International Conference on Transportation and Development 2020: Traffic and Bike/Pedestrian Operations

Bike share systems are commonly found in large cities thus research has focused on large cities. Little research has focused on bike share riders in small to mid-size cities. This study quantifies the factors affecting bike sharing and develops a direct ridership model (DRM) to predict monthly station-level bike share ridership in small to mid-size cities. A DRM is used to estimate how station-level bike sharing ridership correlates with demographics and land-use near bike share stations. The model is estimated using data from small cities and finds that bike share ridership is influenced the most by nearby attractions, such as parks. The results suggest that bike share riders in small cities ride for recreational purposes, rather than avoiding congestion along their commute. The developed DRM is better suited to predict ridership and to identify ideal locations for bike share system expansions in small to mid-size cities.

This research examines sidewalk availability, width, and socio-demographic/socio-economic disparities from cities with differing municipal sidewalk maintenance policies. Sidewalk maintenance policy tends to place the onus on either the city or the abutting landowners. Existing sidewalk studies generally focus on smaller cities or limit analyses to micro-scale sample areas. Utilizing geospatial data and software, this paper evaluates citywide sidewalk availability and width across two cities that take responsibility for sidewalk maintenance (Austin, TX, and Raleigh, NC) and two where the onus is on the abutting landowner (Denver, CO, and Portland, OR). The results suggest an overall lack of sidewalk availability in all cities, but there were no substantial differences in availability based on municipal sidewalk policy. All cities, regardless of policy, show an unexpected relationship between wider sidewalks and lower incomes. Regarding socio-economic disparities, lower income block groups were associated with greater sidewalk availability in the cities that take responsibility and less sidewalk availability in the landowner-maintained cities.

A study of pedestrians crossing behavior is conducted at an uncontrolled mid-block crosswalk in Istanbul, Turkey, to model the pedestrians waiting time, related to their behavior for making the crossing decision. This article focused on the issues encountered in the modeling of the operational behavior of pedestrians. The discrete choice framework is used because of its capacity to deal with individuals’ choice behavior. Pedestrians waiting time is classified into three levels, including low, medium, and high levels based on the level of service of pedestrians waiting time. The pedestrians’ behavior prediction has been improved by analyzing, taking into account three levels for pedestrian behavior.

This research aims to explore the areas where transit service can be improved through analyzing the use of dockless bikesharing systems. It initially introduces the overall characteristics of bike use in Beijing and finds that weekday cycling shows a certain regularity and travel distributions of the hot riding area at different periods are revealed to have little difference. Next, each trip purpose is investigated and categorized by using density-based spatial clustering of applications with noise (DBSCAN) analysis to divide bikesharing origins and destinations into five types based on different built environment features. Results show that most riders prefer to transfer to other public transit systems. On the basis of distinct transit travel mode choice data obtained from mobile navigation software, including expected travel time, walking distance, and travel cost, an evaluation index model is formulated to measure and evaluate transit performance for network optimization. The proposed framework can help local transit authorities make targeted countermeasures that improve service quality and attract more passengers.

Vulnerable road users (VRUs) such as pedestrians and bicyclists, also known as non-motorists, are vulnerable due to lack of protection in traffic and usually sustain the higher injury levels. The objective of this study was to determine factors that contribute significantly to the crash severity of intersection-related crashes involving motor vehicles and VRUs. A three-year crash data, 2013–2015, from the Ohio Department of Public Safety was utilized for this study. The logistic stepwise selection procedure was applied to estimate statistically significant predictor variables that contribute in increasing VRU-related crash severity levels. The model identified five statistically significant predictor variables out of fourteen independent variables tested in this study. The predictors that increase crash severity of VRU-related crashes are: if a VRU is a pedestrian, if the intersection is on a curve and/or grade location, if a VRU is a female, if the crash occurs at nighttime, and if the unit in error is the VRU.

