01663318

Component

The investigation of crashes and other traffic incidents in terms of frequency, rate, severity and duration has received significant attention by researchers. Several advanced statistical and econometric methods are being used to model these phenomena. This paper extends previous research by proposing alternative nonlinear dynamical theories such as the Fractal theory and the Cusp Catastrophe theory for crash and incident rate modeling. Congestion patterns of a motorway in Sydney, Australia are assessed using the Hurst Exponent from Fractal theory. Further, cusp catastrophe models are applied to identify the key variables that can cause sudden changes in crash and incident rates. AADT per lane, the presence of ramps, horizontal curve and the Hurst exponent are found to have a potential effect on the system dynamics. The results are then compared with that of the widely applied Tobit models. The cusp model utilized fewer variables than the tobit model and offered a significantly better model fit for both crash and incident rates. However, comparison of cusp models with a nonlinear logistic model and other diagnostics suggested the use of cusp modeling only for incident rates. For instance, when crashes rates are examined, traditional linear models could also be used to describe the underlying phenomena.

This paper was sponsored by TRB committee ABJ80 Standing Committee on Statistical Methods.

18-05663

English

8p

2018

Transportation Research Board 97th Annual Meeting

Digital/other

Figures; References; Tables

Catastrophe theory; Crashes; Fractals; Incident management; Traffic congestion

Sydney (Australia)

Highways; Safety and Human Factors

Transportation Research Board Annual Meeting 2018 Paper #18-05663

TRIS, TRB, ATRI

Jan 8 2018 11:27AM