Pedestrian modeling using the least action principle with sequences obtained from thermal cameras in a real life scenario

Abstract

Living labs provide the possibility of doing real-time research in an ecological context corresponding to normal daily activities. In particular, it is important to know how humans respond to environmental changes and different scenarios. The appropriate characterization of individual human displacement dynamics within a crowd remains elusive, and for this reason, there is a keen interest in exploring behaviors with general physical models. In this work, we present a theoretical and experimental study of the natural movement of pedestrians when passing through a limited and known area of a shopping center. The modeling problem for the motion of a single pedestrian is complex and extensive; therefore we focus on the need to design models taking into account mechanistic aspects of human locomotion. The theoretical study used mean values of pedestrian characteristics, e.g., density, velocity, and many obstacles. We propose a human pedestrian trajectory model by using the least-action principle, and we compared it against experimental results. The experimental study is conducted in a Living Lab inside a shopping center using infrared cameras. For this experiment, we collected highly accurate trajectories allowing us to quantify pedestrian crowd dynamics. The tests included 20 runs distributed over five days with up to 25 test persons. Additionally, to gain a better understanding of subjects’ trajectories, we simulated a background of different pathway scenarios and compared it with real trajectories. Our theoretical framework takes the minimum error between previously simulated and real point pathways to predict future points on the subject trajectory.