Studying the technical challenges of a customizable multi-user multi-modal simulation platform for traffic safety research

Understanding traffic behavior is critical to ensuring the safety of our roads and all those who use them. Traffic psychologists and engineers utilize a variety of techniques such as large-scale field observations, computational modeling, and simulation in attempts to recognize dangers and give guidance to improve road safety standards. However, large-scale studies are often time-consuming, expensive, and may not capture nuances of individual drivers. Further, while computational modeling allows for the simulation of various traffic scenarios, it is often deterministic and better suited for macro-level analysis such as studying traffic flow. Lastly, in recent years, simulation within virtual environments has become more feasible and accessible. These advances allow researchers to put a user within a virtual traffic environment and perform controlled studies in a variety of traffic scenarios. However, the main drawbacks to this approach are two-fold. First, prior research has focused on individual simulators consisting of a single vehicle, bicycle, pedestrian, etc. While studying an individual along with non-human agents in a traffic scenario may provide significant insight, it does not allow for human-to-human interactions that may be far more nuanced and of interest to a traffic researcher. Furthermore, these systems are often expensive and heavily specialized for a given research center or task. The second main drawback is that there are limitless traffic scenarios that a traffic researcher may want to study but they may not have the necessary skillset to develop those environments. Transitioning from a two-dimensional (2D) image, blueprint, or sketch of a roadway to an immersive 3D model requires a large amount of expertise in three-dimensional (3D) modeling and computer graphics. With the traditional training and education of a traffic researcher, it would be extremely unlikely that they would possess the required expertise. Further, even if they did, it would still take an immense amount of time and effort to complete the transition from 2D to 3D. The research in this dissertation addresses these two main drawbacks of virtual simulation for studying traffic behavior. A multi-user multi-modal traffic simulation platform, named InterchangeSE, was developed to study an infinite range of traffic scenarios. This includes modes such as a vehicle simulator, a bicycle simulator, and a pedestrian simulator all interacting with industry standard traffic computational modeling software. The architecture of this platform not only allowed for a variety of simulator modes, but also networking and interaction between all of them. A scenario may be configured for a variety of individual simulator modes allowing a researcher to explore complex behaviors between any combination of traffic entities. To ensure this platform would be accessible to a wide range of traffic researchers, it was important to target low-cost commodity hardware while still providing a high level of performance to prevent any negative outcomes to a user’s experience. A technical evaluation of the platform was performed to evaluate a range of factors for computer graphics as well as networking. In addition to InterchangeSE providing a means to study a wide range of traffic scenarios, an equally important and challenging problem was the creation of the scenarios themselves. This involved modeling large 3D environments to ensure an immersive experience within the simulation platform. A single virtual environment may encompass multiple square miles and include numerous entities such as buildings, roads, trees, road signs, and various forms of scenery. The next step of the research involved optimizing the process from a 2D image and generating a corresponding 3D model. An individual with 3D modeling experience may take hours to identify and model key pieces of the traffic environment using software such as Blender or Autodesk Maya. This research focused on using computer vision and image processing techniques to automatically identify critical traffic components such as roads, buildings, and trees from publicly available digital maps of the earth. Attributes such as size, orientation, location, and color can be calculated and utilized to recreate the environment in 3D using computer graphics algorithms. A mesh is generated for the roads by extruding the segmented road data while components such as buildings and trees are selected from a library of predefined models and matched based on attribute values. The result is a large 3D representation of a 2D map image that can be used for traffic research. The output of this task significantly reduces the workload for creating traffic scenarios in a 3D environment.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

dissertation

Discipline thesis:degree_discipline

Computer engineering

Department dc:contributor.department

Department of Electrical and Computer Engineering

Grantor

Iowa State University

Year dc:date.issued

2024

Author dc:creator

• Miller, Jack

Advisors dc:contributor.advisor

• Winer, Eliot
• Gilbert, Stephen
• Zhang, Hongwei
• Dorneich, Michael
• Li, Beiwen

Language dc:language.iso

en

DOI

https://doi.org/10.31274/td-20250502-258

OAI identifier oai:identifier

oai:dr.lib.iastate.edu:20.500.12876/6wBlGmnr