Link dynamic vehicle count estimation based on travel time distribution using license plate recognition data

Signal phase and timing (SPaT) information is a necessary input for traffic performance evaluation. However, current SPaT estimation studies mainly focus on estimation of cycle length or green time of a certain movement, and are realized mostly by floating car data whose data quality significantly affects the estimation accuracy. As license plate recognition (LPR) systems are becoming a widely implemented and reliable data source in China, in this study, a SPaT estimation method is proposed using the LPR data for fixed-time controlled intersections. The SPaT estimation problem is formulated as a bi-level programming model to find the optimal match between the phase boundaries and the LPR passing time series in the study period. Evaluation is done with an empirical case and compared with an existing method, results show that the estimation accuracies of the phase duration can reach 90.0%, outperforming the existing method and demonstrating great potential for practical application.

In urban traffic systems, traditional signal control can no longer meet the increasing traffic demand, and local congestion is severe during peak hours. Fixed detector data is characterized by high deployment density, full sample detection and restorable vehicle paths, providing new data support for coordinated signal control. We propose an optimization method for regional coordinated control, with the Changsha road network as the study area. Firstly, based on License Plate Recognition (LPR) data, the road network is divided into sub-networks and combined with the boundary control for regional coordinated control. Then, the critical path is taken as the control object, and the phase coordination rate is introduced as the optimization objective. The particle swarm optimization algorithm improved by the logistic chaotic map is used as the global searcher, and sequential quadratic programming is adopted as the local searcher to solve the optimization strategy for the objective function. Finally, a simulated road network is constructed in VISSIM 6.0 simulation software to verify the effectiveness of the strategy. The results show that the optimization strategy reduces intersection delay and saturation by 20.3% and 19.3% in the critical path coverage area. Road travel time and the average number of vehicle stops are reduced by 21% and 22.1%. This indicates that the regional coordinated control based on the improved particle swarm algorithm can better alleviate the peak hour traffic congestion.

To partition an urban network into several subareas (i.e., subarea partition) is a vital step for regional coordinated signal control. The correlation between intersections must be analyzed for achieving reliable subarea partition results. However, because of the incompleteness of spatial–temporal information in traffic data, previous studies merely explored the relationship between any intersections. Subarea partition considering the correlation of any pair of intersections remains a challenge in a large-scale network. This paper proposes a subarea partition method that integrates a novel correlation-degree model and the Newman fast algorithm with an edge-elimination strategy using automatic license plate recognition (ALPR) data. First, vehicle trips are extracted and a correlation-degree model is developed for measuring the relationship of any pair of intersections. Second, an edge-elimination strategy is proposed to generate candidate subarea partition solutions under conditions with different proportions of correlated edges. Finally, an optimal solution of subarea partition is identified by the ratios of ideal connected intersections to total intersections of different partition solutions’ correlation index (CI). The proposed method was implemented in a real-world urban network in Kunshan, China. The results show that the optimal partition solution can be obtained when the top 33% of correlated edges are maintained, and the ratio of ideal connected intersections’ CI is 72.35% with most of the intersections being connected, which demonstrates the rationality of the proposed partition method in large-scale urban networks.

Discovering patterns and detecting anomalies in individual travel behavior is a crucial problem in both research and practice. In this paper, we address this problem by building a probabilistic framework to model individual spatiotemporal travel behavior data (e.g., trip records and trajectory data). We develop a two-dimensional latent Dirichlet allocation (LDA) model to characterize the generative mechanism of spatiotemporal trip records of each traveler. This model introduces two separate factor matrices for the spatial dimension and the temporal dimension, respectively, and use a two-dimensional core structure at the individual level to effectively model the joint interactions and complex dependencies. This model can efficiently summarize travel behavior patterns on both spatial and temporal dimensions from very sparse trip sequences in an unsupervised way. In this way, complex travel behavior can be modeled as a mixture of representative and interpretable spatiotemporal patterns. By applying the trained model on future/unseen spatiotemporal records of a traveler, we can detect her behavior anomalies by scoring those observations using perplexity. We demonstrate the effectiveness of the proposed modeling framework on a real-world license plate recognition (LPR) data set. The results confirm the advantage of statistical learning methods in modeling sparse individual travel behavior data. This type of pattern discovery and anomaly detection applications can provide useful insights for traffic monitoring, law enforcement, and individual travel behavior profiling.

