Data Fusion of Probe and Point Sensor Data: A Guide (2024)
Chapter: Summary
SUMMARY
Data Fusion of Probe and Point Sensor Data: A Guide
Transportation agencies have long invested in traffic sensors installed on freeways and arterials that collect data such as vehicle speeds and volumes. However, many agencies are no longer investing in these sensors. Instead, they are focusing more on probe-based speed data and volume estimation products and sensors which are installed, operated, and maintained by the private sector. Probe data have the advantage of not requiring agency infrastructure or maintenance, and they can theoretically cover any and all roadways, at all hours of the day, at significantly lower prices than sensors. While often viewed as competing datasets, neither probe nor sensor data are perfect, and neither by itself is currently able to meet all of the needs of agencies. Even private-sector probe data providers must rely on some sensors to calibrate their volume estimation products.
Researchers have identified transportation planning and operations use cases that have the potential to benefit from the combination (or fusion) of probe and sensor data. For example, when volume data from sensors at specific locations along a corridor are combined with probe-based speed data covering the entirety of the same corridor, one can more readily understand significantly more about the real-time and historic performance of the roadway. Researchers can obtain real measurements of user delay, the vehicle hours of delay, the cost of delay, and even estimates of fuel consumption and emissions. Another example use case benefits agencies where probe data quality may be in question, especially on low-volume roadways. Strategically placed point-sensor data may be used to improve the quality of probe data speed measurements or even help in predicting queues and detecting incidents.
Data fusion is a broad topic. This paper covers specific use cases, benefits, and methods for fusing two classes of data used in transportation operations and planning. These two datasets include 1) vehicle speeds and volumes collected from point sensors such as side-fired microwave radar sensors, video detection systems, or inductive loops, and 2) probe-based vehicle speed data from a variety of companies.
This report is divided up into several sections, including a broad overview of data, considerations before embarking on a data fusion exercise, a detailed description of a proposed framework for data fusion of point and probe data, and then a couple of detailed use cases from the real world that show how the framework can and has been applied by agencies.
This document is divided into chapters to serve multiple purposes and audiences. Chapters 1 through 4 are meant for all transportation professionals regardless of background or position. Chapter 5 has been divided into sections that describe each step of the framework. Additionally, each step of the framework has sections written for specific audiences including:
- For the Executive: This is a high-level overview of the data fusion framework step, its importance to the agency, with enough justification and guidance to help the executive better understand its relevance.
- For the Transportation Systems Management and Operations (TSMO) Professional: This section is meant for TSMO professionals who may be interacting with systems implementers. The section strives to give TSMO professionals enough knowledge to be able to have intelligent and meaningful conversations with systems implementers without going into too much detail or math.
- For the Systems Implementer: This section provides deeper details that someone charged with implementing the data fusion algorithms and technologies would likely need to know.
Chapter 6 provides two real-world examples of point sensor and probe data fusion, and Chapter 7 provides additional reference materials that systems implementers may find helpful.
The data fusion space is huge, with hundreds of potential fusion algorithms applicable to point sensor and probe data. This document only scratches the surface of the realm of possibilities in data fusion. Where feasible, there are references to other data fusion algorithm books, papers, and related resources.
This document and the included fusion framework strive to be descriptive; however, this is not a how-to manual. The example implementation code contained in later chapters is meant to be illustrative of techniques that inspire an implementer to build fusion systems for specific use cases. They may not be directly transferable to a specific agency or system.
Last, while there are samples of code shown, this document does not teach the basics of relational database management systems, Hadoop-style distributed file sharing systems, no-SQL technologies, generative artificial intelligence (AI) algorithms, scripting, information technology (IT)/server configuration, or other software development technical implementation basics. It is expected that a systems implementer will have the competencies to implement the framework.
For those who are not systems implementers or software developers, sections of this document are designed to give transportation professionals enough knowledge to be able to communicate effectively with systems implementers.
