27 Transportation Research Record: Journal of the Transportation Research Board, No. 2308, Transportation Research Board of the National Academies, Washington, D.C., 2012, pp. 27–36. DOI: 10.3141/2308-04 C. Bachmann, B. Abdulhai, and M. J. Roorda, Department of Civil Engineering, University of Toronto, 35 Saint George Street, Toronto, Ontario M5S 1A4, Canada. B. Moshiri, School of Electrical and Computer Engineering, University of Tehran, Tehran, Iran. Corresponding author: C. Bachmann, c.bachmann@utoronto.ca. • Reliability, robustness, and redundancy. A system that depends on a single source of input is not robust in that if the single source fails to function properly, the whole system operation will fail. However, a system fusing several sources of data has a higher fault tolerance because multiple sensors providing redundant information serve to increase reliability in the case of sensor error or failure; • Accuracy and certainty. Combining readings from several kinds of sensors can give a system more accurate information. Combining several readings from the same sensor makes a system less sensitive to noise and temporary malfunctions; • Completeness, coverage, and complementarity. Multiple data sources provide extended coverage of information on an observed object or state. Extended coverage is particularly relevant in spatial and temporal environments for the sake of completeness; and • Representation. Data fusion addresses the problem of informa- tion overload. The amount of time required for a person to make a decision increases as the amount of information available increases. By combining data and clearly presenting the best interpretation of the data, data fusion allows the user to make a well-informed and timely decision. In summary, data integration is about bringing data together in one place. Data fusion is about using the data together such that there is some new or better inference to be had. Together, data integration and data fusion create a variety of benefits. RESEARCH OBJECTIVES The objective of this paper is to demonstrate how data integration and data fusion can benefit traffic operations and management. The next section discusses the multiple sources of data conventionally available for traffic monitoring and some emerging technologies. The third section discusses several data fusion techniques for fusing data from competitive sensor configurations, describes the analytical foundation of these techniques, and interprets how each technique might be used most appropriately. The fourth section uses a case study with real-world data to show how data integration and data fusion can improve traffic monitoring and evaluates the accuracy and reliability of each technique. The final section summarizes the main conclusions of this research. DATA INTEGRATION FOR TRAFFIC MANAGEMENT AND OPERATIONS Conventional Sources of Traffic Data Loop detectors are the most widely used sensors in freeway traffic management systems because of their reliability in data measurements and flexibility in design (6). The main function of loop detectors is Multisensor Data Integration and Fusion in Traffic Operations and Management Chris Bachmann, Baher Abdulhai, Matthew J. Roorda, and Behzad Moshiri Widespread technological development and deployment have created an abundance of data sources for traffic monitoring. A database that integrates data from all these technologies would maximize coverage of the network, given the available data. Sometimes, however, there are multiple independent measurements of the current traffic conditions for a particular portion of the network. In these cases, a variety of data fusion techniques can be used to achieve better estimates while helping to overcome information overload. This paper discusses several tech- niques for fusing data from competitive sensor configurations, describes the analytical foundation of these techniques, and interprets how each technique might be used most appropriately. In addition, these data fusion techniques are implemented and compared relative to their ability to accurately and reliably estimate traffic speeds. A real-world case study in Toronto, Ontario, Canada, demonstrates that estimates from data fusion techniques that pull loop detector data and probe vehicle data from an integrated database are more accurate and reliable than estimates based on individual data sources. Consequently, these data fusion–based estimates can be taken with greater certainty and confidence. Data integration is the development of a database in which the resid- ing data come from different sources. When users run a query in an integrated database, they may or may not know that the data have been integrated from multiple sources. Data integration has become an increasingly common process as data become more plentiful, and the need or desire to share data becomes more apparent. Hav- ing multiple sources of data in an integrated database also provides opportunities for data fusion. Data fusion is the process of combining data such that the resulting inferences are in some way superior to those based on independent data sources. The difficulty of data fusion varies depending on how the data are being fused. Fusing data from complementary sensors, that is, sensors that can be combined to provide a more complete image of the phenomenon under study, is generally an easy task as these data merely solve the problem of incompleteness. In contrast, fusing data from competitive sensors, that is, those that provide independent measures of the same property, can be more challenging as the quality of these data sources must be evaluated. Depending on the problem, data integration and fusion can real- ize a variety of benefits (1–5). Their effects can be realized in areas such as the following: