Heng Wei

Study on continuous network design problem using simulated annealing and genetic algorithm

In general, a continuous network design problem (CNDP) is formulated as a bilevel program. The objective function at the upper level is defined as the total travel time on the network, plus total investment costs of link capacity expansions. The lower level problem is formulated as a certain traffic assignment model. It is well known that such bilevel program is nonconvex and algorithms for finding global optimal solutions are preferable to be used in solving it. Simulated annealing (SA) and genetic algorithm (GA) are two global methods and can then be used to determine the optimal solution of CNDP. Since the application of SA and GA on continuous network design on real transportation network requires solving traffic assignment model many times at each iteration of the algorithm, computation time needed is tremendous. It is important to compare the efficacy of the two methods and choose the more efficient one as reference method in practice. In this paper, the continuous network design problem has been studied using SA and GA on a simulated network. The lower level program is formulated as user equilibrium traffic assignment model and Frank-Wolf method is used to solve it. It is found that when demand is large, SA is more efficient than GA in solving CNDP, and much more computational effort is needed for GA to achieve the same optimal solution as SA. However, when demand is light, GA can reach a more optimal solution at the expense of more computation time. It is also found that increasing the iteration number at each temperature in SA does not necessarily improve solution. The finding in this example is different from [Karoonsoontawong, A., & Waller, S. T. (2006). Dynamic continuous network design problem - Linear bilevel programming and metaheuristic approaches. Transportation Research Record (1964), 104-117, Network Modeling 2006.]. The reason might be the bi-level model in this example is nonlinear while the bi-level model in their study is linear.

Examining Headway Distribution Models with Urban Freeway Loop Event Data

Vehicle headway distribution is fundamental for several important traffic research and simulation issues. Many headway models have been developed over the past decades. Each has its own strength and weakness. Selection of the most suitable model for a certain traffic condition remains an open issue. A comprehensive study of the performance of typical headway distribution models on urban freeways is presented. With the advanced loop event data analyzer system, many accurate headway observations were obtained from I-5 in the area of Seattle, Washington. These headway data were used to calibrate and examine the performance of various headway models. The goodness of fit for several most commonly used headway distribution models was investigated by using headways observed on regular lanes and high-occupancy-vehicle (HOV) lanes from different time periods of day. To evaluate the performance of these headway models, the analytical Kolmogorov-Smirnov test statistic and visualized comparison curves were used to measure and reflect their overall goodness of fit to the collected headway data. Although each model has its own practicability to a certain extent, the test results showed that the double-displaced negative exponential distribution model provided the best fit to these urban freeway headway data, especially for HOV lanes at wideranging flow levels. The shifted lognormal distribution also fits the general purpose lane headways very well. As a byproduct, a new standard parameter estimation method was developed for calibrating complex multiparameter headway models.

A Multinomial Logit Model-Bayesian Network Hybrid Approach for Driver Injury Severity Analyses in Rear-End Crashes

Rear-end crash is one of the most common types of traffic crashes in the U.S. A good understanding of its characteristics and contributing factors is of practical importance. Previously, both multinomial Logit models and Bayesian network methods have been used in crash modeling and analysis, respectively, although each of them has its own application restrictions and limitations. In this study, a hybrid approach is developed to combine multinomial logit models and Bayesian network methods for comprehensively analyzing driver injury severities in rear-end crashes based on state-wide crash data collected in New Mexico from 2010 to 2011. A multinomial logit model is developed to investigate and identify significant contributing factors for rear-end crash driver injury severities classified into three categories: no injury, injury, and fatality. Then, the identified significant factors are utilized to establish a Bayesian network to explicitly formulate statistical associations between injury severity outcomes and explanatory attributes, including driver behavior, demographic features, vehicle factors, geometric and environmental characteristics, etc. The test results demonstrate that the proposed hybrid approach performs reasonably well. The Bayesian network reference analyses indicate that the factors including truck-involvement, inferior lighting conditions, windy weather conditions, the number of vehicles involved, etc. could significantly increase driver injury severities in rear-end crashes. The developed methodology and estimation results provide insights for developing effective countermeasures to reduce rear-end crash injury severities and improve traffic system safety performance.

