Classification of pedestrian behavior using real trajectory data

Pedestrian behavior is influenced by various factors and is thus characterized by heterogeneity. The aim of this research is to explore the impact of various parameters, such as pedestrians’ characteristics, mobile phone use and walking pace on pedestrians’ behavior. Classification of pedestrian behavior contributes into understanding how different factors affect pedestrian behavior and allows a finer perception of pedestrian movement, as it helps us distinguish and interpret the way that pedestrians react to different situations. In this research a simple methodology for pedestrian behavior classification is proposed taking into account pedestrians’ heterogeneity. The methodology employs random forest algorithms in order to analyze and classify trajectory data, as well as to estimate the pedestrian behavior state in real time. The analysis utilizes real pedestrian trajectory data collected via an experiment conducted in the Campus of National Technical University of Athens under various conditions (normal and fast pace, distraction from mobile phone). Distributions of these data are explored and a clustering analysis follows, yielding satisfactory results.