International Conference on Transportation and Development 2022: Traffic Operations and Engineering

A recent study contributed a noteworthy automated vehicles’ (AV) data set based on the GPS sensors by a series of empirical experiments by commercial AVs equipped with the adaptive cruise control (ACC) system. We find that the dynamic action of these vehicles cannot respond in real-time to the data detected by the sensors; that is, there is a time delay between real dynamic response and data perception. It can be observed that the best fitting scheme always appears between 1.8 s and 2.0 s by using a simple linear car-following model under different time delays, whether the target speed is set to high or low. A linear car-following model for AVs, including sensor delay and a first-order inertial process, is established to better explain the existence of this time delay with string stability analyses. By combining experimental data and theoretical derivation, the overall effect of the time delay in the ACC control system can be calculated. This study provides more accurate error and delay reference for the future car-following model design of AV in real operation on the road.

This study developed a traffic sign detection and recognition algorithm based on the RetinaNet. Two main aspects were revised to improve the detection of traffic signs: image cropping to address the issue of large image and small traffic signs and more anchors with various scales to detect traffic signs with different sizes and shapes. The proposed algorithm was trained and tested in a series of autonomous driving front-view images in a virtual simulation environment. Results show that the algorithm performed well under good illumination and weather conditions. The drawbacks are that it sometimes failed to detect objects under bad weather conditions like snow and failed to distinguish speed limit signs with different limit values.

Truck climbing lanes are designed to enable a slow stream of heavy vehicles in segments with a high road incline. Research has shown the positive safety effects of these lanes as they separate slow traffic from non-slow traffic. In higher road slopes, many vehicles are restricted to follow a heavy vehicle before they can overtake on the adjacent lane. This becomes a more pressing issue with higher demands on the highway as the frequency of lane changes increases. With the addition of truck climbing lanes onto multilane highways, a change in the speed profile of each lane in a multilane highway is observed. In this study, the addition of a truck climbing lane in the Placer County of California is investigated. The truck climbing lane is constructed in the eastbound direction besides the existing two lanes on the I-80 corridor in Placer County. The before and after analysis of this construction revealed that the average speeds in the two existing lanes increase from 60.6 mph to 64.4 mph in the horizon year (2046) as faster vehicles are not restricted to periods of following heavy vehicles. Also, the increase in speeds increases the highway throughput in the horizon year. Furthermore, additional to the capacity increase, the truck climbing lane can reduce unnecessary rapid braking and passing on the non-truck climbing lanes which would yield more environmental-friendly operation of the multilane highway.

This study provides insights on enhancing snowplow drivers’ training using simulation technologies. Existing literature was reviewed to understand the present snow and ice removal practices of different state Departments of Transportation (DOTs) and corporations, as well as their effectiveness. Two alternatives (in-house capability or contractor training) for setting up a simulation training platform were evaluated using Ohio snowplow drivers as a case study. Based on these, a benefit-cost (BC) analysis was conducted to identify the best alternative. The state-of-the-practice review shows that many DOTs are already adopting simulation training technologies, but with varying degrees of complexities in implementation with reported benefits, including crash reductions, fuel savings, and training cost reductions. Also, a survey of snowplow experts showed that to set up an effective simulation training platform, a driving simulator with powerful motion and visual systems is needed. Finally, a BC analysis showed that the best alternative for agencies is to develop an in-house simulation training capability.

The road condition images provide valuable information for transportation operation managers to enhance their snowplow operation practices and improve roadway safety during winter storms. In recent years, various state Departments of Transportation (state DOTs) in the US have employed GPS-based automatic vehicle location (AVL) systems, consisting of several detached equipment pieces (e.g., communication modem, mobile data computer, and GPS device), along with mounted cameras in snowplow fleet to communicate real-time snowplow locations as well as road condition images. However, using such a multi-part system in states with less frequent winter storms, such as Texas, where the trucks may be used for other purposes than plowing, is undesirably restrictive and costly. The objective of this research is to develop a multi-purpose all-in-one mobile data collection system for snowplow fleet to communicate road condition images along with snowplow locations during winter operation. In this research, a cellular Android tablet with an embedded camera and GPS device is programmed and installed in snowplows to automatically collect and transfer geotagged road condition images at predetermined time intervals (e.g., 10 min) when the snowplow is moving at the speed of 5 mph or more. Moreover, a web-map ArcGIS interface is developed to visualize the collected road condition images and snowplow locations along with weather-related information provided by national weather services (e.g., current ambient temperatures, snowfall forecast, and issued watches, warnings, and advisories). The developed system was implemented in the TxDOT Wichita Falls District on five (5) snowplows for the 2020–21 winter season. The system captured the historic winter storm in Texas in February 2021. The results, including more than 120 geotagged road condition images visualized on the web-map interface, provided helpful information for transportation operation managers to better oversee the real-time road condition and make well-informed decisions during snowplow operation. The developed system is expected to facilitate equipping the snowplow fleet with a mobile data collection system to communicate adverse road conditions, improve snowplow operational decisions, and consequently decrease weather-related crashes on roadways.

