Dynamic group behavior for real-time multi-agent crowd simulation

We present a real-time crowd simulation model with dynamic group behavior in virtual environment. In our model, we adopt a multi hierarchy structure composed of virtual dynamic groups and agents, and then we continue to divide the agents into leader-agent and follower-agents. At the same time, we adopt a situation-based control structure for the groups not for every agent. When groups enter a new situation, dynamic, situation-based group behaviors are composed on the fly. Through this method, we can strike a balance between the individuals´ variability and the time complexity of controlling for the crowd behavior simulation. We demonstrate our model through manifesting the crowd simulation in a virtual street.