A Data-Driven Model of Pedestrian Movement

We present a method for simulating individual pedestrian motion based on empirical data. Our model keeps track of the pedestrian’s position and body confi guration (pose) and uses motion capture data to produce plausible motion. While our ultimate goal is creating 3D animations of crowds, our initial efforts focus on 2D simulations. In this paper, we present a 2D model for an able-bodied male. Using our approach, we could also capture data and build models for a heterogeneous population, including children, the elderly, pedestrians in wheelchairs, and people on crutches. We demonstrate the realism of our model with a small-scale test case and a larger crowd evacuation simulation.