The past few years have seen breakthroughs in automation technologies, enabling the development of Advanced Driver Assistance Systems (ADAS) and Automated Driving Systems (ADS). The introduction of such technologies to the transportation system has resulted in a new set of interactions among roadway users (e.g., human-ADAS and human-ADS interactions) that did not exist a few years ago. As the number of vehicles equipped with these technologies increases on our roadways, more and more interactions between these technologies and humans are expected. These interactions can shape the future of traffic flow theory and fundamentally change the way we model traffic flow dynamics. Unfortunately, our understanding of these interactions remains limited. Capturing the interactions between human drivers and Connected and Automated Vehicles (CAVs) is essential in addressing many future research questions as well as for the improvement of our Analysis, Modeling, and Simulation (AMS) tools and methodologies(1).

To date there have been three major ADAS and ADS data collection efforts funded by the Intelligent Transportation Systems Joint Program Office (ITS JPO) and FHWA. More are on their way. Under these efforts, trajectory rich ADAS/ADS datasets were collected in Northern Virginia, Ohio, Chicago, and Washington, DC. These datasets can be utilized for various traffic flow and safety analyses, including but not limited to (1) investigating the impacts of SAE Level 1 and SAE Level 2 ADAS-equipped vehicles on human behavior, (2) string stability of ADAS-driven vehicles in real-world settings, (3) dynamics and impacts of ADAS-heavy vehicle interactions on traffic flow and safety, (4) underlying dynamics of mandatory and discretionary lane-changing maneuvers, (5) car-following behavior under various traffic flow regimes, and (6) microscopic and macroscopic traffic flow analyses.

The aforementioned efforts focused on collecting the data and their success resulted in the assembly of very large and robust datasets. Unfortunately, as is the case with many Big Data cases, the size and robustness of a dataset is reversely correlated to its useability. Specifically, initial utilization of some of these datasets revealed the wealth of information contained within. However, the original data collection project scopes were relatively narrow, focused on the operations of project provided and operated CAVs and their interactions with the immediately surrounding conventional vehicles on public roads. Since the resulting presentation of the data and dataset format were organized around the project provided CAVs, there are limits to more general dataset usability without considerable effort. For, example, a University of Illinois project team collected data from a helicopter in Chicago and the data are primarily arranged to study the trajectory of the subject vehicles (e.g., the CAV) and vehicles immediately surrounding it. However, trajectories were captured for all the vehicles in the field of view of the camera. There is a tremendous wealth of trajectory data related to all vehicles in the dataset that can be mined and organized better to answer a broader set of research questions. Having secured these datasets, the FHWA is now focusing on the development of data mining, analysis, and visualization methodologies and tools that would increase the useability of these vehicle trajectory datasets and help unlock the wealth of knowledge they contain.

The Research Associate will assume a high level of responsibility for contributing to the methodology and tool development by working both independently and in collaboration with FHWA staff members at the Saxton Transportation Operations Laboratory (STOL), housed within the Turner-Fairbank Highway Research Center in McLean, VA. The goals of this research could be one or more of the following:

1.   Perform a literature search on data needs for AMS research. Focus will be given in information illustrating and exploring human and CAV driving behavior regarding car following, lane selection, and lane change.
2. Develop a framework for data mining and visualization of existing datasets.
3.  Use different conventional and emerging analytics for the development of derivative datasets
4.  Develop data mining techniques and predictive models that can test and evaluate causal relationships in the interactions between CAV and human driven vehicles.
5.  Incorporate those findings into analysis methodologies and tools to transfer them to practice

This overall objective is to develop data mining tools that will by themselves serve as a resource to future users of the datasets as well as expand the current datasets through post processing of the existing raw data.

FHWA is seeking candidates with innovative ideas and expertise in one or more of the following: highway safety, data science, statistical modeling, analyzing datasets involving geospatial information, large volumes of text data, and video data using Artificial Intelligence (AI) and Machine Learning (ML). Candidates demonstrating skills with statistical programs such as R, SAS and proficiency in using Python and Matlab are highly desirable.

(1). Mahmassani, H. S., Elfar, A., Shladover, S. E., & Huang, Z. (2018). Development of an Analysis/Modeling/Simulation (AMS) Framework for V2I and Connected/Automated Vehicle Environment (No. FHWA-JPO-18-725). United States. Department of Transportation. Intelligent Transportation Systems Joint Program Office.