High-Fidelity Simulation and Highway Safety

Traffic simulations have long been major tools to be used in obtaining operations measures. For many years now, researchers in highway safety have been looking at the possibility of using simulation in evaluating highway safety including estimating/predicting crashes.

Background

This effort to evaluate highway safety with simulation began more than two decades ago with the development of the Driver/Vehicle Module (D/VM) of the FHWA Interactive Highway Safety Design Model (IHSDM) software (1). The D/VM is a single-vehicle simulation program that simulates the interactions of the driver, the vehicle, and the road at a high-fidelity microscopic level.

Research conducted by M. Banihashemi and M. Dimaiuta showed that there exists a real potential use of this type of high-fidelity simulation in evaluating highway safety (2). This research related the D/VM simulation output results to single-vehicle departure crashes on two-lane highways.

The D/VM simulates driving behavior and vehicle dynamics on a two-lane highway. It mimics the interactions of the driver and the vehicle dynamic on a two-lane highway in fractions of second intervals. It provides profiles of predicted output measures (response variables), via the simulation of a single driver/vehicle combination. The main outputs of this simulation are:

• Lateral Offset (Lane Position),
• Friction Ratios (X, along the highway length and Y, laterally),
• Lateral Acceleration (g’s), and
• Roll Over Index that shows the lateral load transfer indicating the fraction of vehicle weight borne by the right or left tires.

Existing Research as Proof of Concept

M. Banihashemi and M. Dimaiuta showed that the D/VM simulation output measures are normally distributed. They define encroachment as vehicle encroaching out of the lane boundaries from either side and investigated the relation between Lateral Offset and encroachment. They assumed the simulation outputs were normally distributed and predicted encroachments by estimating the probability that Lateral Offset exceeds some threshold values. Finally, their research identified the number of encroachments as an effective surrogate measure for departure crashes for two-lane highways with narrow shoulders.

Data used in this research were from rural two-lane highways in Washington State, obtained from the FHWA Highway Safety Information System (HSIS) (3). For normality tests 44 stretches of highway - each 1.0 mile or longer with at least 2 horizontal curves were used. To calculate probabilities the above 44 stretches, plus 53 other stretches of highways with 5.0 mi minimum length were used. As a result, for highways with shoulder width less than or equal to 3-ft the following linear model was found to relate the number of departure crashes to the number of encroachments calculated from the Driver/Vehicle simulation:

                                                                                          NDC = 0.248 + 0.00008 NE

Where:

           NDC = Predicted number of departure crashes per year per mile

           NE   = Number of encroachments calculated from the Driver/Vehicle simulation outputs per year per mile for the highway’s AADT

Research Opportunity: Expansion of the D/VM Simulation to Multi-Vehicle Simulation

The focus of this research opportunity is the expansion of the Driver/Vehicle simulation program to include multi-vehicle simulation along different types of divided and undivided facilities. Identifying the potential crash types associated with different simulation outputs and estimating the probabilities of different outputs going beyond certain thresholds may be used to estimate the number of different crash types. The following subsections show preliminary lists of crash types and associated simulation results of different driving outcomes exceeding certain thresholds. 

Single-vehicle Crashes

• Collision with animal -----> Probability (area and highway type) *
• Collision with bicycle -----> Probability (area and highway type) *
• Collision with pedestrian -----> Probability (area and highway type) *
• Overturned -----> Roll-over Index
• Run-off the road -----> Lateral Position
• Collision with fixed object -----> Lateral Position

Multiple-vehicle Crashes

• Rear-end -----> Longitudinal Position
• Head-on -----> Lateral Position, Passing SD Estimation
• Angle -----> Longitudinal Position
• Same direction sideswipe -----> Lateral Position, Lane Change
• Opposite direction sideswipe -----> Longitudinal Position, Passing

* Crashes not being associated with any simulation outcome would be estimated differently and not by using this simulation outputs.

Data Needed to Conduct This Research

Field data collection is not part of this research opportunity.  The expectation is that the researcher uses already available sources of data in conducting this research.  Examples of such data include but are not limited to:

1. Naturalistic Driving Study (NDS) Data such as SHRP 2 NDS Safety Data.
2. Roadway Information Database (RID) from SHRP 2 Safety Data that is linkable to the SHRP 2 NDS data.
3. Roadway and crash data from Highway Safety Information System (HSIS) (3).

References

1. Federal Highway Administration (FHWA), Interactive Highway Safety Design Model (IHSDM) Public Website:  https://highways.dot.gov/research/safety/interactive-highway-safety-design-model/interactive-highway-safety-design-model-ihsdm-overview , Last accessed September 2022.
2. Use of The Output of the Driver Vehicle Module of the IHSDM as Surrogate Measures for Departure Crashes” presented in the TRB 3rd International Conference on Road Safety and Simulation, 2011, https://onlinepubs.trb.org/onlinepubs/conferences/2011/RSS/1/Banihashemi,M.pdf
3. Federal Highway Administration (FHWA), Highway Safety Information System (HSIS): https://www.hsisinfo.org/ , Last accessed April 2023.