A Flexible Framework for General, Multi-agent Based Real-time Crowd Simulation

In this paper, we present a generic and flexible framework for multi-agent based, real-time crowd simulation. Due to the modular arcitecture and the simple structure, all components can easily be modified or replaced. In contrast to many other crowd simulation systems available, our system is able to handle very different scenarios and behaviors with minimal adaptation effort. Furthermore, the presented system is able to simulate and visualize all scenarios in realtime on standard PCs without precalculation steps. Optimization techniques like multi-threading and cell space partitioning are discussed to further improve the simulation’s runtime performance. Finally, different scenarios with different behavior models are discussed to show the flexibility of our framework.