Copyright © 2017 Jing Zhao et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

The dynamic lane assignment at signalized intersections is a possible countermeasure to address the traffic demand variability problem. However, the operational efficiency is affected by the unfamiliarity of the drivers. This paper evaluates the operational efficiency of the intersections with dynamic lane assignment using field data collected at five intersections in China. A total of 63488 vehicles were captured, which were divided into four groups according to the lanes they drive on: Group 1, the variable approach lane; group 2, the lane adjacent to the variable lane with the same lane-use; group 3, the lane adjacent to the variable lane with different lane-use of the variable lane; and group 4, the lanes with the same lane-use as the variable lane at other approaches of the intersection. The statistical analysis was conducted to identify the difference of saturation flow rate among the four groups. A saturation flow rate adjustment model was established accordingly. Results indicate that the using of the dynamic lane assignment decreases the saturation flow rate of the variable lane and the adjacent lane with different lane-use 22.86% and 9.80%, respectively. For the variable lane, the reduction of the saturation flow rate comes from three aspects: the unequal distribution of traffic (8.9%), the mandatory lane-changing (10.7%), and the lane blockage (4.9%).

1. Introduction

In many cities, congestion at intersections continues to worsen as the direct result of ever-increasing traffic demand. Moreover, real-world travel demands are intrinsically fluctuating, and traffic flows at intersections may vary significantly even for the same time of day and day of week [1]. The variability in volumes of turning movements exacerbates the perennial problem of congestion [2].

As one of numerous possible countermeasures to address the traffic demand variability problem, the concept of dynamic lane assignment at signalized intersections was proposed and has been carried out in practice [3, 4]. Different from the commonly used signal timing method, the dynamic lane assignment method directly changes the lane-use dynamically according to the traffic demand. Therefore, it is designed to accommodate situations in which the relationship between supply and demand varies significantly [5].

To date, many lane assignment optimal models have been established to produce the best operational strategy for the intersections. At the beginning, the lane assignment was designed on a trial-and-error basis. The design scheme was revised by the engineer according to his/her experience until the performance of the intersection was satisfactory. This method, however, may not always produce the optimal set of lane markings [6, 7]. Then an integrated model with mix-integer program was proposed to minimize the sum of flow ratios of all the phases [6]. Results show that substantial improvement in intersection performance can be achieved by combining the design of lane markings and signal timings into a unified framework. Along the same line as the integrated design method, Wong et al. [7–9] further developed lane assignment optimization models using the lane-based approach. In these lane-based models, all key design variables were given on a lane basis. This simplifies the expression of the set of constraints as linear equations and thus ensures the feasibility of the solution algorithm for the optimization model. Wong et al. also use this method to deal with the problem that the traffic demand varies in various periods, for example, the morning peak, off peak, and evening peak periods [10, 11].

However, the computing times may be a problem using these models into online settings for real dynamic optimization because of the model complexity. To overcome the computing time difficulty, some simplified models was established to reduce the total number of variables. Zhong et al. narrowed the study scope from an intersection to an approach [12]. An optimization model was established to minimize the delay of an approach by dynamically changing the number of through and left-turn lanes at the approach. Zeng et al. [13] reduced the number of variables by fixing the signal control policy, which was proposed by Webster [14], in his integrative model. Then, a two-step optim