Investigation of pedestrian dynamics in circle antipode experiments: Analysis and model evaluation with macroscopic indexes.

• Present a series of circle antipode experiments and recognize the trajectories. • Discover the motion navigation strategies and conflict avoidance behaviors. • Obtain the distribution law of route length, route potential, travel time and speed. • Build a macroscopic index based evaluation framework for pedestrian models. We organized a series of novel circle antipode experiments in which pedestrians are uniformly initialized on the circle and they are required to leave for their antipodal positions simultaneously. In the experiments, a conflicting area is naturally formulated in the center region due to the converged shortest routes, so the practical motion navigation strategies, as well as the conflict avoidance behaviors, can be explored and summarized. Besides, the symmetric experimental conditions of pedestrians, e.g., symmetric starting points, symmetric destination points, and symmetric surroundings, lay the foundation for further quantitative comparisons among participants. In the paper, the pedestrian trajectories are recognized, and the pedestrian dynamics are accordingly investigated from the perspective of macroscopic indexes, i.e., route length, route potential, travel time, velocity and time-series. It is found that the route length has a log-normal distribution, the route potential obeys the exponential distribution, and the travel time is normally distributed as well as the speed. Furthermore, a model evaluation based on the experimental results is performed, using four distribution indexes and two time series indexes in both space and time dimensions. We graded the simulated distribution indexes by applying a Kolmogorov-Smirnov (K-S) test based method and the simulated time series indexes by a Dynamic Time Warping (DTW) based method. We evaluated two models: the traditional social force model, the modified model with default parameters and with calibrated parameters. The evaluation results showed that the two models cannot well reproduce the six macroscopic indexes, which indicates that the models do not capture the route choice and conflict avoidance behaviors of pedestrians. Our study is expected to shed light on the pedestrian modeling. Specifically, it indicates that apart from classical indexes such as a fundamental diagram, more comprehensive indexes are needed to evaluate the pedestrian flow models. [ABSTRACT FROM AUTHOR]