Journal of Transportation Engineering, Part A: Systems

The interaction between pedestrians and right-turning vehicles at a signalized intersection is a usual scenario in urban locations in which both have to share a common space on the road. Due to the rapid growth of vehicular traffic in developing countries such as India, pedestrian–vehicle conflicts are increasing tremendously. This study evaluated pedestrian safety by examining the interaction between pedestrians and right-turning traffic (left-hand driving) at signalized intersections using the traffic conflict technique. Five time-based conflict indicators, namely postencroachment time (PET), time to vehicle (TTV), deceleration to safety time of pedestrians (${\mathrm{DST}}_{\mathrm{ped}}$), time to accident (TTA), and deceleration to safety time of vehicles (${\mathrm{DST}}_{\mathrm{veh}}$) were used to analyze the conflict situation. Four signalized intersections were selected in two cities (Kolkata and New Delhi) of India. For every possible conflict situation, PET, TTV, ${\mathrm{DST}}_{\mathrm{ped}}$, TTA, and ${\mathrm{DST}}_{\mathrm{veh}}$ were estimated and the severity of conflicts was identified based on pedestrian demography and their crossing behavior. The $k$-means clustering technique was used to classify the conflict indicators into four levels of severity. Silhouette plots were developed to validate the clusters. A binary logistic regression model was developed to identify significant contributing factors to the risk-taking behavior of pedestrians. The model’s results showed that pedestrians’ age, gender, waiting time, and speed; type of crossing; pedestrian with company; the occurrence of conflicts in different quarters of the green interval; and right-turning vehicle volume have a significant effect on the risk-taking behavior of pedestrians. The findings of this study may help to identify various factors that affect pedestrian-vehicle conflicts at signalized intersections.

Estimation of passenger car unit (PCU) values is very important for traffic capacity analysis and other relevant applications such as level of service (LOS) measures, determination of saturation flow rate, signal design and coordination, and development of traffic flow models. Because of such wide applications, the accuracy of PCU values is highly significant in traffic flow analysis. Many research works have been carried out on estimation of PCUs in the past few decades. In developed countries, various methods were developed for estimating PCU values for different types of facilities which carry trucks and buses of lesser composition. But these methods are not completely analogous for mixed traffic due to the presence of wide variety of vehicle types, nonlane discipline, intraclass variability of vehicles and their maneuverability. With this consideration, various methods used for estimating PCUs for different facility types under homogeneous and mixed traffic conditions are reviewed and the drawbacks of the existing methods are identified in this paper. The challenges in estimating PCU values and future directions for improving the PCU estimation methods are also presented.

An unconventional left-turn treatment called contraflow left-turn lane (CLL) design has been increasingly used in China to relieve traffic congestion associated with left-turn movements at signalized intersections. This study proposed a procedure for estimating the left-turn queue length at signalized intersections with the CLL design. Field data were collected at six approaches at five signalized intersections in the city of Handan, China, and 40 h of traffic data were recorded. A binary logit model was developed to estimate the probability of a driver stopping at the presignal when there are still vacant spaces in the conventional left-turn lane between the main and presignals. A modified shockwave-based method was then proposed for estimating the maximum left-turn queue length for the signalized intersections with the CLL design by considering the unique queuing behavior at the presignal. Sensitivity analyses were conducted to identify the impacts of the key parameters in the proposed queue length model on the maximum left-turn queue length. The results suggest that the effective red time duration, the offset between the main and presignals, and the queuing behavior of left-turning vehicles at the presignals greatly affect the maximum left-turn queue length. The proposed queue-length model was validated with the field data considering varying left-turn traffic demands. The results indicated that the proposed model provides reasonable estimates for the maximum left-turn queue length at the signalized intersections with the CLL design. A procedure was also proposed to help determine the minimum distance between the upstream and downstream signalized intersections for the deployment of the CLL design.

Variable speed limits (VSLs) are components of active traffic management (ATM) systems that are increasingly being used as safety and operational treatments in urban areas. Studies have shown safety and operational benefits associated with VSLs; however, the level of benefit is dependent on the degree to which drivers comply with the VSL. Although several studies have quantified macroscopic safety and operational benefits of VSL, limited studies have examined the degree to which drivers adhere to the posted VSL signs. This study evaluated how speed changed in response to an advisory VSL system on Interstate-66 (I-66) in Northern Virginia. Compliance in this study was defined as when vehicle speeds in a time interval were at or less than the speed posted on an advisory VSL sign. The study generally revealed that noncompliance increased as the posted VSL speed increased. The analysis further showed that the observed speed of all drivers was at or below the posted VSLs’ speed 37% and 40% of the time downstream and upstream of I-66 VSL gantries, respectively. About 47% and 51% of observed interval speeds exceeded the posted VSL by more than 2.235 m/s ($5\mathrm{mi}/\mathrm{h}$) downstream and upstream of the VSL gantries, respectively. The analysis also showed relatively higher compliance levels when VSLs were spaced closer together and had lower VSL posted speeds than when higher VSLs were posted with longer distances between gantries. These higher levels of compliance at low VSLs likely indicate that traffic was restricting the ability to travel above the posted VSL. The results comparing speeds upstream of gantries showed a similar pattern based on the speed posted, but no pattern was observed with respect to compliance over space. This study may be useful to transportation and law enforcement agencies in determining when to perform targeted enforcement and in planning for future deployments.

