Experiment to Improve Estimation of Vehicle Queue Length at Metered On-Ramps

Two types of algorithms for on-ramp queue estimation are discussed; a Kalman filter and a conservation model. A volume-balancing ratio is introduced to both models to account for unavoidable detector miscounting behavior. Estimation results are compared with queue data observed in the field. The volume-balancing ratio improves both models. Although the conservation model may provide more accurate prediction with balanced volumes, the Kalman filter tends to provide better estimation when the volume-balancing ratio deviates from 1. Generally the Kalman filter provides a better prediction, but the conservation model is simpler to implement. Attempts to improve the Kalman filter further are also explored.