A Feedback-Based Dynamic Tolling Algorithm for High-Occupancy Toll Lane Operations

Dramatically increasing travel demands and insufficient traffic facilities have induced severe traffic congestion. High-occupancy toll (HOT) lane operation has been proposed as one of the most applicable and acceptable countermeasures against freeway congestion. With balanced pricing and vehicle occupancy constraints, HOT lane operations can realize the optimal traffic allocation and enhance overall infrastructure efficiency. However, few previous studies have concentrated on optimal tolling strategies. Two major problems with inferior tolling strategies degrade HOT lane system performance. First, an undersensitive tolling algorithm is incapable of handling the hysteretic properties of traffic systems and may cause severe response delays. Second, oversensitive characteristics of imperfect tolling strategies may cause unfavorable flow fluctuations in HOT and general-purpose lanes that disrupt traffic operations. A new feedback-based tolling algorithm to optimize HOT lane operations addresses these problems. To decompose the calculation complexity, a second-order control scheme is used in this algorithm. On the basis of traffic speed conditions and toll changing patterns, the optimal flow ratio for HOT lane use is calculated by using feedback control theory. The appropriate toll rate is then estimated backward by using the discrete route choice model. This algorithm is simple, effective, and easy to implement. VISSIM-based simulation tests were conducted to examine its practicality and effectiveness. Test results show that the proposed tolling algorithm performed reasonably well in optimizing overall traffic operations of the HOT lane system under various traffic conditions.

Artificial Neural Network Method for Length-Based Vehicle Classification Using Single-Loop Outputs

Classified vehicle volumes are important inputs for traffic operation, pavement design, and transportation planning. However, such data are not available from single-loop detectors, the most widely deployed type of traffic sensor in the existing roadway infrastructure. Several attempts have been made to extract classified vehicle volume data from single-loop measurements in recent years. These studies used estimated speed for length calculation and classified vehicles into bins based on the calculated vehicle lengths. However, because of the stochastic features of traffic flow, deterministic mathematical equations based on certain assumptions for speed calculation typically do not work well for all situations and may result in significant speed estimation errors under certain traffic conditions. Such errors accumulate when estimated speeds are used in vehicle-length calculations and degrade the accuracy of vehicle classification. To solve this problem, an artificial neural network method was developed to ...

CHARACTERIZING AND MODELING OBSERVED LANE-CHANGING BEHAVIOR: LANE-VEHICLE-BASED MICROSCOPIC SIMULATION ON URBAN STREET NETWORK

Key findings are discussed regarding characteristics of lane-changing behavior based on observations of an urban street network. An in-depth exploration of observed lane-changing behavior and its modeling were conducted using vehicle trajectory data extracted from video observations using VEVID, a software package developed by the authors, integrated with a video-capture system. As a result, rules for modeling lane-changing behavior are proposed with respect to various types of lane changes. A lane-changing model consists of three components: a decision model, a condition model, and a maneuver model. Drivers’ decisions to change lanes depend on travel maneuver plans, the current lane type (i.e., the relationship between the current lane and the driver’s planned route), and traffic conditions in the current and adjacent lanes. A lane-changing condition model is the description of acceptable conditions for different types of lane changes. A lane-changing maneuver model describes a vehicle’s speed and duration when a certain type of lane change occurs. All of these models are established in a heuristic structure.

Observation-Based Study of Intersection Dilemma Zone Natures

Yellow phase dilemma zone is dynamically distributed at high-speed signalized intersections because of varying driving behaviors in response to yellow indications. This article presents an observation-based study of the natures of dynamic dilemma zones. A case study was conducted at a high-speed intersection in Fairfield, Ohio. Time-based yellow-onset trajectories were obtained using the video-capture-based technique and then they were used to analyze three types of dynamic dilemma zone models: that is, Type I and II dilemma zones, and option zone. The results reveal that the contributing factors of Type I dilemma zone and option zone are not constant but dynamic at different speeds. Vehicle type has been proved a factor that significantly affects drivers’ stopping behavior. Drivers of trucks are more likely to make pass decisions than drivers of passenger vehicles. Cars, sport utility vehicles, vans, and light trucks have similar downstream boundaries of Type II dilemma zone whereas heavy trucks have the furthest upstream boundary of Type II dilemma zone. Finally, the comparison between the option zone and the Type II dilemma zone is analyzed using the sample data.

