This project was presented at Motion Interaction and Games 2024 and received the Best Student Paper (Honorable Mention).

It enhances the realism of crowd simulation by incorporating social rules and gaze-driven attention.

Traditional crowd simulations mainly focus on goal-directed movement and collision avoidance, often resulting in robotic behavior. Our approach introduces gaze-driven steering, social interaction, and realistic attention shifts, making agents behave more like real humans.

By integrating social behaviors and dynamic gaze control at both the steering and animation levels, our method significantly improves the naturalness of crowd dynamics. A user study confirmed that our model is perceived as more realistic compared to traditional approaches.