Preventing Intersection Rear-End Collisions with an Optimized Dynamic Two-Stage Actuated Control

Abstract

To minimize the likelihood of having rear-end collisions at signalized intersections, this study presents two algorithms for use with an actuated signal controller and the wide-ranger sensor, deployed for dynamic all-red extension. The first algorithm features its replacement of conventional gap-out control with vehicles detected in a field-calibrated dilemma zone and executes a phase transition only if no vehicle is within the detected zone. To further assure no rear-end collisions occur during either the defaulted green max-out or extended max-out, this study has further proposed a dynamic two-stage control algorithm that can optimally divide the duration between the minimum and maximum greens into two stages, where the signal within stage-1 duration will operate the first algorithm for green termination to ensure its efficiency, but shall execute the phase transition within stage-2 only at the time point projected to have the least likelihood of incurring rear-end collisions between any pairs of leading-following vehicles within the sensor’s monitoring zone. Because the effectiveness of such algorithms depends on the accuracy and reliability of their key parameters, the study has detailed their operational logic and the procedures for system calibration with field data, traffic simulator, and optimal searching heuristics. A predeployment evaluation at one candidate intersection has also been conducted with field data and a well-calibrated traffic simulator. Results of extensive experimental analyses with both peak and midday data convincingly confirm the effectiveness of both proposed algorithms, based on the surrogate variable of vehicle conflicts for potential read-end collisions as used both in research and practice.