Quantifying Dynamic Factors Contributing to Dilemma Zone at High-Speed Signalized Intersections

The issue of the dynamic yellow light dilemma zone (DZ) has been raised by researchers for many years. However, quantitative study of the inherent factors contributing to the dynamic DZ remains an issue, perhaps because of the lack of effective means for collecting the trajectory data. This paper presents an analysis of the dynamic characteristics of major contributing factors for Type I DZ and the option zone on the basis of vehicle trajectory data during yellow intervals. The qualified trajectory data of 1,445 vehicles were extracted from 46-h high-resolution videos shot at four high-speed signalized intersections in Ohio with the use of the cost-effective software VEVID, developed and upgraded by the first two authors. The statistical analysis of the obtained trajectory data quantitatively revealed the dynamic nature of major DZ contributing factors. Results indicated that the minimum perception–reaction time of drivers was greatly influenced by speed and could be modeled as a function of the speed. The maximum deceleration rate for stopping and the maximum acceleration rate for running a yellow light were greatly dependent on speed and the 85th percentile speed of the intersection approach. The rates could be expressed as a function of those two variables. On the basis of the new findings, the traditional Type I DZ model was greatly modified and improved. The new model provides a theoretical base for updating the existing DZ tables with the identified dynamic characteristics of the contributing factors.

Spatial distribution and characteristics of accident crashes at work zones of interstate freeways in Ohio

In this paper, we identify the spatial distribution and characteristics of fatal and injury crashes at different work zone locations on interstate freeways in Ohio within the last three (3) years. Work zone accident data obtained from the Ohio Department of Public Safety and Ohio Department of Transportation from year 2001 through 2003 has provided necessary information for the analysis. The study provides an insight on fatal and injury crashes at different locations within work zones. Concentrating only on fatal and injury crashes allowed for evaluation of the predominant factors that are responsible for these types of crashes. The results indicate that there is a significant difference in the proportion of fatal and injury crashes as well as rear-end crashes at different locations within a work zone. The activity area comes out to be the most dangerous area in a work zone and safety efforts must be concentrated on this area to improve the work zone safety. Predominance of rear-end crashes at all locations within a work zone indicates a high speed variance. Any countermeasures to reduce the speed variance will definitely improve the work zone safety

Observation-Based Lane-Vehicle Assignment Hierarchy: Microscopic Simulation on Urban Street Network

A lane-assignment model in a vehicle-based microscopic simulation system describes a vehicle’s position during its journey on an urban street network. In other words, it is used to estimate an individual vehicle’s location, speed, routing plan, lane-choice plan, lane-changing plan, and car-following plan from its entrance to a street network until the end of the trip. From the authors’ observations and study of lanechoice and lane-changing behavior, it is concluded that a vehicle is assigned to a lane in a logical manner depending on the relationship between its route-planned motivation and traffic conditions in the current lane and other lanes. A lane-assignment model consists of three components: lane choice, car following, and lane changing. The lane-changing component is composed of three submodels—a decision model, a lane-changing condition model, and a lane-changing maneuver model. Rules are discussed for lane-choice and lane-changing modeling based on videotaped observations over four-lane urban streets. Then a heuristic structure of a lane-vehicle-assignment model is proposed, which exposes the inherent relationship between vehicle-based travel behavior and lane-vehicle assignment on an urban street network. With the addition of a lane-assignment model derived from observed data, a simulation may be developed to correctly represent travel behavior and dynamic traffic assignment at the lane level and provide a more effective tool for design and evaluation of the performance of strategies for traffic control, traveler information, and congestion alleviation.