An Application of Chaotic Time-series Analysis to Car-following Data at Capacity Bottleneck

Abstract

Modelling car-following behavior is a promising approach to clarifying mechanisms of capacity bottleneck phenomena in an access-controlled section on an expressway. Since the car-following behavior has nonlinearity, it is possible that deterministic chaos is observed in the behavior. The chaos in the car-following behavior as a microscopic element in traffic flow may relate to a complex phenomenon of traffic flow, such as the stochastic nature of the breakdown. This paper applies chaotic time-series analysis to car-following behavioral data called "Zen Traffic Data (ZTD)" collected on Hanshin Expressway. The analysis estimates the Lyapunov exponent which means the ratio of extension on the attractor in reconstructed phase space derived from the original time-series of car-following data. The positive exponent indicates the chaotic nature in the data. In morning commuting hour (7-8 am), approximately 60% of the following cars have possibly the chaotic characteristics because the Lyapunov exponent of these car-following data are positive. The findings will direct how the car-following is modeled to accurate illustration of the breakdown phenomena in traffic flow on capacity bottlenecks.