A framework for analyzing the sensitivity of traffic data quality to sensor location and spacing

This paper presents a framework and tools developed to study the sensitivity of traffic data quality to detectors location and spacing. Our ultimate objective is to formulate generalized detector deployment guidelines that are based on the functional needs of practitioners, and for which funding can be objectively justified. Our approach consists in using trajectory sets obtained from field experiments and traffic simulation models as ground truth, and to run a traffic detector model from which we extract information that would normally be available to practitioners. Ground truth information and detector-generated information are compared through selected quality benchmark measures, and we search detector configurations that optimize this comparison. We test both model-based and so-called naïve traffic estimation techniques, and find that while the former is superior, the difference becomes negligible as detector density increases. 1/2 mile spacing seems to always yield reasonably good information, but no such analysis should overlook detector failure rates. We conclude that those must be taken into account in the formulation of deployment guidelines, a step we defer to further studies.