Comparison of Queue-Length Models at Signalized Intersections

Abstract

Several measures of effectiveness (MOEs) are associated with the queuing process at traffic signals, including delay, number of stops, fuel consumption, emissions, and queue length. The focus in this study is on queue length in general and on the storage requirements for left turns in particular. Queue length is an important MOE because queues that overflow the available storage space have an adverse effect on the overall operation of the intersection. Many traffic models now provide queue-length estimates, but the procedures used by these models are based on different queue definitions and have different computational approaches that lead to different results. A classification framework is developed for the existing models, their behavior is compared with that of the proposed Highway Capacity Manual (HCM) 2000 queue model, and queue conversion factors are provided for translating the various model outputs to their HCM 2000 equivalent. The proposed HCM 2000 model and its parent model from the Signalized and Unsignalized Intersection Design and Research Aid (SIDRA) provide a comprehensive treatment of the queuing process, accounting for control parameters such as controller type and progression quality as well as for the random and overflow effects associated with traffic flow. As such, the queue-length estimates from these models are more analytically defensible than those of the simpler theoretical models. The SIDRA and HCM 2000 queue estimates are generally higher than those of most other models and are somewhat higher than what conventional wisdom would suggest. It is suggested as a result of the comparisons presented that the queue estimates from some models are unduly optimistic when demand approaches capacity and that a goal of 90 percent confidence in the adequacy of left-turn storage lanes may be difficult to achieve under these conditions.