Chaotic Analysis of Pedestrian Flow Series for Vehicular Networking in Complex Interaction Environment

Interaction between self-driving cars and pedestrians is a key issue for the design of vehicular networking and auto-drive algorithms. Pedestrian movement usually exhibits irregular and complex behaviors, dramatically increasing the difficulty of such design. By choosing a suitable method to analyze the pedestrian flow and obtain its major features, self-driving cars and service robots could intelligently avoid collision with pedestrians. Therefore, this paper analyzes the chaotic characteristics of the pedestrian flow time series of different locations in real campus scenes qualitatively and quantitatively. The optimal delay and optimal embedding dimension of pedestrian flow time series are obtained by the mutual information method and Cao’s method, respectively. The largest Lyapunov exponent, correlation dimension, and Kolmogorov entropy are used to quantitatively analyze its chaos, and recurrence plot is used for qualitative demonstrations. According to this study, pedestrian flow time series are definitely chaotic, which provides essential theoretical support for pedestrian flow prediction and auto-drive algorithm design in future vehicular networking.