January 7-11, 2018
Washington, D.C.
TRB 97th Annual Meeting

Big Data Applications and Methods in Transportation

Sunday, January 07, 2018, 1:30 PM-4:30 PM, Convention Center
Mecit Cetin, Old Dominion University; Kristin Tufte, Portland State University; Francisco Pereira, Technical University of Denmark, presiding
Sponsored by Standing Committee on Urban Transportation Data and Information Systems; Standing Committee on Information Systems and Technology; Standing Committee on Geographic Information Science and Applications; Standing Committee on Artificial Intelligence and Advanced Computing Applications; Standing Committee on Visualization in Transportation; Standing Committee on Transportation Demand Forecasting; and Standing Committee on Transportation Issues in Major Cities

As the volume of transportation data grows, users grapple with the basics of big data implementation. The highest-priority request from more than 200 attendees of the 2017 Big Data Analytics Workshop was for case studies and examples of implementation. This workshop provides case studies of big data methods used in practice and hands-on analysis. The results of the workshop will be documented in a workbook and disseminated through the sponsoring committees.

Maricopa Association of Governments: A Case Study
Vladimir Livshits, Maricopa Association of Governments
Big Data and Big Data applications became a necessary part of analytical and forecasting work for many planning agencies. MAG case study will highlight applications that de-facto became state -of-the-practice for many large MPOs and DOTs, applications that are relatively new and have not yet received wide acceptance, and innovative techniques and applications that has been recently developed by MAG. The presentation also will discuss methods for data validation and QC and data applicability for specific analytical purposes. Specific data set utilized in the case study will include speed data, truck GPS data, vehicle traces and personal travel trajectories from time-lapse aerial photography and GPS surveys, origin-destination and volume data from commercial vendors.
Deep Learning in Transportation
Anuj Sharma, Iowa State University

Transportation  agencies  and  Departments  of  Transportation  (DOTs)  rely  on  a  dense  network  of  CCTV cameras primarily for real time viewing and monitoring of traffic conditions on roadways .  Very  minimal  to  no  automated  analytics  is  performed  on  such  high-valued  asset.  This  presentation  talks about computer  vision  systems  that  leverages  recent  developments  in  high  performance  computing and advanced machine learning algorithms to maximize the use of data streaming from  CCTV cameras. The system could potentially provide various video analytic capabilities that will enable CCTV  cameras to automatically extract important traffic information: traffic data (volume, density, speed  and  vehicle  classification),  traffic  condition  (congested,  stalled  vehicle,  free  – flow)  and  road  condition information (snow, wet or dry pavement).

Tools for Management and Monitoring of Transport Big Data
Nikola Ivanov, University of Maryland, College Park

This presentation will briefly introduce several specific technical concepts related to storage and analysis of transportation big data, including distributed computing, Hadoop, and related ecosystem tools. The main portion of the presentation will focus on use of those back-end technologies to power big data analytics and tools to solve specific real-world transportation operations and planning problems. Finally, the presentation will address emerging technologies, such as blockchain, and their potential impact on transportation.

UC Berkeley's and Caltrans' New Cloud Based Data Hub
Qijian Gan, University of California, Berkeley

This presentation introduces the most recent development on the I-210 Connected Corridors Project, with particular emphasis on the new cloud based data hub. This presentation consists of two parts. In the first part, we will talk about the cloud based architecture and how we integrate various traffic systems and standardize heterogeneous data feeds using a common data format (e.g., TMDD). We will also talk about the data streaming technique using Kafka, the data storage methods using relational (e.g., Postgres) and non-relational (e.g., MongoDB and Cassandra) databases, and the data processing framework using SPARK.  In the second part, we will talk about how to utilize the data to support the development of the decision support system, which includes the topics of arterial performance dashboard, traffic state estimation, Aimsun modeling, calibration and prediction, and development of the rules engine.