Exploratory Advanced Research Program

VASTO - Evolutionary Agent System for Transportation Outlook

Agent-Based Modeling and Simulation in The Dilemma Zone

3 Overview of Research Methodology

Our research efforts focused on two elements: (1) collecting human behavioral data via a driving simulator, and (2) modeling and analysis.

While collecting human behavioral data via a driving simulator, our first goal was to develop an experimental plan, which helps to understand the individual and interactive impacts of factors on output results with minimal experimental and systematic error. ⁽¹⁰⁾ The goal was to capture driver behavior in the DZ under a variety of conditions. An experimental plan was developed in collaboration with the University of Arizona, Leidos, and AAI, where the factors considered include the (1) facility speed limit of the roadway (FacilitySpeed), (2) degree of driving in a hurry (InAHurry), (3) presence of a red-light photo enforcement camera (RedLightCam), (4) presence of a pedestrian countdown signal (PedCountSig), and (5) behavior of an adjacent vehicle as the participant approaches the intersection (AdjVehBeh). The University of Arizona, in collaboration with researchers at AAI and Leidos, developed a simulation work plan that was used to collect human behavior data when encountering intersections via a driving simulator at Federal Highway Administration's Turner-Fairbank Highway Research Center. More details about the simulation work plan can be found in section 4, and a video of the driving simulator is available at https://www.youtube.com/watch?v=vrJiVJaZW4g. ⁽¹¹⁾

The responses of participants in the driving simulation experiment were used to develop an ABMS under the E-BDI framework (see section 5.1.2). In addition to the five factors previously mentioned, two factors provided by the ITE DZ model were considered: a vehicle's approaching speed (AppSpeed), and a vehicle's distance to the stop line (Distance). More details of the experiment design can be found in section 5.

In terms of modeling and analysis, our first goal was to identify the essential factors relevant to drivers' decisionmaking and to mimic their behavior when running into a yellow phase at intersections. To this end, an analysis of the DZ was conducted based on the responses of participants in the driving simulation experiment. 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 with respect to the five factors (i.e., FacilitySpeed, InAHurry, RedLightCam, PedCountSig, and AdjVehBeh). The second goal was to develop a DZ simulation model where drivers' decisionmaking behaviors are represented in the E-BDI framework. ^{(1) (12)} In addition to the five factors previously mentioned, two factors provided by the ITE DZ model were also considered, AppSpeed and Distance. To test the impact of these factors on drivers' proceeding behavior when encountering a DZ, several microscopic traffic simulation experiments were conducted based on the constructed and validated ABMS model. More details of this work can be found in section 5.