Autonomous vehicles (AVs) have the potential to minimize car crashes caused by human errors, reduce emissions to allow a sustainable environment, and provide smaller headways and reaction times to reduce congestion and facilitate mobility. Since AV adaptations are still in their early stages, the use of traffic modeling and simulation can justify the positive impacts of AVs on our transportation systems. This study aims to use the PTV VISSIM microscopic simulator to evaluate the impact of AVs on traffic mobility at an urban intersection in Stillwater, Oklahoma. To prepare for the VISSIM simulation, 15-min field traffic counts during the afternoon peak hour, signal timings, and the intersection geometry were collected at the intersection. Three AV behavior types (AV normal, AV cautious, and AV allknowing) were penetrated at different rates and compared with the conventional driving through several simulation scenarios. The simulation results demonstrate that AVs can significantly reduce the queue delay, stopped delay, travel time, and queue length, leading to an improvement in traffic flow and mobility.

Highway congestion often occurs, and dynamic speed limit control plays a good role in congestion control. By combing the relevant research contents of dynamic speed limit, the shortcomings of the current research on expressway dynamic speed limit are found, and the mechanism of expressway congestion is analyzed; what’s more? The process of expressway dynamic speed limit is determined; otherwise the principle of dynamic speed limit is clarified; furthermore the expressway dynamic speed limit scheme is given based on capacity, design speed, and traffic volume. Based on VISSIM simulation, the traffic operation of an expressway in Ningbo is simulated, the simulation model is calibrated by orthogonal test method, and then the secondary development of VISSIM is carried out by python programming, so as to realize the simulation evaluation of the effect of expressway dynamic speed limit control. The simulation results show that compared with static speed limit control, dynamic speed limit can significantly improve the traffic congestion of expressway, and the operation delay is reduced by 14.9%; the travel time is also shortened by 16.3%; in addition, the traffic efficiency of congested sections is improved.

An adaptive traffic signal controller (ATSC) combined with a connected vehicle (CV) concept uses real-time vehicle trajectory data to regulate green time and has the ability to improve travel time in a signalized corridor. This study introduces an innovative “slow poisoning” cyberattack for a waiting-time-based ATSC algorithm and a corresponding detection strategy. The objectives of this paper are to (1) develop a “slow poisoning” attack generation strategy for an ATSC and (2) develop a prediction-based “slow poisoning” attack detection strategy using a recurrent neural network—i.e., long short-term memory model. We have generated a “slow poisoning” attack modeling strategy using a microscopic traffic simulator—Simulation of Urban Mobility (SUMO)—and used generated data from the simulation to develop both the attack model and detection model. Our analyses revealed that the attack strategy is effective in creating a congestion in an approach and that our detection strategy can capably flag the attack.

Geographic information systems (GIS) are frequently used to represent traffic information in the form of shapefiles. Compared to GIS, building information modeling (BIM) provides extra useful features that can be utilized to extend and improve the functionality of the GIS model. This study was conducted to convert an existing GIS traffic data system to a more robust BIM-based model for traffic data management. In order to achieve the model interaction from GIS to BIM, a visual programming tool, Dynamo, was utilized to transfer the shapes and attributes of the shapefiles. This paper presents an innovative and automatic interaction method for the model conversion from a GIS model to a BIM platform. The traffic model interaction from GIS to BIM is described through an example with a local area traffic database. It is expected that the transformed BIM model can be used for comprehensive road traffic information management.

This paper provides a system architecture for an infrastructure-based cooperative perception fusion engine for next-generation traffic control. This engine will provide a complete state-space digital representation with measurable accuracy to support a wide-range of applications. The architecture includes inputs, functional flow, data standardization recommendations, outputs, and supported applications. The cooperative perception engine addresses critical needs with respect to accelerating the benefits of automation through intelligent roadway infrastructure, which complements and accelerates connected and automated vehicle (CAV) technology. The cooperative perception acquires and fuses information from sensors (radar, LiDAR, and cameras) and CAVs to perceive roadway traffic states of moving objects, creates a complete 3D digital representation of that state-space, and communicates it to downstream application such as intelligent signal control, safety and energy applications, and cooperate driving applications. The intelligent roadway infrastructure approach, as opposed to a vehicle-centric approach, is more scalable because it can be deployed to the roughly 300,000 signalized intersections more readily than over 300 million vehicles in the United States, and accrues early-stage benefits equitable to all roadway users addressing safety, equity, fuel efficiency, and greenhouse gas reduction.