In the past 2 years dockless electric scooters, shared bikes, and electric skateboards have risen in popularity and are being integrated into the urban transportation ecosystem. As these micro-mobility devices increase in prevalence, several cities have raised concerns about their safe operating speeds and interaction with pedestrians and vehicles. This study collected travel speeds for pedestrians, bicycles, electric scooters (e-scooters), skateboards, and electric skateboards (e-skateboards). A total of 754 travel speeds were recorded using a laser speed gun setup on bike paths, sidewalks, and streets. Results show that around 95% of e-scooter and 93% of bicycle speeds recorded were below 15 mi per hour (mph). Interestingly, more than 50% of e-skateboard speeds were above 15 mph, with 4% above 20 mph. Statistical tests were also performed to compare the mean speeds of all the modes. There was statistically significant difference in speeds between all of the above mentioned transportation modes, with the exception of bicycle and scooter speeds.

The issue relating to pedestrian with motor vehicles crashes has received more attention in recent years. In this paper the two-level random intercept model for pedestrian crashes severity prediction that accounts for the interdependent characteristics was introduced. A thought is proposed in this paper by using Bayesian probability inference to calculate the parameters rather than the maximum likelihood method, setting parameter distribution for each parameter, using Monte Carlo Markov (MCMC) algorithm to generate model parameter distributions. To demonstrate this approach, pedestrian crashes data from Colorado is divided into spatial categories: urban road intersection, urban road section, and driveway access. The result shows that the prediction accuracy of Bayesian probability inference is better than the traditional ones of urban road intersection and urban road section. The results prove that causes of pedestrian death or injury crashes at intersections are: pedestrian age, pedestrian direction, vehicle type, and driver speed. Significant factors causing pedestrian death or injury on urban road sections including: road condition, lighting condition, belt use, and driver speed. The analysis results reflect the objective feasibility of the model and the method for model parameter estimation optimization. It is expected that the two-level intercept model and Bayesian probabilistic inference model optimization method can help more researchers, transportation officials, and urban city community planners to make more efficient treatments to proactively improve pedestrian safety.

This study investigates the roles of nonmotorized travel, telecommuting, and the built environment in health. Multilevel structural equation models have been developed to examine the complex interrelationships between county-level nonmotorized travel and telecommuting mode shares, county-level health status indicators, as well as built and social environment characteristics at various spatial levels. The results indicate that higher levels of nonmotorized travel within a county are linked with improved health outcomes for residents, whereas higher rates of telecommuting are related to adverse health effects. The results further show that county-level health outcomes are linked with built environment attributes including the extent of compactness and mixed-use development, pedestrian-friendliness of the street network, distance to transit, access to healthcare and recreational facilities, and prevalence of fast food restaurants. Identifying the key factors that influence the health profiles of communities, these findings can assist in developing more effective transportation planning policies that foster sustainable travel and promote healthier communities.

Pedestrian fatalities in the U.S. increased in 2018 to its largest death toll since 1990. Approximately 72% of those fatalities happened at mid-block locations. A driving simulation experiment was carried out to analyze driver-pedestrian conflicts at an uncontrolled mid-block location of a multi-lane urban arterial. The effect on driver performance of lighting conditions, the pedestrian crossing direction, and the use of a “PEDESTRIAN CROSSING AHEAD” overhead warning sign with flashing beacons was evaluated. Vehicle positions and speeds were analyzed for those drivers who faced the potential conflict with the simulated pedestrian. Driver-pedestrian conflicts occurred in 77% of the simulation runs with the existing signage on the simulated road, whereas conflicts occurred in only 23% of the runs with the proposed warning sign configuration. The reduction of 70% in the number of driver-pedestrian conflicts reflects the positive effect of the proposed overhead warning sign configuration in enhancing safety at the mid-block crossing.

This paper presents the results of a research effort in recognizing fatality patterns in large truck-involved pedestrian/bike crashes in the state of Florida. Using 2007–2016 crash data, machine learning models, including random forests and decision tree were applied. Model results showed that fatality patterns and contributing factors tended to be different by traffic volume and road type classification. For local roads with an AADT higher than 38,000, proximity to intersection and vehicle speed were found as decisive parameters in increasing crash injury severity. On high volume roads not classified as local, crashes that involved young/middle-aged truck drivers happened on divided roadways with a speed limit of 50–75 mph, and during the midday period had a high probability of a fatal outcome. For low volume roads, under clear weather conditions, a combination of median barrier and curb shoulder close to a signalized intersection were associated with high fatalities.

