BACKGROUND
Data fusion is the process of integrating multiple data sources to produce more consistent, accurate, and comprehensive information than that provided by any individual data source. In a transportation context, transportation agencies are seeking to define the types and characteristics of data for entry into data fusion engines. Research is needed to identify the challenges, issues, and proven or potential practices for performing data fusion to measure or forecast travel time, speed, reliability, and other aspects of operational performance on roadway networks. Traffic datasets of interest include point sensors; Bluetooth; data from GPS devices embedded in smartphones, personal navigation devices, taxis, and fleets; third-party travel time data; and emerging connected vehicle (CV) data sets.
Research that enables better knowledge of the network state could help improve traffic management and planning decisions to address impacts of recurrent and non-recurrent (e.g., incident-related) congestion. Improved network state estimates could also enhance safety outcomes by identifying locations with high crash rates and anomalous traffic flow conditions.
OBJECTIVES
The objectives of this research are to:
1. Develop a process to (a) identify specific objectives for data fusion; (b) identify data sources available for fusion; (c) select the most suitable data for fusion; and (d) facilitate the fusion itself.
2. Develop guidelines for transportation agencies to facilitate data fusion, improve data reporting, and ultimately improve traffic management.
Accomplishment of the project objective will require at least the following tasks.
RESEARCH TASKS
Task 1. Determine data fusion state of practice and develop stakeholder engagement plan.
Task 1a. Review literature and research in progress. The review shall include (a) research by public agencies, academia, and industry; and (b) relevant resources from outside the transportation sector.
Task 1b. Develop draft survey of stakeholders on data fusion. Identify stakeholders to be surveyed in Task 2 and develop draft survey. The survey shall be aimed at identifying stakeholder objectives for data fusion, data sources utilized, and challenges with data fusion. Both public and private sectors shall be included in the survey and organizations within and outside the transportation sector, including the AASHTO Committee on Data Management and Analytics and pertinent TRB committees.
Submit Technical Memorandum 1 covering the findings of Tasks 1a and 1b. Present findings to NCHRP at an online meeting. NCHRP approval of Technical Memorandum 1 is required before proceeding with subsequent tasks.
Task 2. Conduct stakeholder survey and develop recommendations on research deliverables.
Task 2a. Survey stakeholders. The survey shall initially be conducted electronically, with follow-up contacts by phone as needed. The survey shall be sent to all pertinent personnel within an agency or firm.
Task 2b. Develop draft plan for project deliverables and evaluation. First, a set of potential data fusion use cases shall be defined. Based on the identified use cases and the results of the literature review and survey, a set of candidate data fusion methods for testing shall be identified. Data fusion methods shall be selected by identifying those that can be reasonably integrated into transportation agency operations and that are computationally feasible to encourage agencies to adopt them. A series of data sets shall be compiled to assess the accuracy of the various methods and identify the advantages and limitations of the methods across a range of facility types and traffic conditions.
Task 3. Submit interim report. The report shall summarize the results of Task 2, including the survey, recommendations on research deliverables to be developed in subsequent tasks, and draft development and evaluation plans. The interim report shall be presented to the panel at an in-person or online meeting. NCHRP approval of the interim report is required before proceeding with subsequent tasks.
Task 4. Develop and evaluate deliverables.
Task 4a. Deliverable development. Develop project deliverables using the findings from Tasks 1-3. These deliverables must be suitable for evaluation per specifications in Task 4b.
Task 4b. Deliverable evaluation. Evaluation of the deliverables developed in Task 4 shall be conducted using real data from real networks. The evaluation shall include multiple data sources, e.g., probe data and point sensors.
Task 5. Develop implementation guidelines: To support data fusion implementation in real-world applications, the guidelines shall, to the extent possible, be data source-independent and allow for minimum agency expertise in data fusion algorithms and modeling. Submit the draft guidelines to NCHRP for panel review. NCHRP approval of the guidelines is required before proceeding with remaining project tasks.
Task 6. Submit final report and other project deliverables. Final deliverables include (1) the guidelines, (2) a conduct of research report describing how the research was conducted, (3) an implementation plan (see Special Note C), and (4) presentation slides summarizing the research and its results.
Note: Following receipt of the draft final deliverables, the remaining three (3) months shall be for NCHRP review and comment and for research agency preparation of the final deliverables.
STATUS: Research is complete; publication decision is pending.
