Signal Green Time Estimation Method for Connected Vehicle-to-Infrastructure Applications

Abstract

Connected and autonomous vehicles (CAV) are becoming more integrated with traffic signal infrastructure for V2I applications, such as traffic light indication and automated driving. However, modern traffic signal controllers allocate green time using vehicle sensors and therefore the anticipated green time has significant stochastic variation. This study develops a methodology to characterize green time stochastic variation for actuated-coordinated operation. During the peak hours where the demand was highly consistent, green intervals can be predicted with high certainty. In contrast, during midday and late evening, stochastic variation increased significantly due to the varying arrival patterns and associated real-time responsiveness of the traffic signal controller. The statistical characterization methods presented in this paper are important for green light optimized speed advisory (GLOSA) and eco-driving, technologies that rely on having an accurate estimate of the beginning of green (BOG) and end of green (EOG). Prior knowledge on typical values of how early to stop or shutdown the vehicles at a traffic signal approach can significantly improve efficiency and manage emissions for CAV. The paper concludes with a proposed graphical performance measure chart that can be used by traffic engineers and automotive vendors to frame the discussion on traffic signal operation.