AM Peak Period Research Articles

Traffic Congestion and Safety: Mixed Effects on Total and Fatal Crashes

This paper examines the effects of traffic congestion on total crashes, fatal or serious injury (FSI) crashes, and fatal-only crashes in peak periods using a zone-level safety analysis in Greater Melbourne. Bayesian mixed-effect negative binomial models are employed to investigate the relationship between a congestion index and the frequency of total and FSI crashes. In addition, Bayesian mixed-effect binary logistic models are adopted to explore the association between the congestion index and the likelihood of having fatal crashes in Statistical Area Level 2 (SA2) zones. Modelling results indicate that traffic congestion tends to increase total crashes in both the AM and PM peak periods and FSI crashes in the AM peak period. In contrast, traffic congestion tends to decrease the likelihood of having fatal crashes at both the AM and PM peaks. These findings suggest that many policies to reduce traffic congestion may also enhance road safety by lowering the overall number of crashes. However, it is crucial to incorporate careful speed management within these policies to reduce the risk of fatal crashes effectively.

Needs-gap analysis of street space allocation

Streets have long been designed to maximize motor vehicle throughput, ignoring other street users. Many cities are now reversing this trend and implementing policies to design more equitable streets. However, few existing tools and metrics enable widescale assessment, evaluation, and longitudinal tracking of these street space rebalancing efforts, i.e., assessing how equitable the current street design is, how it can be improved, and how much progress has been made. This paper develops a needs-gap methodology for assessing the discrepancy between transportation supply and demand in urban streets using existing datasets and automated methods. The share of street space allocated to different street users is measured in 11 boroughs of Montréal, Canada. Travel survey data is used to estimate the observed and potential travel demand in each borough in the AM peak period. A needs-gap analysis is then carried out. It is found that bus riders and cyclists face the greatest needs-gap across the study area, especially in central boroughs. The needs-gap also increases if only trips produced or attracted by a borough are considered. This shows the potential of applying an equity-based framework to the automated assessment of street space allocation in cities using large datasets.

Traffic and Welfare Impacts of Credit-Based Congestion Pricing Applications: An Austin Case Study

This study seeks smart credit-based congestion pricing (CBCP) solutions for maximally improving travelers’ welfare by varying toll levels and locations across the Austin, Texas network. Scenarios evaluated include selecting links with maximum delays by variably tolling bridges and by recognizing congestion externalities across all links. Travel demand models deliver inputs for normalized logsum differences to quantify and compare consumer surplus changes across traveler types, around the region. Results suggest limited tolling locations under four broad times of day can do more harm than good, unless travelers shift out of the PM and AM peak periods or revenues are returned to travelers as credits. When using CBCP across all congested links at congested times of day (with 10% of revenues retained to cover system administrative costs), an average net benefit of $1.61 per licensed driver per weekday is estimated, with almost all travelers benefiting. For example, 95% of the traffic analysis zones’ lowest value of travel time (VOTT) group (VOTT1 = $5/hour) are expected to benefit from the CBCP policy. Tolling at twice the difference between marginal social cost and average travel cost (on each subset of congested links) appears to benefit more people, although tolling high on various links adds to congestion elsewhere. For example: tolling Austin’s highest-delay-producing or “top 500” links is estimated to benefit 98.5% of the zones’ highest VOTT (VOTT5 = $45/hour) travelers, while raising vehicle-miles traveled by just 0.8% (as a result of more circuitous, congestion- and toll-avoiding travel).

Replication of Daily and Monthly Freeway Demand Variations for Travel Time Reliability Procedures

The sixth edition of the Highway Capacity Manual (HCM) incorporates a travel time reliability assessment procedure for freeways and urban streets. Several demand adjustment factors, referred to by demand multipliers, are used to capture traffic demand variation across different days and months. These factors are currently produced by referencing the average daily traffic volume of each day-month combination to a base daily volume. However, practitioners usually perform traffic analyses during specific times of the day, for example, peak periods, off-peak periods, or even peak hours, demand multipliers may therefore replicate demand variation more accurately if they are based on traffic volumes concurred in time intervals narrower than a day. This paper investigates six criteria or periods to derive demand multipliers: full-day, pre AM-peak, AM peak-period, midday, PM peak-period, and post PM-peak. The study explores how these periods affect the scale of demand multipliers and the travel time reliability assessment. It was found that the main statistics of demand multipliers, that is, the mean, range, and standard deviation, greatly differ across the different multiplying periods. If analyzing peak periods on oversaturated corridors, the adoption of daily-volume multipliers was found to significantly overestimate the mean travel time index and planning time index during both the AM and PM peak periods, the accuracy of the travel time reliability estimation was considerably influenced. The study concludes with major findings and recommendations for possible enhancements to the HCM travel time reliability procedure.