Time saving and cost saving are two of the most discussed and perceived benefits of cycling over the world. Bicycle is an emerging solution for commuters in Dhaka city which is notorious for traffic congestion. Based on a stated preference survey on commuters (N = 300), this paper studies the alternate modes for cyclist. Calculating monthly time and cost saving, this paper attempts to develop a logit model and find correlations of these savings with income, trip distance, and alternate mode choice. According to analysis, cost-saving is usually high for the higher income group. As an alternative to bicycle, public transit was the most preferred mode. Being the first-ever base study in Bangladesh on commuting cyclists’ this study develops a comprehensive and flexible model for making decisions and advocating in favor of promoting a sustainable transportation system.

Quantitative evaluation of urban walking traffic is an important way to optimize the layout of daily life facilities, improve the convenience of residents’ walking, ease congestion and pollution problems, and improve urban living environment. By constructing the gravity model which uses the number of different weight POIs (Building points that attract people to travel) that the urban transportation network can access within a certain distance from each building in the Yanta area, the corresponding accessibility evaluation index can be calculated and used to evaluate walking traffic. The results show that within the acceptable range of travel, the area in Yanta where the walking gravity index above 42.1 accounted for 73.2%. It shows that the overall walking accessibility of Yanta District is higher, but there is a difference in the distribution of east and west; the area in Yanta where the walking gravity index is above 24.4 accounted for 88.6%. And the distribution is even.

Logistics systems operations, such as intermodal container shipping, are subject to the use of transportation corridors with inherent factors related to the safe, reliable, and efficient movement of freight. Risk analysis and resource allocation of freight operations must consider a diverse set of corridor attributes, system goals, and impacts of future and emergent conditions on operational priorities. A corridor trace analysis method is developed to explore the impact of inland waterway corridor factors on system performance. Transportation corridors are segmented into system-appropriate intervals. Each segment is evaluated using a number of attributes and performance metrics. Segments with potential to pose greatest disruption to operations are identified through a multifactor analysis of dynamic attributes that vary along the length of the corridor and across temporal domains. The methodology is demonstrated for maritime corridors necessary for the movement of containers through harbor terminals to inland terminals for a marine container port in the United States. The approach is of interest to enterprises utilizing transportation networks for the movement of freight, equipment, or personnel.

Short-term traffic flow prediction, which is useful to improve traffic congestion and road efficiency, has been a hot issue in the field of transportation. However, only considering Euclidean space, conventional methods are always unable to make good use of the spatial-temporal correlation of traffic flow data which is usually a topological structure. In this paper, a deep learning model, GCN-LSTM (graph convolutional network-LSTM), was proposed with encoder and decoder structure. GCN-LSTM will simultaneously capture the spatial and temporal characteristic of traffic flow by embedding GCN into the structure of LSTM. Training with the traffic flow data of previous T moments and adjacent section, GCN-LSTM effectively perform short-term traffic flow prediction. Experiments on real data demonstrate that our method, considering both of spatial and temporal features, has a more powerful representation ability and higher prediction accuracy compared with LSTM.

Different modeling and analysis methods describe different characteristics of the traffic network and lead to different conclusions for the identification of important road sections and intersections. In this paper, the road network and bus network of Langfang city are integrated to model the composite traffic network, and the actual traffic flow data are added to analyze. A bioinformatics approach called SPRING is introduced to determine the importance of certain nodes. This method gives the ranking result of node importance with a combination of various characteristics, which include the similarity of network characteristics and traffic conditions between certain nodes and others. Through comparing the node importance of multi-feature fusion and the node importance of single traditional network characteristic, it is concluded that multi-feature fusion method has the feasibility, advantages, and disadvantages in identification of important nodes and shows effectiveness when considering the real traffic situation and the demand of traffic analysis.

