一种基于多线程加速的大规模群体仿真方法  

CROWD SIMULATION METHOD BASED ON MULTI-THREAD ACCELERATION

作　　者：傅正扬 姜忠鼎[2] Fu Zhengyang;Jiang Zhongding(Shanghai Key Laboratory of Data Science, Fudan University, Shanghai 201203, China;Software School, Fudan University, Shanghai 201203, China)

机构地区：[1]复旦大学上海市数据科学重点实验室,上海201203 [2]复旦大学软件学院,上海201203

出　　处：《计算机应用与软件》2019年第3期154-161,242,共9页Computer Applications and Software

基　　金：国家自然科学基金项目(60803064)

摘　　要：大规模群体仿真技术可以提高虚拟场景的临场感和趣味性,在游戏娱乐、演习训练、建筑设计等领域具有重要的应用价值。然而,常用于虚拟场景开发的游戏引擎难以实现大规模群体仿真。设计一种基于多线程加速的大规模群体仿真方法,提升大规模群体仿真包含的群体动画渲染和群体行为模拟性能。针对大规模群体动画渲染,在GPU蒙皮渲染算法上加入群体动画状态的多线程计算,实现群体动画状态与蒙皮动画的加速计算;针对大规模群体行为模拟,将自主性群体行为算法修改为速度模型,与多线程优化后的最优互相碰撞避让算法结合,实现群体的避让行为,提升群体行为模拟效率。实验结果表明,该方法可以很好地与游戏引擎结合,实现复杂场景的大规模群体仿真,提升大规模群体场景帧率。Crowd simulation technology can improve the vividness and interest of virtual scenes, and has important application value in the fields of entertainment, training, architectural design. However, the game engine commonly used in virtual scene development is difficult to achieve crowd simulation. We designed a crowd simulation method based on multi-thread acceleration, which improved the performance of crowd animation rendering and crowd behavior simulation included in large-scale crowd simulation. For crowd animation rendering, the multi-thread calculation of crowd animation state was added to the GPU skinning instancing to accelerate the calculation of group animation state and skinning. For crowd behavior simulation, the autonomous group behavior algorithm was modified to a speed model and was combined with the optimal collision avoidance algorithm optimized by multi-thread, so as to achieve crowd avoidance behavior and improve the efficiency of crowd behavior simulation. The experimental results show that the method can be well combined with game engine to achieve large-scale crowd simulation of complex scenes and improve the frame rate of large-scale crowd scenes.

关 键 词：多线程 群体仿真 GPU蒙皮渲染 碰撞避让 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]