Loop Detector Data Diagnostics Based on Conservation-of-Vehicles Principle

Abstract

Inductive loop detectors are the most popular method for continuously collecting traffic data over an extended period and over large spatial areas. These data are used in several applications, including real-time traffic information graphical displays, traffic forecasting programs, and incident management systems. The accuracy and reliability of the data generated by the loop detectors ultimately determine the quality of the results of the end application. However, the main criticism of detector data is the high probability of error because of equipment malfunctions. Traditionally, gross errors are identified by using threshold checking on speed, volume, or occupancy observations, either individually or in combination at individual locations. A more robust approach would include a check for conservation of vehicles over a series of detectors. However, this approach has received little attention. A constrained nonlinear optimization approach is presented for identifying and correcting loop detector data obtained from the field, in situations in which the data violate the vehicle conservation principle. The generalized reduced gradient method is adopted, and the objective function and constraints are selected so that the result will follow the conservation principle with the least change of the original data. This proposed technique can also be used to impute missing data and to locate suspect detector stations among a series of detector locations. Loop detector data from San Antonio are used as a test bed.