With the rapid development of urban traffic, the problem of urban traffic congestion is becoming more and more serious. Solving the traffic congestion of the main arterials of urban area is of great significance to alleviate the urban traffic congestion. Arterial coordination optimization is an effective method to improve arterial capacity, service level, and alleviate the urban traffic congestion. In this paper, an arterial coordination optimization model has been proposed with the analysis and processing of the vehicle trajectory. First, the characteristic trajectories are extracted according to the velocity distribution of the arterial line. Then, the intersection delay model is established to optimize the characteristic trajectories to obtain the phase difference for each intersection. The simulation result by SUMO software shows our method has 45.48% improvement in parking delay and 14.08% improvement in the average queue length at each intersection, which proves effectiveness of the proposed method.

Hard shoulder running (HSR) is increasingly used in recent years as a more sustainable way to alleviate congestion and increase the efficiency of freeways in many vibrant cities. With HSR, the existing paved shoulder is opened for use as a travel lane when congestion exceeds predefined thresholds. This study proposes a hybrid hard shoulder running operation system (H-HSROS) that can be used by transportation agencies to operate their HSR operations in a safe and efficient manner. The proposed H-HSROS adopts a fuzzy approach based on archived speed data from roadway sensors. Interstate 75 (I-75), located in Cincinnati, is used as a testbed to illustrate the operation of the proposed H-HSROS. There are three major models in the proposed H-HSROS, which are a data manager, a computational engine, and a transportation management center. The data from sensors on I-75 in Cincinnati, Ohio, showed an increasing trend in traffic volume during the weekday compared to a weekend with an increase in traffic congestion. The congestion on the corridor could be improved by implementing the HSR technique. The data manager of H-HSROS processes the information coming from the roadway sensors and cameras. The computational engine works with data and provides preliminary traffic analysis such as volume, flow, mean, standard deviation, and reliability. Real-time traffic information of H-HSROS is analyzed using a hyperbolic fuzzy membership function. Fuzzy membership function uses the computation engine to analyze the coming data provided and determine whether to open the HSR lane for travel or not. Before the opening of HSR to travel, the shoulder must be inspected continuously. The inspection should be through a closed-circuit camera, which must be distributed along with the facility for safety issues. For rush hour congestion reduction, the shoulder might be opened every weekday during peak hours. Also, if the traffic volume is high, the H-HSROS will be able to extend the operating hours. HSR provides additional capacity on the freeway and can potentially mitigate the negative effects of congestion.

Connected vehicles (CVs) navigate through multiple routing to select the quickest path. In non-possible multi routing at freeway sections, CVs can make a car platoon to increase the gap between non-CVs. In this study, different traffic conditions, volume levels with four different penetration of CVs have been simulated in PTV Vissim on I-26 in South Carolina. A statistical t-test has been conducted to evaluate the impact of CVs on travel time in different traffic conditions. The results show CVs reduce travel time until 75% penetration when there is no accident on a freeway. Although, with an accident on freeway, the increase in CVs reduces the average speed unless all the vehicles are connected.

The diverted traffic from the freeways to alternative arterials due to incidents deteriorates the intersection movement performance causing delays, long queues, spillbacks, spillovers, and so on. At the critical intersections, the diversion most likely affects only specific movement(s) in the direction of the diverted traffic rather than the whole intersection. This study investigates the use of data from connected vehicle (CV) technology combined with the high-resolution controller (HRC) to automatically identify traffic congestion patterns at the signalized intersections due to diversion. Four different clustering techniques: K-means, K-means with principal component analysis (PCA), K-means with t-distributed stochastic neighboring embedding (t-SNE), and deep embedded clustering (DEC) are investigated for use to identify the patterns, and their performances in identifying the congestion patterns are evaluated. The results indicate the ability of the select method to identify distinctive patterns of congestions that agencies can use in implementing signal timing plans to better accommodate the diverted traffic without significantly deteriorating the performance of other movements at the intersection.