Weather conditions may significantly impact a series of everyday human decisions and activities. As a result, engineers seek to integrate weather-related data into traffic operations in order to improve the current state of practice. Travel times and speeds are two of the elements of a transportation system that may be greatly affected by the weather resulting in deterioration of roadway network performance. This study aims to investigate the impact of different intensities of rain, snow and temperature levels on macroscopic travel times in the Greater London area (UK) during the period 1 October–10 December 2009. The analysis was carried out for three 2-h periods on weekdays during the morning, afternoon and evening periods. Automatic Number Plate Recognition (ANPR) data obtained from more than 380 travel links are used in the analysis. The main finding is that the impact of rain and snow is a function of their intensity. Specifically, the ranges of the total travel time increase due to light, moderate and heavy rain are: 0.1–2.1%, 1.5–3.8%, and 4.0–6.0% respectively. Light snow results in travel time increases of 5.5–7.6%, whilst heavy snow causes the highest percentage delays spanning from 7.4% to 11.4%. Temperature has nearly negligible effects on travel times. It was also found that the longer links within outer London generally yield greater travel time decreases than those in inner London, and even higher decreases than the shortest links in central London. This research provides planners with additional information that can be used in traffic management to modify planning decisions and improve the transportation system control on a network scale under different weather conditions. In order to determine whether the weather effects are region-specific, continued research is needed to replicate this study in other areas that exhibit different characteristics.

Understanding the nature of human movement is important in many fields, such as transportation management, urban planning, and epidemiology. In this study, we focus on travelers who are grouped by their travel modes in the Tokyo Metropolitan Area. Notably, for each group of travelers, the distribution of daily travel time follows a lognormal distribution. For any travel mode, the distributions of normalized travel time for different years (1968–2008) and distances to the city center converge to a curve, although the average travel time and share of travel modes vary. By employing the concept of utility, we developed a disaggregate model wherein an individual maximizes utility from traveling (=benefit - cost) by modifying daily travel time. This model theoretically explained the observed distributions of travel time. The model was further used to demonstrate travel mode choices. The model parameters were estimated using the distribution of travel time, travel mode share, and manageable assumptions. These results provided interesting interpretations of the perceptual utility of the daily travel time of a population. Moreover, the obtained curves of utility versus time are similar to the shapes of the overall benefits assumed in previous studies. Overall benefit is a concept that has been developed and examined since the 1980s, and our study provides a mathematical foundation for this idea.

This study explores the problem of Multi-Agent Path Finding with continuous and stochastic travel times whose probability distribution is unknown. It is often the case with real-world applications (e.g., automated delivery services in office buildings) that the time required for the robots to traverse a corridor takes a continuous value and is randomly distributed because pedestrians and a wide variety of robots coexist, and the prior knowledge of the probability distribution of the travel time is limited. We propose 1) online re-planning to update the action plan of robots while it is executed and 2) parameter update to estimate the probability distribution of travel time using Bayesian inference as the delay is observed. Through simulations, we empirically compare the performance of our method to those of existing methods.

