FbCrowd: Interactive Multi-agent Simulation with Coupled Collision Avoidance and Human Motion Synthesis

We present an interactive algorithm to generate plausible trajectories and full-body crowd simulations. Our formulation is based on a novel two-way coupling between 2D multi-agent collision avoidance and high-DOF human motion synthesis. We present a collision-free navigation algorithm that takes into account human motion and biomechanics constraints to compute smooth trajectories. Furthermore, we present a hybrid motion synthesis algorithm that seamlessly transitions between motion blending and semi-procedural locomotion, thereby balancing control and naturalness of the synthesized motion. The overall full-body crowd simulation algorithm can generate plausible motions with lower and upper body movements for multiple agents in dynamic virtual environments at interactive rates. We demonstrate its benefits over prior interactive crowd simulation algorithms.