New stochastic highway capacity estimation method and why product limit method is unsuitable

TL;DR

This paper critiques the use of the product limit method for stochastic highway capacity estimation, demonstrating its unsuitability, and proposes a new method that better fits traffic flow data, validated through accuracy analysis and a case study.

Contribution

It introduces a novel capacity estimation method that accounts for traffic flow behavior and demonstrates its advantages over traditional methods like PLM and MLE.

Findings

PLM and MLE are unsuitable for traffic capacity estimation.

The new method provides more accurate and sensitive estimates.

Harmonisation system improves highway capacity.

Abstract

Kaplan-Meier estimate, commonly known as product limit method (PLM), and maximum likelihood estimate (MLE) methods in general are often cited as means of stochastic highway capacity estimation. This article discusses their unsuitability for such application as properties of traffic flow do not meet the assumptions for use of the methods. They assume the observed subject has a history which it went through and did not fail. However, due to its nature, each traffic flow measurement behaves as a separate subject which did not go through all the lower levels of intensity (did not "age"). An alternative method is proposed. It fits the resulting cumulative frequency of breakdowns with respect to the traffic flow intensity leading to the breakdown instead of directly estimating the underlying probability distribution of capacity. Analyses of accuracy and sensitivity to data quantity and censoring rate of the new method are provided along with comparison to the PLM. The results prove unsuitability of the PLM and MLE methods in general. The new method is then used in a case study which compares capacity of a work-zone with and without a traffic flow speed harmonisation system installed. The results confirm positive effect of harmonisation on capacity.