Temporal validation of a multimodal transit assignment model

With the increasing demand and range of urban mobility, public transport systems are playing an increasingly important role in providing people with efficient and affordable access to education, employment, markets, and other key services. Public transport planners can predict passenger loads and levels of service by using transit assignment models. Therefore, having a consistent transit assignment model has become an important issue. In addition, estimating the path choice factors, considering that network cost attributes might be non-deterministic, is complicated. In view of these, this paper adapted a transit assignment model, based on the Bayesian Model proposed in the authors’ recent study, to validate the results of the model for multimodal transit networks under uncertainty and random variations of path choice parameters. In that model, path choices are represented by a multinomial logit model, and its coefficients are estimated via a Markov Chain Monte Carlo (MCMC) method. In this paper, the path choice parameters of the model are first calibrated by the individual travel history data in the AM peak period, and then are used to predict the passenger flow on different paths in the PM peak period. Second, based on the percentage error between each posterior estimate and the actual observation from AFC data on the segment level, the path segments are categorized into four groups. Then, by defining the segment attributes, the strengths and weaknesses of the model are analysed for the four groups.

Reinforcement-Learning-Based Cooperative Adaptive Cruise Control of Buses in the Lincoln Tunnel Corridor with Time-Varying Topology

The exclusive bus lane (XBL) is one of the most popular bus transit systems in the U.S. The Lincoln Tunnel utilizes an XBL through the tunnel in the AM peak period. This paper proposes a novel data-driven cooperative adaptive cruise control (CACC) algorithm that aims to minimize a cost function for connected and autonomous buses along the XBL. Different from existing model-based CACC algorithms, the proposed approach employs the idea of reinforcement learning, which does not rely on accurate knowledge of bus dynamics. Considering a time-varying topology, where each autonomous vehicle can only receive information from preceding vehicles that are within its communication range, a distributed controller is learned real-time by online headway, velocity, and acceleration data collected from the system trajectories. The convergence of the proposed algorithm and the stability of the closed-loop system are rigorously analyzed. The effectiveness of the proposed approach is demonstrated using a well-calibrated Paramics microscopic traffic simulation model of the XBL corridor. The simulation results show that the travel time in the autonomous version of the XBL are close to the present day travel time even when the bus volume is increased by 30%.

Ranking statistically similar highway travel corridors using speed datasets within the USA

Commercially available probe vehicle speed data are becoming a reliable means to evaluate roadway performance within the United States of America (USA). These temporally defined data are used to identify roadway congestion along fixed-location, variable-length highway segments called traffic message channels (TMC). Current performance measures that define congestion as an increase in travel time are inherently skewed towards longer, arbitrarily defined segments. To overcome this limitation, this paper defined congestion as the relative travel time increase, a performance measure that directly accounts for the variation in TMC lengths. Next, concurrent TMC with statistically similar levels of congestion were grouped into corridors. The resulting corridors provide a more accurate assessment of regional congestion, simplify data collection and analysis and allow agencies to compare congestion across different highway segments. The proposed methodology was applied to 83 TMCs along Interstate 80 East in New Jersey, USA. The 83 TMCs had in excess of 30 million speed records in 2013 for the AM peak period that were grouped into over 300 000 15 min intervals. An analysis of variance of time intervals indicated that the 83 TMCs could be grouped into 34 corridors. The congestion along each corridor was then characterised and visualised.

