大规模汽车群组动画的关键技术研究

城市车水马龙场景是动画、虚拟现实、游戏中不可缺少的，其必然涉及大量的汽车。汽车可视作由人驾驶的智能体，它们的运动受道路、交通规则等约束。因此，如何方便、有效地控制模拟大规模汽车流的真实运动具有重要的理论意义和应用价值。本项目旨在研究虚拟动画场景中大规模汽车运动的关键技术，具体包括基于宏观模型的全局路网中交通流的控制，基于微观模型的局部关注区域内交通场景的逼真模拟，基于真实视频的驾驶行为参数学习，基于驾驶意图的灵活变道模拟，交通灯控制路口及各种汇流分流等连接结构附近汽车的智能反应行为,车辆驾驶路径搜索，非理想交通秩序场景模拟等动画技术。同时，还将为用户提供三维路网创建编辑工具，用户可以导入GIS道路相关数据，处理得到真实路网拓扑结构和几何结构，也可以直接创建所需路网结构。项目将最终研发出一个计算高效、控制灵活、交通细节丰富的汽车群组动画模拟原型系统，以广泛应用于动画制作实践中。

Virtual urban environments with heavy traffic are indispensible parts of 3D animations，virtual reality, and computer games, in which, there are always lots of automobiles driving on the city road. Since automobiles are driven by drivers, each automobile can be viewed as an agent and its motion is constrained by roads and traffic rules. Therefore, it is important to investigate how to effectively control and simulate massive automobiles in a convenient and realistic way because of its theoretical value and potential applications. This project aims at investigating the key techniques in simulating large-scale automobiles' motion in virtual urban environments. The research contents include the traffic flow control in a global road network based on the macroscopic model, detailed and realistic simulation of traffic scenes in local focused regions based on the microscopic model, learning realistic driving parameters from captured traffic videos, flexible lane changes based on the intention and various intelligent behaviors of drivers when approaching diverse junction structures like signalized crossing, merging and weaving parts, driving route searching, non-ideal traffic scenario simulation with violation of traffic regulations. In addition, we will provide a three dimensional road network editing tool, in which, a user can either import and process the GIS road data to get real road topological structures and geometry representations, or manually create the needed road system. The project will finally develop a prototype system for detailed automobile animations in large scale with efficient computation and flexible control for animation practice.