A calibration method structured on Bayesian Inference of the HCM speed-flow relationship for freeways and multilane highways and a temporal analysis of traffic behavior

TL;DR

This paper introduces a Bayesian inference-based calibration method for the HCM 2016 speed-flow model, analyzing traffic behavior over time using extensive sensor data and revealing significant temporal variations.

Contribution

The paper develops a Bayesian calibration approach for the HCM 2016 model, providing probabilistic parameter estimates and temporal traffic behavior analysis.

Findings

Free-flow speed and capacity follow a Normal distribution.

Segment where speed remains constant is often near zero.

Traffic behavior varies significantly across different time periods.

Abstract

This paper presents a calibration method for the speed-flow model of the HCM 2016 for freeways and multilane highways allied to temporal analysis of traffic stream. The proposed method was developed using a sample of more than one million observations collected by 23 traffic sensors on four highways in the state of S{\~a}o Paulo. The method is structured on Bayesian inference and provided for each model parameters a probability distribution function. The free-flow speed and capacity presented a probability density function that approximates a Normal distribution. The segment in which the speed of traffic stream remain constant with the increase of the traffic flow is lower than described in HCM 2016, being in some cases close to zero. Along with the proposed calibration method an analysis of temporal variation is performed which shows a significant variation in traffic behavior for different periods. The free-flow speed, capacity and breakpoint distributions obtained through monthly and annual calibration were considered equal by means of Kolmogorov-Smirnov test, different for the model calibration coefficient.