A torsional social force model for simulating rotation behavior of pedestrians under multiple scenarios

Abstract

Pedestrian behavior is a complex and important research subject, which has been studied extensively by simulation and experiment. As the basis or premise of other behaviors, rotation behavior is the most common pedestrian behavior in life. However, the current pedestrian simulation method is difficult to depict the rotation, which results in insufficient authenticity of the simulation effect. This paper proposes a systematic work on the modeling and simulation problem of rotation behaviors under multiple scenarios. Firstly, we improve the body model based on elliptical structure to simulate the microscopic behavior of pedestrians in the most realistic way. Then, we modeling the pedestrian rotation behavior in multiple scenarios by proposing two indexes: rotation angle and rotation timing, which include the Case I: Traversing behavior in narrow areas, Case II: Avoidance behavior in counter flow and Case III: Overtaking behavior in medium density flow. After that, we propose a torsional social force model (T-SFM), which significantly improves original model from two aspects: (a) fusing the torsional force; (b) adding the rotation willingness. More than 500 times real experiments are conducted, and according to the analysis results of real experiments, this paper verifies the authenticity of rotation model from both qualitative and quantitative perspectives. Moreover, we calibrate the general parameters to verify the generality of the rotation model. The simulation results show that the T-SFM can not only simulate the rotation behavior with high authenticity, but also has good generality. Our research significantly improves the authenticity of pedestrian simulation and provides support for heterogeneous pedestrians modeling in complex environments.