Natalia Ruiz Juri

Integrated Traffic Simulation-Statistical Analysis Framework for Online Prediction of Freeway Travel Time

This paper introduces a novel approach to the online short-term prediction of point-to-point freeway travel time, combining statistical forecasting techniques with traffic simulation. At every freeway entrance point, a time series analysis model based on traffic detector counts is used to predict traffic demands, whose flow through the freeway segment is simulated by a cell transmission model. This procedure, applied within a rolling-horizon framework, generates online travel time predictions consistent with traffic flow theory. Experimental results obtained from synthetic data strongly suggest that the estimates obtained with this methodology are robust and accurate. For a wide range of congestion conditions and freeway segment lengths, more than half of the predictions errors were found to be smaller than 15%. Moreover, 80% of these errors fell below 40 s when the actual travel times ranged between 3 and 10 min. Further analyses of the model sensitivity to traffic detector coverage revealed that detector separations of approximately 1 mi resulted in the most precise travel time estimates. In addition to its satisfactory performance, the proposed framework is flexible, and it can make use of additional online data and easily incorporate different forecasting and simulation techniques. Therefore, this work provides a powerful tool for online travel time prediction, suitable for a variety of practical implementation conditions and requirements.

A Model of Ridesourcing Demand Generation and Distribution

Ridesourcing has experienced exponential growth in recent years, yet its impact on individual travel are unclear and have not been adequately examined. Recently, an Austin-based ridesourcing company released a large dataset containing disaggregate trip-level information. In this research, we use this new dataset in tandem with several publicly available data sources to estimate two models: a spatially lagged multivariate count model, which is used to describe how many trips are generated in a specific zone on both weekdays and weekend days; and a fractional split model, which helps us identify the characteristics of zones that attract ridesourcing trips. Our results show spatial dependence in ridesourcing trips among proximally located zones, as well as correlation between weekday and weekend day trips originating in a zone. Another interesting finding is the identification of a possible substitution effect between ridesourcing and transit use for weekday trips. Moreover, our results suggest that different income segments in the population may use ridesourcing for different activity purposes. From a travel behavior researcher perspective, the results in this paper identify aggregate area-level variables impacting ridesourcing, which can guide future efforts to better understand the demand for ridesourcing as well as the demand for autonomous and connected vehicles.

Investigation of Centroid Connector Placement for Advanced Traffic Assignment Models with Added Network Detail

Advanced traffic assignment models, such as simulation-based dynamic traffic assignment, typically incorporate more detailed network representations than do traditional planning models. In this context, the placement of centroid connectors may have a significant effect on model performance, and attention must be paid to their number and location to avoid unrealistic congestion or low utilization of minor roadways by local traffic. Given that the manual inspection of centroid connector placement may be too time-consuming in large regional networks, this paper proposes two simple automatic centroid connector placement strategies for dynamic traffic assignment applications. The first approach radially distributes the connectors to the nearest nodes and is intended to exemplify some limitations of the most common techniques in practice. The second strategy involves dividing the centroid and subsequent demand into two parts, distributing the demand across one sub-centroid linked to nearby nodes and one linked to the periphery, and thus effectively establishing a bilevel distribution. A modification of this strategy involves eliminating nodes at signalized intersections as viable candidates for connection. As part of the evaluation of the methods, a new metric, the locality factor, has been introduced to describe the use of minor streets by local traffic. The numerical experiments, conducted on two real-world networks, exemplify the effects of the incorporation of local streets and the placement of centroid connectors on model results. Sensitivity testing and limited field data comparisons suggest that the bilevel centroid connector placement strategy achieves more realistic results.

Using National Performance Management Research Data Set for Corridor Performance Measures: A US-281N Corridor Case Study

The National Performance Management Research Data Set (NPMRDS), made available by Federal Highway Administration in 2013, provides fine-resolution travel-time data, which have been used in numerous network performance management and operations applications. This article discusses corridor-level performance measures computed using the NPMRDS. Three measures are analyzed on a 20.2-mile long corridor in San Antonio, Texas, including corridor travel time, corridor travel-time reliability, and day-to-day variation in travel time. The primary contributions of this article are the analysis of the impact of using two different approaches for travel-time aggregation across segments—instantaneous and time-dependent approaches—and defining a mean absolute error-based method to identify days when travel times significantly deviate from typical traffic conditions. The findings suggest that the temporal patterns of corridor travel times obtained using instantaneous and time-dependent aggregation approaches are similar; however, instantaneous travel-time estimates lead to inaccuracies that become more apparent during peak hours and for longer segments. In addition, it is found that a k -means clustering analysis performed on daily travel-time profiles provides a useful statistic for corridor performance analysis. Using this methodology, 9.23% of weekdays in the 20-month study period are classified as atypical for the corridor. The numerical results reinforce the value of the NPMRDS in estimating corridor performance measures and highlight potential limitations of traditional techniques for evaluating corridor performance measures when applied in practice to support enhanced traffic planning and operations.

Supporting large scale connected vehicle data analysis using HIVE

Connected vehicles (CVs) are vehicles that can exchange messages containing location and other safety-related information with other vehicles, and with devices affixed to roadside infrastructure. While the main purpose of vehicle connectivity is to enhance safety, the data generated by CVs has an enormous potential to support transportation planning and operations. However, handling the vast volume of data produced by CVs presents considerable challenges for researchers in the transportation domain. This paper presents a case study of using HIVE to facilitate CV data analysis based on the largest CV data set publicly released to date. We characterize the data analytic tasks that are expected to enable transportation planning research, and investigate several approaches to increase the corresponding query efficiency and throughput. This study compares the use of HIVE in conjunction with the MapReduce and Spark programming frameworks, analyzes its performance using different data storage formats, and exemplifies potential use cases.

Computation of Skims for Large-Scale Implementations of Integrated Activity-Based and Dynamic Traffic Assignment Models

Integrated activity-based model (ABM) and dynamic traffic assignment (DTA) frameworks have emerged as promising tools to support transportation planning and operations, particularly in the context of novel technologies and data sources. This research proposes an approach to characterize the implementation of integrated ABM-DTA models and seeks to facilitate the interpretation and comparison of frameworks and, ultimately, the selection of appropriate tools. The importance of the dimensions considered in this characterization is illustrated through a detailed analysis of the computation of skims. Skims are the level of service (LOS) metric produced by DTA models, and the computation of skims may impact the performance and convergence of ABM-DTA applications. Numerical results from experiments on a regional ABM-DTA model in Austin, Texas, suggest that skims produced at relatively small time steps (10 to 30 min) may lead to a faster integrated model convergence. Finer time-grained skims are also observed to capture sharper temporal peaking patterns in the LOS. This work considers two skim computation methods; the analysis of the results suggests that simpler techniques are adequate, as the inherent variability of travel times from simulation overshadows any gain in precision from more complex methods. This study also uses promising techniques to visualize and analyze the model results, a challenging task in the context of highly disaggregate models and the subject of further research. The insights from this research effort can inform future research on the implementation of ABM-DTA methods and practical applications of existing frameworks.