STATISTICAL ANALYSIS OF GEOMETRIC ROADWAY FEATURES WITH SAFETY SURROGATES: A MICRO SIMULATION CASE STUDY

Open Access
Author:
Kwon, Daniel Wontak
Graduate Program:
Civil Engineering
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
June 26, 2009
Committee Members:
  • Venkataraman Shankar, Thesis Advisor
Keywords:
  • safety surrogates
  • geometric roadway features
  • micro simulation
Abstract:
Previous research conducted on the reliability of microscopic simulation programs has validated its usefulness as a tool for analyzing network efficiency. The characteristic that makes micro simulation so effective is the synthetic data that it produces. Information that is not easily attainable from the field can be replicated through an artificial network. The accuracy of the simulation output is dependent on how closely the artificial network can mimic conditions from the field. Upon completing calibration, the resulting output from the simulation can account for an entire population of vehicles within a given network and can expose the factors linking the fundamental components of roadway design with mobility and safety. This research aims to utilize the micro simulation program VISSIM as a tool for providing data for intensive modeling efforts that will reveal the relationship between geometrics and performance measures such as mobility and safety. The resulting output from the simulated network discloses the variations in speed and density for specific roadway segments over time. Manual manipulations of the simulated results aggregate the speed and density data into a time series that presents the output in a dynamic context. A random coefficients model structure is employed to establish the relationship between SSAM simulated conflicts, based on various safety surrogates, and roadway geometrics. In addition, using actual accident data, a crash model is developed that relates recorded crashes with simulated conflicts. The connection that exists between simulated conflicts and actual crashes will be modeled and expressed through a range of values for forecasting accident probability. Integrating the two models will be restricted to the estimation of the PM Peak accident probability.