Real-time Simulation Of Massive Crowd Evacuation Based On Dynamic Implicit Potential Field In Venue

With the rising of crowd evacuation applied widely to the aspects such as the rationality of building’s design, the group behavior drilling and the video production,how to realize people path planning efficiently and construct complicated behavior model has attracted attention of many researchers. Although the crowd simulation of evacuation has been reviewed on several aspects such as the crowd behavior and path panning while simulation system can’t realize complicated render effect in real time when the size of crowd reach a certain high level, it still poses great challenges for obtaining the real movement effect meanwhile giving consideration to the real time rendering. The cause of above phenomenon is that the complicated simulation model over occupies the computing resource for too many factors taken into account.Fortunately, as the General Purpose GPU for concurrent computation arisen, the time consuming of traditional great scale compute-intensive tasks with GPU are much less than the calculation operated on CPU, which brings new opportunity to further promote the simulation efficiency of great scale crowd in real time.Dynamic implicit potential field( potential field) is a classical crowd model in the field of fluid simulation, which overcome the crowd simulation’s technique confines concretely represented in the aspect of synthesizing the factors which influence crowd’s movement into a continuum field. Although the above method has improved the simulation efficiency greatly, we still need to deal with a lot of problems, such as the rough calculation of crowd’s movement and method of terrain description compared with complicated scene with large scale crowd, the serial algorithm becoming the bottleneck for constructing potential field and so on. Therefore, the rendering of tens of thousands of people evacuation in real time form the major research object for satisfying realistic and high performance simultaneously. This paper has the following innovations:(1) We propose a crowd simulation method for evacuation controlled by a combination of dynamic potential field and local navigation field in parallel computation. According to the phenomenon that there exists a lot of obstacle with small volume and Multi exits in venue, we construct a global potential field vectoring crowd’s movement in means of scene information, and take use of local navigation field in order to avoiding collision and planning path in certain areas simultaneously.(2) New methods of describing discrete data in complicated scene and crowd’s motion redirection are proposed on the basis of Parallel Computing Environment. The optimization of constructing potential field and crowd’s motion model has accomplished by the establishment of a common terrain data description method.(3) According to the character of constructing potential, we map the process intomulti iterative calculation in GPU and propose a parallel method with multi wave fonts to solve path planning in scene existing multi exits.The simulation of evacuation fully utilizes the high concurrence of GPU in the field of path planning and virtual human’s rendering. The results show that our method can satisfy the simulation of evacuation with tens of thousands people in real time.