Sensitivity analysis and validation of a multi-agents pedestrian model

We present a pedestrian movement model, which use a multi agents system for pedestrian traffic analysis. The model captures the dynamic microscopic interaction between pedestrians, which cannot be addressed using traditional macroscopic approach. The pedestrians are modeled as autonomous agents with non-linear system differential equations. The pedestrian agents may avoid other pedestrians, passing and overtaking slower pedestrians and to form a self-organization behavior of lane formations as in real pedestrian studies. A critical issue for such multi-agent pedestrian models, however, is the validation of the model against real world data. We show that the sensitivity analysis of control variables and parameters of the multi-agents model form the basis in ensuring the validation step. The model was automatically validated using real world data by minimizing the difference between the speed distributions. With the validated model, we can utilize model applications to evaluate pedestrian facilities.