Abstract
Modern crowd simulation needs two important characteristics: i) capability to simulate how human beings plan their paths, and ii) capability to cope with a large number of agents, particularly, in congested areas. Most research in this area does not focus on both issues at the same time because it can be computationally expensive. This research proposes a framework for crowd simulation that embraces these two capabilities and offers low computational cost. We combine BDI (Belief-Desire-Intention) model for planning and making decision, and RVO (Reciprocal Velocity Obstacle) for coping with congested areas. We achieve satisfactory results as a large number of agents can be sim- ulated smoothly. We found that by combining BDI with RVO, the simulation time, i.e., how fast agents can move, is at least 20% faster in all scenarios. The execution time is also reduced dramatically, 35% on average. We can simulate up to 40,000 agents in a large scale area in a typical computer. The framework allows for visualized simulation of 600 agents at 24 frames per second and can go up to 6,400 agents at 1.5 frames per second.