Sufficiency of detector information under incomplete configuration for intersection OD estimation

To generate effective control strategy, traffic signal control systems require reliable estimates of turning movement flows. However, current control systems are often hungry for extensive detector information from a rich detector configuration. Due to various factors, maintaining satisfactory functionality in surveillance systems, although not impossible, is somehow difficult. A question is raised as to whether the surveillance system provides sufficient detector information for the controllers. Detector information sufficiency can be evaluated in both temporal or spatial dimensions. This paper focuses on investigating the detector information quality as a function of the density of spatial distribution. With a full set of detectors at intersections, it has been established that useful estimates of turning movement proportions can be generated using OLS-based methods, given time-series of input and output counts are available. A partial set of detectors will provide less information, leading to estimates with less accuracy and greater variability. The important issue is how dense does the detector configuration need to be to provide sufficient information to support parameter estimation. This paper investigates this issue based on the notion of identifiability. A numerical test is also proposed to evaluate the sufficiency of particular detector configuration.