Cycle-by-Cycle Analysis of Signalized Intersections for Varying Traffic Conditions with Erroneous Detector Data

Analysis of traffic at a signalized intersection requires quantification of the number of vehicles in the queue and the corresponding delay. Such analyses can be used for optimal signal control as well as for intelligent transportation system (ITS) applications such as advanced traveler information systems (ATISs). However, the direct measurement of these variables is difficult because of their spatial nature. Hence, they are usually estimated using location-based data such as count, speed, occupancy, and so on that can be obtained from point-based detectors such as loop detectors installed on roads. However, driving maneuvers such as lane shifts and free right turns can lead to inaccurate queue estimates in lane-based analysis. To check this, analysis of lane-based data was carried out and discrepancies in the count data obtained from loop detectors were observed. To address these issues, the present study proposed a model-based queue estimation scheme using the Kalman filtering technique, taking into account the statistical properties of detector errors. The detector data and the signal timing information were used as inputs in the estimation scheme. The estimation was carried out for two cases-one where the queue ends within the advance detector and one in which the queue extends beyond the advance detector. Field data collected from four different intersections were used to corroborate the estimation scheme for the "queue within advance detector" scenario. Because of the lack of availability of field data for "queue beyond advance detector," simulated data were used to corroborate the corresponding results. Results showed that the estimation scheme that incorporated the statistical properties of the detector errors performed better than the scheme that did not incorporate errors. (C) 2017 American Society of Civil Engineers.