Probe-based data provides an opportunity to explore incident identification and travel time prediction. While it has led to considerable research on freeways, little research has been reported on arterials. There are various uncertainties such as interrupted traffic flows and regular variation of signal timing schemes in existence of a signalized corridor. To address the challenge, this paper presents a case study on US-27 corridor in the state of Ohio by developing an innovative framework. The exponential smoothing technique, artificial neural network (ANN) techniques, and Bayes theorem are integrated into the developed framework for predicting signalized corridor travel time, especially under the influence of an incident. The models are tested with real data source. The results indicate a good performance of the developed models with a higher accuracy, implying that the developed mechanism has a great potential to apply into CAV environment as CAV data evolves more widely available toward ubiquity.

As the complexity of traffic situation in each intersection is different, the applicability of the evaluation model established through one or more fixed parameters to all intersections will be greatly affected. Effectively assessing the congestion state of intersections to guide the optimization of the signal light timing plan has become one of the important methods to solve the traffic congestion problem. First, this paper establishes a set of ten traffic flow parameters, including capacity, average delay, parking rate, and obtains a set of sample data sets containing these ten parameters every 10 min through mathematical models and simulations. Next, the K-Means algorithm is used to cluster all the samples multiple times, and the mutual information method is used to filter the correlation between each parameter and the clustering result label. The set of traffic flow parameters with high correlation with the state is obtained by sorting the correlation degree, and data normalization is performed on them. Finally, a tower structure is used to comprehensively evaluate the multi-influencing factors at the intersection, and the traffic states and the thresholds of each state are determined through K-Means clustering method. The results of the simulation experiment show that the dynamic evaluation model proposed in this paper can more accurately reflect the changes of the traffic state.

In this work, we build a traffic signal performance ranking online database for the state of Tennessee, using segmented probe vehicle data from Regional Integrated Transportation Information System (RITIS). The goal is to develop a low-cost and yet effective way to rank traffic signals in the scale of 1 to 10 for performance evaluation and improvement purposes. First, we form intersections using the TMC segments for all road classes except interstate highways. We then use several metrics, including average speed (aka congestion), worst-case travel time (aka planning time index), and the bottleneck ranking data from RITIS to develop a ranking formula, which also takes different days of the week and different times of the day into consideration. Lastly, we create an online database using MySQL, Amazon webservices, and PHP scripting language. The database is web-based: Agencies and the public can browse and search intersections and view their rankings easily using a browser. We hope that our work is helpful to other states that wish to develop similar systems.

Congestion in metropolitan areas of California has many causes. Traffic congestion significantly affects the population in areas where it occurs. A large number of people are adversely affected by congestion each day in California. Congestion causes individuals to suffer delays and incur extra costs that the auto user must pay. Congestion is measured by factors such as the Travel Time Index and the Roadway Congestion Index. Measuring and identifying the causes of traffic congestion and its relative impact is the first step in reducing congestion. Congestion causes environmental degradation, significant loss of productive time, health impacts, and significant economic impacts. This paper explores some of the factors that account for congestion and their relative impacts on congestion in select metropolitan areas in California. The main goal of the study was to investigate the possible impact of road data, road connectivity, socioeconomic, trip, and weather data and on congestion. The results of this research showed a significant effect of some infrastructure and demographic factors on congestion, namely, that congestion is negatively related with lane miles per person, and positively related to the number of links per area. The findings of this study could be used to improve our understanding of causes of urban traffic congestion and methods to alleviate it.

Discretionary lane changing (DLC) decision on freeways is influenced by several factors and varies across vehicle classes. This research aims to investigate the role of vehicle attributes (e.g., length, width) and flow characteristics (e.g., headways and lead-lag gaps) on the DLC decision making process. We analyzed the vehicle trajectory data extracted from the Federal Highway Administration’s (FHWA) Next Generation Simulation (NGSIM) program for the Interstate 80. The data were collected during rush hours (between 4:00 p.m. and 4:15 p.m. on April 13, 2005) on a segment of Interstate 80 in Emeryville, San Francisco, California. Our data set contains only the discretionary lane change that made by autos and trucks only. As a result, two lanes out of six lanes were used for the purpose of this research. The logistic stepwise selection procedure was applied to estimate statistically significant predictor variables that contribute to increasing the likelihood of DLC. The dependent variable was whether discretionary lane change was executed or not. Nine explanatory variables were included to find out the associated factors for DLC. The logistic regression model identified three statistically significant predictor variables out of seven independent variables considered in the current research. The predictors that increase the probability of discretionary lane change are distance between vehicles (space headway) on the original lane, distance between subject vehicle and vehicle on the target lane (lead gap), vehicle class (auto and truck). Overall, the model results provide an overview of factors associated with discretionary lane change that could be addressed when providing highway safety improvement.

