collision detection, are systematically studied. The first part of this thesis describes the implementation of ROBVR, a software system for design and simulation of product line with robotic workcells. The main goal of this project is to accelerate design process of product line. Based on WTK, the system provides full support fox" VR technology and implements many function, including generation of virtual environment, interactive layout of product line and robotics simulation. In the study of robotics kinematics simulation, this thesis provides a method to detect and avoid kinematic singularities in trajectory planing. This thesis also presents an algorithm for level of detail representation of virtual environment. This algorithm is built on mesh simplification techniques and then a new mesh simplification algorithm is presented in the thesis. The main contributions of the mesh simplification algorithm lie in several aspects: 1. Based on analysis of local theory of surface, geometry error and topology error between meshes are carefully defined, then an algorithm is presented to calculate these errors. 2. Using vertices offsetting method, we can produce offset mesh of original model. Simplified mesh is confined within the space between two offset meshes to avoid large geometry deviation. 3. In a local area, we use quadric surface to approximate original surface. By calculation the curvature of quadric surface, we can calculate topology error between two meshes. 4. In the process of mesh simplification, feature points are retained to keep the topology of original mesh. It is demonstrated by experiments that our algorithm is effective and has high triangle simplification rate. Fast and accurate collision detection between geometry models is a fundamental problem in computer graphics, robotics and virtual environments. In this thesis, we propose a multi-level collision detection algorithm based on hierarchy architecture. This algorithm uses a four-level collision detection strategy: bounding box, sphere octree, simplified meshes and original model. It can provide user-specified error collision detection. In practice, it can also accurately detect all the contacts between objects. Based on the hierarchy architecture above, the distance between two objects can be easily calculated. Experiments demonstrate that our multi-level collision detection algorithm is much faster than other collision detection algorithms.
修改评论