Estimation of Route Choice Model with Euler - Lagrange Observation based on Information Theory

Abstract

The estimation of behavior model based on non-aggregate analysis of pedestrian behavior is required in order to analyze pedestrian behavior in space of about 1km square with high accuracy. In this study, we propose the estimation of behavior model from fixed observation acquired by video footage and movement observation for each person acquired by GPS and Wi-Fi. Previous methods based on movement observation require route data as input. In contrary, we define the estimation of behavior model from movement observation based on entropy and propose the estimation based on cross-entropy as the estimation of behavior model from fixed observation. This method is applied for simulation data by twin experiment. As a result, we show that this method has validity against previous estimation methods and the estimation of entropy has robustness against observation error.