Probabilistic travel time progression and its application to automatic vehicle identification data

The global outbreak of coronavirus disease 2019 (COVID-19) has affected the urban mobility of nations around the world. The pandemic may even have a potentially lasting impact on travel behaviors during the post-pandemic stage. China has basically stopped the spread of COVID-19 and reopened the economy, providing an unprecedented environment for investigating post-pandemic travel behaviors. This study conducts multiple investigations to show the changes in travel behaviors in the post-pandemic stage, on the basis of empirical travel data in a variety of cities in China. Specifically, this study demonstrates the changes in road network travel speed in 57 case cities and the changes in subway ridership in 26 case cities. Comprehensive comparisons can indicate the potential modal share in the post-pandemic stage. Further, this study conducts a case analysis of Beijing, where the city has experienced two waves of COVID-19. The variations in travel speed in the road network of Beijing at different stages of the pandemic help reveal the public's responses towards the varying severity of the pandemic. Finally, a case study of the Yuhang district in Hangzhou is conducted to demonstrate the changes in traffic volume and vehicle travel distance amid the post-pandemic stage based on license plate recognition data. Results indicate a decline in subway trips in the post-pandemic stage among case cities. The vehicular traffic in cities with subways has recovered in peak hours on weekdays and has been even more congested than the pre-pandemic levels; whereas the vehicular traffic in cities without subways has not rebounded to pre-pandemic levels. This situation implies a potential modal shift from public transportation to private vehicular travel modes. Results also indicate that commuting traffic is sensitive to the severity of the pandemic. This may be because countermeasures, e.g., work-from-home and suspension of non-essential businesses, will be implemented if the pandemic restarts. The travel speed in non-peak hours and on non-workdays is higher than pre-pandemic levels, indicating that non-essential travel demand may be reduced and the public's vigilance towards the pandemic may continue to the post-pandemic stage. These findings can help improve policymaking strategies in the post-pandemic new normal.

The acquisition of real-time and accurate traffic arrival information is of vital importance for proactive traffic control systems, especially in partially connected vehicle environments. License plate recognition (LPR) data that record both vehicle departures and identities are proven to be desirable in reconstructing lane-based arrival curves in previous works. Existing LPR data-based methods are predominantly designed for reconstructing historical arrival curves. For real-time reconstruction of multi-lane urban roads, it is pivotal to determine the lane choice of real-time link-based arrivals, which has not been exploited in previous studies. In this study, we propose a Bayesian deep learning approach for real-time lane-based arrival curve reconstruction, in which the lane choice patterns and uncertainties of link-based arrivals are both characterized. Specifically, the learning process is designed to effectively capture the relationship between partially observed link-based arrivals and lane-based arrivals, which can be physically interpreted as lane choice proportion. Moreover, the lane choice uncertainties are characterized using Bayesian parameter inference techniques, minimizing arrival curve reconstruction uncertainties, especially in low LPR data matching rate conditions. Real-world experiment results conducted in multiple matching rate scenarios demonstrate the superiority and necessity of lane choice modeling in reconstructing arrival curves.

A probabilistic approach for queue length estimation using license plate recognition data: Considering overtaking in multi-lane scenarios

The paper studies the problem of predicting traffic equilibrium (TE) in a transportation network within the framework of decision making among discrete choices in a probabilistic and uncertain environment. Conventional approaches to predict TE typically assume that travel times are deterministic and perceived accurately by the travelers; and some new TE models have considered probabilistic travel times or inaccurate perceptions but not both. In our generalized model, which we refer to as GTESP, the travel time on each route is random and each traveler perceives, possibly inaccurately, a travel time probability distribution for each route which may vary from traveler to traveler. Each traveler uses a disutility function of travel time to evaluate each route and chooses that route which minimizes his expected disutility. GTESP is difficult to solve for general nonlinear disutility functions. However, special cases—in particular when arc travel times are statistically independent and the disutility functions to evaluate route travel times are linear, exponential, or quadratic—are solvable, at least approximately. GTESP is general in the sense that most existing TE models can be shown to be special or limiting cases of GTESP. Furthermore, this paper demonstrates, with illustrative examples, that GTESP appears to capture travelers' risk-taking behavior more realistically than existing TE models.

