Determining Vehicle Operating Speed and Lateral Position Along Horizontal Curves Using Linear Mixed-Effects Models

Abstract

Objective: Despite recent improvements in highway safety in the United States, serious crashes on curves remain a significant problem. To assist in better understanding causal factors leading to this problem, this article presents and demonstrates a methodology for collection and analysis of vehicle trajectory and speed data for rural and urban curves using Z-configured road tubes.

Methods: For a large number of vehicle observations at 2 horizontal curves located in Dexter and Ames, Iowa, the article develops vehicle speed and lateral position prediction models for multiple points along these curves. Linear mixed-effects models were used to predict vehicle lateral position and speed along the curves as explained by operational, vehicle, and environmental variables. Behavior was visually represented for an identified subset of “risky” drivers.

Results: Linear mixed-effect regression models provided the means to predict vehicle speed and lateral position while taking into account repeated observations of the same vehicle along horizontal curves.

Conclusions: Speed and lateral position at point of entry were observed to influence trajectory and speed profiles. Rural horizontal curve site models are presented that indicate that the following variables were significant and influenced both vehicle speed and lateral position: time of day, direction of travel (inside or outside lane), and type of vehicle.

Keywords:

• traffic safety
• trajectory
• speed profile
• horizontal curve
• linear mixed-effects model
• pneumatic road tubes

Acknowledgments

The authors thank the Institute for Transportation (InTrans) and Survey & Behavioral Research Services (SBRS) at Iowa State University for funding this research and providing statistical assistance, respectively. The authors also thank Damion Pregitzer of the City of Ames, Iowa, and Josh Sebern of Guthrie County, Iowa, for providing technical assistance during the data collection effort.