The determinants of travel demand between rail stations: A direct transit demand model using multilevel analysis for the Washington D.C. Metrorail system

Transit demand models have become indispensable tools for transit planners and mangers in the 21st century. By quantifying the relationship between transit ridership, the cost of travel, the character of the built environment, and the socio-economic characteristics of riders, such models enable transit planners and managers to make more informed decisions regarding transit routes, levels of service, transit fares, transit oriented development (TOD) and other transit supply parameters. Direct ridership models (DRMs) are now able to address transit ridership at each station directly with higher sensitivity to built environmental characteristics in well-defined station areas. More recently Origin-Destination DRMs have begun to use data on ridership between each origin and destination pair to facilitate more precise estimation of transit demand by origin-destination pair.In this study, we developed a time-of-day Origin-Destination Direct Transit Demand Model (OD-DTDM) that uses fare-card data from the Washington DC Metrorail system, applying a multilevel (or hierarchical) model to address the statistical problem due to the presence of groups or clusters of observations. We examine the research questions: (1) what are the determinants of transit demand between the origin and destination stations in the DC Metrorail system by time of day? and (2) what are the magnitudes of impacts that land use factors, as well as factors of fares and travel time of other modes, have on transit demand vary by time of day? To address statistical complexities introduced by the fact that each station represents both an origin and a destination, we applied multilevel (or hierarchical) modeling techniques. Using these techniques, we found that the number of households and the number of jobs within a walkshed serve as trip generating and attracting factors, respectively, in the AM peak period, but with higher positive coefficients for jobs; these two factors reverse their roles in the PM peak period. Other variables with substantial effects on ridership include transit fares per mile, travel time between OD-stations by car and by bus, parking capacity, the level of feeder bus service, and train service levels. While these findings are not surprising, the time-of-day OD-DTDM provides more detailed information regarding the determinants of transit demand with temporal variation, and enables transit planners and managers to adopt policies and plans, such as transit oriented development, fare structure, and service levels, more fine-tuned for each origin and destination pair and by time of day.

Application of smart card data in validating a large-scale multi-modal transit assignment model

The accuracy of transit assignment plays an important role in the successful design and operation of a transit system. The majority of previous studies on validating transit assignment models has used limited survey data or has lacked a large-scale multimodal and high quality dataset. Considering the advantages of smart card [automatic fare collection (AFC)] systems, the aims of this study are to put forward a methodological framework to validate existing transit assignment models and to quantify the performance of these models. Our study combines data from three sources: the General Transit Feed Specification, an AFC system, and a strategic transport model from a large-scale multimodal public transport network, namely the South-East Queensland (SEQ) network in Australia. The AFC system in SEQ has provided a very large and highly accurate dataset on passenger boardings and alightings for the three transit modes of bus, rail and ferry. Following a data analysis, an origin–destination trip matrix is estimated for the AM peak period using AFC data as an input to the transit assignment model. Then, the results of the transit assignment model are compared with the actual passengers’ route choices over the same period, at different levels of aggregation. The model performance is quantified by each route (and direction), by each segment of each route (and direction), and by each stop. The results indicate that relatively tighter thresholds are required to validate the transit assignment at the segment level than at the stop level. Furthermore, the validation results indicate that the greatest error is realized for the bus mode, while the level of accuracy in the rail mode is the best. The results suggest a segment-level analysis should be used as the most useful level of aggregation for future calibration and validation of transit assignment models.

Large-scale application of MILATRAS: case study of the Toronto transit network

This paper documents the efforts to operationalize the conceptual framework of MIcrosimulation Learning-based Approach to TRansit Assignment (MILATRAS) and its component models of departure time and path choices. It presents a large-scale real-world application, namely the multi-modal transit network of Toronto which is operated by the Toronto Transit Commission (TTC). This large-scale network is represented by over 500 branches with more than 10,000 stops. About 332,000 passenger-agents are modelled to represent the demand for the TTC in the AM peak period. A learning-based departure time and path choice model was adopted using the concept of mental models for the modelling of the transit assignment problem. The choice model parameters were calibrated such that the entropy of the simulated route loads was optimized with reference to the observed route loads, and validated with individual choices. A Parallel Genetic Algorithm engine was used for the parameter calibration process. The modelled route loads, based on the calibrated parameters, greatly approximate the distribution underlying the observed loads. 75% of the exact sequence of transfer point choices were correctly predicted by the off-stop/on-stop choice mechanism. The model predictability of the exact sequence of route transfers was about 60%. In this application, transit passengers were assumed to plan their transit trip based on their experience with the transportation network; with no prior (or perfect) knowledge of service performance.

