Deformable polygonal agents in crowd simulation

To produce impressive virtual worlds, real-time crowd simulations require large and detailed scenes populated by agents with complex shapes and geometry. For efficiency reasons, these agents are usually approximated by point-like representations to optimize the performances of collision avoidance and interactions between agents. This paper addresses the issue of handling deformable polygonal agents with arbitrary shapes in real time crowd simulations. The proposed multiresolution framework supports environments with arbitrary topologies and provides tools for efficient proximity queries.