U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590

Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

This report is an archived publication and may contain dated technical, contact, and link information
Back to Publication List        
Publication Number:  FHWA-HRT-15-082    Date:  December 2015
Publication Number: FHWA-HRT-15-082
Date: December 2015


Exploratory Advanced Research Program

VASTO - Evolutionary Agent System for Transportation Outlook

Agent-Based Modeling and Simulation in The Dilemma Zone


1 Executive Summary

The dilemma zone (DZ) is an area in which a driver cannot stop gently before the stop line or clear the intersection safely before the red phase.(1) When decisions are made in the DZ, there could be a risk of a car crash, such as a rear end or right-angle crash. To mitigate the risk inherent in the DZ, many researchers have attempted to identify critical factors and develop a driver decisionmaking model. (1) The purpose of this study is to develop a realistic DZ model that considers the effects of surrounding factors at an intersection. The factors examined in the assessment of the influence of the road environment on driver decisionmaking behavior within the DZ simulation were:

Participants were separated into four groups based on the facility speed limit and the degree of driving in a hurry. The four groups were: (1) group A had a 40 mi/h speed limit and did not drive in a hurry; (2) group B had a 40 mi/h speed limit and drove in a hurry; (3) group C had a 55 mi/h speed limit and did not drive in a hurry; and (4) group D had a 55 mi/h speed limit and drove in a hurry.

The Federal Highway Administration's Highway Driving Simulator (HDS) was used to conduct experiments with licensed drivers (i.e., residents of the Washington, DC, metropolitan area, which includes the District of Columbia, southern Maryland, and Northern Virginia). The HDS used a late model compact car chassis providing acceleration, deceleration, handling, and braking functions, along with all audio (e.g., driving noise) and a 210-degree horizontal high-resolution visual scene. The experimental scenarios had 24 signalized intersections, 8 of which had signal changes intended to produce a DZ.

Ninety-nine licensed drivers provided useable data, 48 females (mean age 46 years, range 18-80) and 51 males (mean age 47 years, range 19-82).

Generalized estimating equations (GEE) with a binomial response distribution and logit link function were used to assess whether the probability of proceeding through a DZ intersection varied regarding the five factors (FacilitySpeed, InAHurry, RedLightCam, PedCountSig, and AdjVehBeh). The results showed that all factors influenced the drivers' decisions. Participants in the 40 mi/h (64.37 km/h) speed condition had a higher probability of continuing through a DZ intersection than those in the 55 mi/h (88.51 km/h) speed condition. The probability of proceeding through a DZ intersection was greater for participants who were not tasked to drive in a hurry than for participants who were driving in a hurry. In addition, the probability of proceeding through a DZ intersection without a red-light photo enforcement camera increased across all drivers. The absence of a pedestrian countdown signal also increased the likelihood that participants continued through a DZ intersection. When an adjacent vehicle traveled through the intersection, participants were also more likely to proceed through a DZ intersection.

From these driver response findings in the HDS, a DZ model was developed via ABMS under the E-BDI framework, which represents the perception and decisionmaking behaviors of humans regarding an uncertain property by using probabilistic models. (2) Four types of E-BDI-based DZ models were classified by the facility speed limit and the degree of being in a hurry. E-BDI-based DZ models were calibrated with two hypothetical constructs: internal information and external information. Internal information includes variables that depend on the vehicle's approach speed (AppSpeed) and distance to the stop line (Distance), as set forth in the Institute of Traffic Engineers' (ITE) DZ model. External information includes three environmental factors, i.e., RedLightCam, PedCountSig, and AdjVehBeh. Because both hypothetical attributes were unobserved during the experiments, i.e., they were latent variables, a structural equation modeling (SEM) approach was used to analyze the relative contributions of internal and external attributes on drivers' decisions in a DZ. The results of the SEM were used to calibrate the E-BDI framework (i.e., E-BDI-based DZ model). Thus, the E-BDI-based DZ model was able to represent stopping and proceeding decisions of drivers from external and internal information.

The ABMS with E-BDI-based DZ model depicts the impact of external and internal information via a speed-distance (SD) diagram. In ABMS, driver agents' stopping and proceeding decisions at the onset of the yellow phase were counted to create SD diagrams under different configurations of intersections. The results showed that the presence of a pedestrian countdown signal or a red-light photo enforcement camera mitigates the number of red-light violations. According to the post-participation survey of the drivers, drivers were able to predict the time of the traffic signal change via the remaining time of the pedestrian countdown signal so that they could make a clear decision at the intersection. In addition, when drivers were uncertain of their decisions, they followed the decisions of adjacent vehicles or tried to stop if a red-light photo enforcement camera existed on a roadway.



Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101