Short-term Forecasting of Travel Time and Reliability

In this paper, a traffic flow simulator (TFS) is presented to forecast the short-term travel time and reliability that can be served as a predicted travel time database for the Hong Kong transport information system or advanced traveler information system. The forecasted travel time information is useful for the pre-trip planning of the drivers. The drivers can decide their departure time and mode choice based on the forecasted travel times and reliabilities. A case study has been carried out in the Hong Kong Island during the AM peak period. A stochastic deviation coefficient of link flow is incorporated in the TFS so as to simulate the minute-by-minute fluctuation of traffic flows within the peak hour period. The purposes of the case study are: 1) to show the applicability of the TFS for larger-scale road network; and 2) to illustrate the short-term forecasting of travel time and reliability in practice. The results of the case study show that the TFS can be applied for real network effectively. The TFS forecasted travel times and the observed travel times are compared. The findings show that the observed travel times fell in the 90% confidence intervals of the TFS forecasted travel times. To distribute the traffic information to the drivers, a TIS website is currently being developed in The Hong Kong Polytechnic University for displaying the traffic information in Hong Kong.

A specialized equilibrium assignment algorithm for air quality modeling

A multiple user class equilibrium assignment algorithm is formulated to determine vehicle trips and the vehicle miles of travel (VMT) in various operating modes on highway links. A heuristic solution algorithm based on the Frank–Wolfe decomposition of the equilibrium assignment problem is presented. The treatment of intrazonal trips, which are very important for emission studies is also discussed. The solution algorithm is implemented in a traffic assignment program for emission studies, referred to as TAPES. The use of the algorithm is demonstrated through a TAPES model case study on a Charlotte, NC network database for 1990 AM peak period. The operating mode mix of VMT in cold transient, hot transient and hot stabilized modes, also known as the mix of cold-starts, hot-starts and stabilized mode trips, is derived on a link by link basis. The results are aggregated by facility type and the location of link segments. It is observed that the operating mode fractions in transient and stabilized modes could vary widely across different facility types geographic locations. The aggregated operating mode fractions derived from the assignment analysis indicates that a lesser proportion of VMT operates in cold and hot transient modes when compared to the operating mode mix derived from the Federal Test Procedure (FTP).

Longitudinal Analysis of Bicycle Count Variability: Results and Modeling Implications

This research examines composition, weather, and time-of-year count variability for a longitudinal bicycle count program. The results indicate greater variability in the PM peak period than in the AM peak period, and this variability is at least partially attributable to the presence of nonutility cyclists. Second, all locations display the same functional form for count volume by both temperature and amount of precipitation, and it appears that temperature may be a better predictor of volume than the amount of precipitation for these data. Third, the results of a Poisson model statistically confirm many of the factors thought to influence cyclists, and identify that a single count volume may be biased by as much as ±15% depending on the time of year in which the count was undertaken. Finally, a new bicycle functional classification system based on PM peak-period composition is proposed.

Effect of traffic information on commuters' propensity to change route and departure time

AbstractConsiderable public and private resources are devoted to the collection and dissemination of real‐time traffic information in the Chicago area. Such information is intended to help individuals make more informed travel decisions, yet its effect on behavior remains largely unexplored. This study evaluates the effect of traffic information on travelers' route and departure time changes and provides a stronger basis for developing advanced information systems.Downtown Chicago automobile commuters were surveyed during the AM peak period. The results indicate that a majority of the respondents access, use and respond to information. For example, individuals use travel information to reduce their anxiety—even if they do not change travel decisions; this indicates that information may have “intrinsic” value. That is, simply knowing traffic conditions is valued by travelers. More than 60% of the respondents had used traffic information to modify their travel decisions. Multivariate analysis using the ordered probit model showed that individuals were more likely to use traffic reports for their route changes if they perceived traffic reports to be accurate and timely, and frequently listened to traffic reports. Respondents were more likely to change their departure times if they perceived traffic reports to be accurate and relevant, and frequently listened to traffic reports.The implication for Advanced Traveler Information Systems are that they may be designed to support both enroute and pre‐trip decisions. ATIS performance, measured in terms of accuracy, relevance and timeliness would be critical in the success of such systems. Further, near‐term prediction of traffic conditions on congested and unreliable routes (where conditions change rapidly) and incident durations is desirable.