As traffic demands increase rapidly on highways, effective merging strategies are necessary to cooperate with intelligent vehicles and improve traffic operations. Existing merging algorithms for connected vehicles rarely consider the inflow from on-ramps. Also, the merging order of vehicles is generally generated based on very simple rules. In this paper, a cooperative merging strategy is developed for vehicles wirelessly connected to other vehicles and roadside infrastructure. The cooperative merging is formulated as an optimization problem, which takes as objectives the minimization of travel time of mainline vehicles and maximization of the number of merging vehicles. This problem is solved by a genetic algorithm. The effectiveness of this strategy is verified in MATLAB with various simulation scenarios. By comparing the simulation results with a platoon-velocity-based merging strategy, the cooperative merging scheme proves to improve traffic performance in terms of traffic efficiency and fuel consumption. Significant improvements are obtained, especially when mainline and on-ramp demand are both particularly high. To conclude, the proposed strategy is applicable to cooperative merging operations under saturated traffic conditions.

Heterogeneous traffic conditions prevail in all countries across the world, but the degree of heterogeneity is different in developed and developing countries. In the United States, heavy vehicles (trucks and buses) are mixed with passenger cars, whereas more than seven different categories of vehicles in India are observed in the traffic stream. Such heterogeneity is also observed at tollbooths, where dedicated lanes are provided for each vehicle category; however, during peak hours, lane discipline is broken, and the same lane is used by different categories of vehicles. Such mixed traffic conditions lead to wide variations in service time for the same vehicle category and, hence, reduce tollbooth capacity. To convert the mixed traffic flow into the equivalent passenger car, the tollbooth equivalency factor (TEF) is proposed in the present study. The TEF is based on the service time and clearance time of a vehicle with respect to a passenger car at the same tollbooth. The TEF is found not to be a fixed value but varies with the approaching mixed traffic volume and composition of traffic in the queue. Vehicles in the traffic stream are divided into seven categories, and simultaneous equations are developed to determine the service time of a vehicle type based on approach volume and traffic composition. These equations are further used to depict variations in TEF with varying approach volume and composition. The change in TEF values is explained on the basis of the relative change in service time of a vehicle type with respect to a standard car at different volume levels. The accuracy of the TEF values estimated through simultaneous equations is checked by comparing the estimated values with those calculated directly from the field data. TEF values obtained for each type of vehicle are multiplied by the number of that vehicle type to obtain the approaching mixed traffic volume in an equivalent homogenous mix ($\mathrm{TEF}/\mathrm{h}$). This method will naturally require an estimation of TEF for each vehicle type from the field data. To avoid this exercise, the stream equivalency factor (SEF) at the tollbooth is proposed, which provides an overall multiplying factor for the entire traffic volume to convert heterogeneous traffic into the homogenous equivalent. The results presented in this paper will be useful for planners and tollbooth managers for the design and performance evaluation of toll plazas and to identify the number of lanes required during peak and nonpeak hours.

Calibration factors are applied in the Highway Safety Manual predictive method for rural two-lane, two-way roadway segments to adjust the estimate for local conditions. This research aims to evaluate and recommend improvements related to the estimation of these calibration factors. An aggregated and disaggregated analysis was performed to study the influence of different calibration factors on the prediction of the number of crashes in North Carolina. As a result, those calibration factors based on both types of road elements (horizontal curves and tangents) led to overestimating and underestimating the number of crashes on tangents and horizontal curves, respectively. Furthermore, the calibration factors based on fatal and injury crashes allowed a more accurate estimation of the predicted number of crashes than those calibrated considering all severity levels. Therefore, it is recommended to apply different calibration factors for each type of road element and each type of crash severity.

Modeling efforts for driver behavior parameters on horizontal curves have mostly focused only on the 85th percentile value. However, predicting whole distributions would help improve alignment design by allowing reliability-based design and design consistency evaluation. This paper used naturalistic driving study data to model distributions of speed and comfort threshold on horizontal curves of two-lane rural highways. Several variables along the approach tangent and curve were extracted and examined. This analysis helped determine the driver behavior parameters needed to evaluate driver behavior on horizontal curves and the headway threshold for free-flow conditions. Driver level models (DLM) and panel models (PM) were developed to predict distributions of curve speed and comfort threshold in addition to the traditional 85th percentile models. The models developed can be used in evaluating vehicle stability, driver comfort, and design consistency. Thus, the models can act as the basis for reliability analysis of horizontal curves, for which analysis methods are already established but realistic data are relatively scarce.

This research compared kinematic and behavior parameters of light and heavy vehicles at turbo-roundabouts. This is of interest because heavy vehicles can influence turbo-roundabout functionality more than that of conventional roundabouts. The traffic arrival process at arms of a typical turbo-roundabout was studied. Using image-processing techniques, the trajectories of about 1,400 light vehicles and 300 heavy vehicles were examined at a roundabout entry, on the ring, and at the roundabout exit. Light vehicles had moderate average speeds lower than $25\mathrm{km}/\mathrm{h}$ at the entry lanes, 15 m prior of the yield line, whereas heavy vehicles had an average speed of about $20\mathrm{km}/\mathrm{h}$ on the same section. Significant differences were found in average acceleration: they were generally below $2\mathrm{m}/{\mathrm{s}}^2$ on arm lanes and $1.5\mathrm{m}/{\mathrm{s}}^2$ on circulating lanes for light vehicles, whereas those of heavy vehicles were $0.5\mathrm{m}/{\mathrm{s}}^2$, much lower than the corresponding acceleration values of light vehicles. The values of the critical gap $t_c$ and follow-up time $t_f$ estimated for light vehicles ranged between $t_c=4.03-5.48\mathrm{s}$ and $t_f=2.52–2.71\mathrm{s}$ respectively, whereas heavy vehicles had $t_c$ and $t_f$ values 15%–47% higher than those of light vehicles.