Parking problems have hindered the smooth traffic in cities for a long time. In this paper, we propose reservation-based private parking slots allocation mechanisms from the e-parking platform’s perspective in which proximity and parking cost are considered as demand preferences. Two allocation mechanisms are proposed. The advanced FBFS (first-book-first-serve) provides immediate feedback to each user while the PBA (preference-based-allocation) is an overall optimization mechanism which collects demands in a certain period and allocates at one time at a decision point. A binary integer linear programming model aiming at maximizing platform’s profits is proposed for PBA. Results show that PBA guarantees the highest profits when the acceptance rates of the original FBFS, the advanced FBFS and PBA are in mere difference. In conclusion, PBA demonstrates unparalleled superiority while the advanced FBFS outstripped the original FBFS and the proposed mechanisms make good use of idle private parking resources.

Express lanes are a type of managed lane strategy that use congestion pricing by varying the toll amount to maintain a performance threshold and provide a more reliable travel time. The aim of this study was to develop an analysis methodology to assess the complex or multiple weave segments formed in express lane projects due to the placement of access points. Various scenarios were tested using the Highway Capacity Software (HCS) and VISSIM microsimulation to estimate the capacity of such complex weave segments. The study recommended reporting results from HCS and volume-to-capacity (V/C) ratio obtained from microsimulation to estimate the adequate capacity of complex weave segments. This study also found that there is value in dividing the critical segment into merge, basic freeway, and diverge sections for analysis. The speed differential between weaving and non-weaving volumes was found to be minimal, if adequate length was provided and no congestion was observed.

Trucker fatigue has become a national epidemic. Unavailability of sufficient parking spaces during various periods at rest areas results in illegal and unsafe parking at on/off-ramps and other unauthorized areas, which may lead to traffic safety hazards. To provide better trucking parking information for truck drivers and improve freight operations safety, eight state members of the Mid-America Association of State Transportation Officials (MAASTO) have joined together and developed a real-time, multi-state truck parking information management system (TPIMS). TPIMS has been fully operational from January 4th, 2019, and is providing real-time parking availability information to truck drivers and dispatchers and traffic management centers (TMCs). A centralized TPIMS data archival system (TPIMS-DAS) is developed to archive real-time TPIMS data from all the eight participating MAASTO States in order to evaluate and track TPIMS performance. This paper presents the system architecture of TPIMS-DAS, system performance measures, and the next phase applications. By July 2019, TPIMS-DAS has archived more than 23 million records about parking availability and site information for 153 parking sites across eight states. Additional TPIMS-DAS functionalities, such as automated performance metric calculation, the web interface for data access, and visualization, are also under development. Data and performance metrics provided by TPIMS-DAS are expected to provide insights for states to improve and coordinate parking facility improvements and improve corridor traffic safety.

In this paper, a methodology framework is developed on systematic traffic count balancing for regional road networks. The framework includes three major components: two intercept link extractors, and one traffic count balancing model. The intercept link extractors are developed based on breadth search algorithm (BFS) designed to traverse networks to extract intercept links. The traffic count balancing model is a constrained weighted least square optimization model with constraints of traffic conservation equations based on intercept links. The proposed framework has the following features: (1) it can automatically traverse networks to identify and extract the intercept link counts for traffic conservation; (2) it can balance multiperiod traffic counts in regional level; (3) uncertainty of traffic counts can be considered by weight factors; and (4) the traffic balancing model can be solved efficiently due to its convex structure.

Traffic signals are typically retimed to accommodate traffic conditions during a period of observation. A set of robust measures are needed to determine the effectiveness of signal improvement technique. The present study utilizes five such measures developed in previous research: median travel rate, within-day variability of travel rate, median travel rate dispersion, and overall travel rate variability to evaluate the effectiveness of signal improvement. Cumulative distribution function plots are constructed initially using travel rate, which is used to identify the atypical days. The atypical days, representation of variable travel rate reduced for 67% of the segments suggesting an improvement of travel time variability. These atypical days are removed, followed by an evaluation to assess the change of travel rate and its variability. The within-day variability was the only parameter that improved significantly. A statistical measure on weekly measures stated that there were few segments with actual change in before-after condition.