To find optimal routes, many drivers use navigation systems, especially during congested times. This study analyzed traffic patterns during a no-notice evacuation event based on two different scenarios, socially optimal routing (following the primary evacuation routes) and selfish routing (drivers using guidance from navigation systems). The chosen road network was ideal for this study because there existed multiple paths with differing capacities. This study was the first known analysis of the price of anarchy (PoA) during an evacuation event. The PoA was quantified as the difference between socially optimal route scenarios and selfish routing scenarios, utilizing the costs of delay and vehicle emissions. The findings show that the PoA for an evacuation can vary between 1.01 and 1.07, during the timespan of an evacuation. As expected, the network with socially optimal routing was found to have less traffic congestion compared to the network with selfish routing. In addition, the selfish routing scenarios had significantly higher variation in both travel time and delay time. These findings suggest that navigation systems may cause unintended delays and variability during an evacuation. Future research could evaluate different demand profiles, road networks, and/or intersection operation schemes.

In order to ensure that a car-following model’s output is as close to reality as possible, the model’s parameters must be calibrated carefully. As simulation models are computationally expensive and black-boxes, robust and efficient calibration algorithms are required. A Bayesian optimization (BO) framework is presented in this research paper for the calibration problem of car-following models. Like any other optimization methods, BO also comes with its own configurations that need to be chosen wisely in order to achieve the best result. A comprehensive study of BO has been done to find its best configuration for calibrating intelligent driver model (IDM) over a synthetic data set. BO is then used to calibrate the IDM over the real-world highD data set, and it is compared to a well-known method named simultaneous perturbation stochastic approximation (SPSA). BO shows better convergence and higher computational efficiency than SPSA. Moreover, the comparative results of multiple experiments confirm its robustness in response to random initializations, as opposed to SPSA, which may sometimes get stuck in a local optimum.

Trajectory prediction has long been a basic task in traffic situation awareness. However, challenges in interactions, continuity, and multi-modal behavior have to be handled for more accurate prediction. In order to obtain a better performance, previous works have started to incorporate recurrent neural networks, graph neural networks, even transformer into their models. Some models construct the temporal and spatial features independently, and then synthesize the processed features to predict the trajectories; meantime other models construct the these features together and add modules to distinguish the features in pursuing of better effects. Their models, however, may have not jointly trained temporal and spatial modules so they may neglect agents’ temporal and spatial features that supposed to be pertinent, resulting in the separation among features and decrease its performance; in addition, those models encoding temporal and spatial features simultaneously are compelled to add functional modules to identify where is an agent’s histories. Given the above disadvantages, instead, we propose a novel agents-separated prediction-former. We think that the input can be taken in a two-dimension form but not in a sequence, which easily tackles the task of distinguishing the agents’ identity. Since we can only put in one scene in one time, we incorporate agents’ velocities through differential. Whereafter, we make it auto-regressive to confirm the trajectories’ temporal continuity. Thus, our model can jointly incorporate temporal and spatial features as well as identify each agent’s histories. Based on our novel idea, we then construct our final trajectory predictor adopting a probability model to obtain probability outputs to generate multi-modal trajectories that accord with the reality. Experiments reveals that our model can improve the performance in several proposed data sets.

Traffic measures are conventionally evaluated in terms of vehicular traffic over time expressed by the annual average daily traffic (AADT). Estimation of vehicular traffic requires a significant effort and data availability. Connected vehicle data provides an alternative approach to associate journeys of a subset of vehicular traffic at roadway facilities with AADT estimates and expand this association to estimate traffic measures where traffic counts are not available. In this study, an AADT prediction model was developed as a function of connected vehicle journeys. The correlation analysis indicated a statistically significant correlation factor of 0.954 (p-value < 0.001). Assuming a normal distribution, linear regression was implemented, and the results indicated that AADT estimates were equal to the number of journeys multiplied by a factor of 0.315 (model R² = 0.941). Traffic estimates from connected vehicle data contribute to traffic analysis and safety evaluations as a function of measures of exposure at locations where regular traffic count data collection is not feasible.

The emerging connected vehicle (CV) technology provides an opportunity to formulate an ambulant data platform that allows the actual data transfer among multiple vehicles and roadside infrastructure. However, the challenge to achieve such an envisioned cooperation lies in the connectivity and interoperability of the CV data and the signal control infrastructure. To address this challenge, we developed distributed computing data intelligence algorithms to make roadside units (RSU) more intelligent. To test the efficiency of the developed algorithms, a simulation-based test bed was established within the VISSIM platform, and a corridor consisting of four signalized intersections has been evaluated at different CV penetration rates under three different traffic conditions (light, mild, and heavy traffic conditions). The simulation results indicate that average vehicle delay and queue length can be reduced by 46.04% and 56.15%, respectively, under a 50% CV penetration rate after optimization. The benefits can be foreseen from improved mobility performance, including increased throughput, reduced travel time reduction, and reduced fuel consumption and emissions.