Measurement and estimation of traffic oscillation properties

The paper proposes a frequency spectrum analysis approach to improve measurements of traffic oscillation properties (e.g., periodicity, magnitude) from field data. The approach builds on standard signal processing techniques to effectively distinguish useful oscillation information from noise and nonstationary traffic trends. Compared with conventional time-domain methods, the proposed methodology systematically provides a range of information on oscillation properties. This paper also shows how to estimate oscillations experienced by drivers using detector data. Applications to real-world data from two sites show that the dominant oscillation period remains relatively invariant at each site when an oscillation propagates. Although the average oscillation periods displayed in detector data significantly vary across sites, the range of oscillations experienced by drivers are found to be more consistent.