Riverbank filtration (RBF) is a surface water treatment technology that uses the aquifer adjacent to the river as a natural filter. A production well in the aquifer near the river draws water from both the aquifer and the river. As river water travels through the aquifer to the well, contaminants in the water undergo biologic and chemical reactions while traveling through the subsurface materials, leading to higher quality water at the well as compared with the river. A key parameter in siting the RBF well is the travel time of the contaminants from the river to the well, which must be sufficiently long to allow adequate time for biologic and chemical reactions to occur. Because the well induces flow into the aquifer from all along the river channel, multiple travel paths exist between the river and the well, each with a different contaminant travel time. The water produced at the well contains a collection of contaminant particles from these multiple travel paths and can be characterized by a distribution of travel times. The travel time of the contaminant depends on the groundwater velocity along the flow path, the sorption properties of the contaminant and the aquifer matrix, and the chemical reactions that occur during transport. In this work, we demonstrate the use of an adjoint-based modeling approach to characterize the distribution of travel times at the RBF well. In adjoint modeling, transport is modeled backward in space and time, from the point of view of the RBF well. The output of an adjoint simulation is the distribution of solute travel times at the RBF well at any time of interest. We consider different sorption models and sorption parameters, and we consider different chemical reactions and reaction rates. We simulate the travel time distributions using either MT3DMS or RT3D, depending on the type of chemical reaction, using a flow field generated using MODFLOW.

Travel time distribution studies are fundamental for supporting transportation system reliability studies, particularly for urban road networks. However, such studies are generally based on travel time data sets with limited sample sizes, which provide inconsistent findings. In this paper, a large amount of travel time data collected from the emerging radio frequency identification (RFID) technique are used to conduct empirical investigations and estimations of travel time distributions, and three major findings are determined. First, travel time data are shown to have a complex statistical structure: the travel time distribution is in general peaky, multi-modal, and skewed to the right, which cross validates findings shown in previous publications. Second, unimodal distribution models are shown to be unable to capture the complex statistical dynamics embedded in the travel time data; therefore, a multistate distribution model is more appropriate for modeling travel time distributions. In this respect, a three-component gaussian mixture model (GMM) is tested and results consistently outperform those of unimodal distribution models. Finally, the aggregation time interval is shown to have a trivial effect on the shape of travel time distributions: the travel time distribution is stable under different aggregation time intervals. Future work is recommended to investigate further travel time variabilities and travel time distribution estimations.

Predicting unsaturated solute transport using measured hydraulic parameters has been difficult due to the inherent variability of soil properties, and the difficulty in obtaining accurate estimates of hydraulic properties in situ. The objective of this study was to determine if in situ measurement of soil hydraulic conductivity, the α soil parameter, and the water content (θ) vs. pressure head ( h ) relationship could be used to predict field‐scale solute transport. A series of steady‐state solute transport experiments were conducted on a Fox sand (fine‐loamy over sandy or sandy‐skeletal, mixed, mesic Typic Hapludalf) soil in Ontario, Canada. The transport of Cl − under steady‐state water flux was monitored in three separate experiments using solution samplers. Steady‐state water flux densities applied at the soil surface were 9.72 × 10 −6 , 1.53 × 10 −5 , and 8.68 × 10 −8 m s −1 , respectively, for the three sites. After completion of the transport experiments at Site II, measurements of soil hydraulic conductivity and the α parameter were made using the Guelph pressure infiltrometer (GPI) beside each location and depth where solute breakthrough curves (BTCs) were measured, as well as at the soil surface. Undisturbed soil cores were taken at each location where GPI measurements were made for estimating the parameters in the θ( h ) using van Genuchten's equations. The GPI‐ and core‐measured hydraulic parameters obtained at Site II were used to predict the field‐scale solute travel time probability density function (PDF) at the same site, and at Sites I and III using a stochastic‐convective model. Observed solute travel time PDFs were predicted quite well at high surface water fluxes, which were close to the field saturated hydraulic conductivity, K fs by both the GPI and core methods. Both methods underpredicted the variability of the observed travel time PDF.