Fingerprinting Traffic From Static Freeway Sensors

Ask most commuters and they will agree that congestion has reached an intolerable level. To reduce this congestion, engineers need detailed traffic information. Highly detailed information is also prized by traffic scientists, as a prerequisite to improve current traffic theories. Ideally, engineers and scientists would like to obtain from field data the position of each vehicle on a particular facility at every moment in time. The technology to record space-time vehicle trajectories on a massive scale is in its infancy; therefore, analysts must work with much less data. Many freeways are equipped with primitive sensors that can record only anonymous vehicle passages at specific locations with a time series of 0's and 1's. Typically, these detectors are installed on all lanes at sites, called stations, which are spaced about 1 km apart. This article will show that, despite the anonymity and the spatial discreteness of the measurements, a treasure trove of detailed information can be recovered from 0-1 detector data, if one analyzes the data with the right tools. Field data from a 5-lane freeway in Oakland, California (see Figure 1) is used to demonstrate the ideas.