Connected vehicles (CV) that communicate with traffic signals and notify the driver of expected signal status changes have emerged in the last few years. For fixed-time operations it is a deterministic exercise to predict the signal state for a given movement. However, under actuated-coordinated operation, there are stochastic variations of phase start and end times. When traffic engineers enable additional traffic responsive logic to accommodate for changes in demand, the phase timings become more challenging to predict. This paper proposes a methodology to evaluate traffic signal prediction algorithms for vehicle-to-infrastructure CV application. Data was gathered using video footage on minor and major movement of an intersection that operated in actuated-coordination. Data was collected for 176 cycles over two days. Results showed that the CV application can predict the mean start of green within +/−2.7s. The paper concludes by recommending the evaluation methodology and graphical summaries proposed be used as tools for traffic engineers and vehicle manufactures to characterize the stochastic nature of actuated traffic signals and manage expectations of motorists and other stakeholders.

The main objective of this research was to estimate the maximum queue lengths generated by two new designs, including synchronized and Milwaukee B interchanges, in comparison to four existing designs. The methodology consisted of three main parts: (1) building a simulation model calibrated to the geometric features of 30 existing service interchanges; (2) running over 1000 simulation tests using VISSIM and Synchro; and (3) conducting two-way analyses of variance (ANOVAs) on simulation outcomes. Queue lengths were measured for many combinations of traffic volume, turning traffic percentage, traffic distribution, and traffic composition. Based on the results, the Milwaukee B design had the shortest mean queue lengths among all the interchanges considered in this study. The Milwaukee B had queue lengths averaging about 68% shorter than a parclo B interchange, which was the design with the shortest mean queue length among conventional designs. The synchronized and the DDI had similar mean queue lengths.

One of the designs for future highways with mixed flow of connected automated vehicles (CAVs) and manual cars will be a dedicated lane for the CAVs to form platoons and travel with higher speeds and lower headways. Connectivity between vehicles and the infrastructure will enable control strategies to become more advanced. The connectivity will enable the formation of platoons of CAVs which will travel beside non-platoon lanes. Every connected vehicle joining the highway from an onramp will have the challenge of when and where—while traveling on the acceleration lane—to join the adjacent mainline of the highway. In this research, the aim is developing a model for investigating the effects of different merge actions yielding a more efficient and systematic method for the vehicles joining the mainlines comprised of platoon and non-platoon lanes. The actions tested for assisting the merge process include deceleration in the mainlines, joining a platoon, etc. The model targets to direct every CAV entering a multi-lane highway segment from an on-ramp, to the rightmost lane of the highway based on the appropriate action which is selected according to the traffic conditions on the mainline. To account for car following behavior, the platoon lanes are assumed to have a simplified CACC (cooperative adaptive cruise control) and the non-platoon lanes the IDM+ car-following model. Numerous simulation experiments proved the efficiency of the model in increasing the throughput in the merge area. The results of this research will help in the design and development of advances systems for controlling on-ramp merge sections in the future with CAVs.

This study incorporates drivers’ route diversion behavior (DRDB) into a dynamic origin destination (OD) demand estimation and prediction (DODE) model and establishes a group route guidance model based on an integrated demand–diversion prediction model. Test results of the case study show that the integrated demand–diversion prediction model can consider time-varying OD demand, time-varying traffic characteristics, and dynamic DRDB under information provision. The maximum deviation between the real OD volumes and DODE model under information provision is approximately 11.46%, MAPE is 4.53%, and normalized RMSE is 5.29%. Under the scenarios with fixed time-dependent OD demand and fixed compliance rates, the effectiveness of group route guidance is significantly reduced compared with that under the scenario with real time-dependent OD demand and real compliance rate. The proposed model can accurately estimate and predict the possible DRDB and the effects of traffic information on OD demand prediction by using real-time traffic detected data. Moreover, the proposed model can enhance the accuracy of OD demand and traffic state prediction under information provision, thereby increasing the effectiveness of the proposed network route guidance strategies.