Queue length estimation from connected vehicles with range measurement sensors at traffic signals

Abstract

This paper presents a method for queue length estimation from connected vehicles equipped with range measurement sensors such as light detection and ranging, stereo camera, and ultrasonic. Simple plug-and-play models are developed for queue length estimations that do not require ground truth queue lengths. The proposed method is simple to implement and can be adopted to cyclic queues like traffic signals with known phase lengths. The models are evaluated with microscopic traffic simulations and connected vehicle field test data. Compared to queue length estimations without range sensors, the numerical experiments performed using microsimulations show that the model with range sensors improves the errors by 25% in variance-to-mean ratio and 5% in coefficient of variation at less than 20% market penetration rates. Using the connected vehicle field test data, it is observed that the fusing range sensor can improve close to 10% without sensor estimators, 30% of the queue back, and 48% of the Highway Capacity Manual’s control delay-based estimations at about 25% market penetration rate.