Search

Network state degradation probabilities increase with traffic density. Specifically, at critical or congested densities, any over-input traffic flow may cause degradation, local traffic jams, and even hysteresis of the road network. Existing macroscopic traffic control methods usually target the maximum capacity or minimum delay to improve the road network efficiency and do not consider the negative effect of perimeter control on the state transfer of the road network. This study provides a method to prevent the transfer of the network state when implementing boundary control. First, according to the data detected for the road network, the concept of a conditional value at risk is used to establish a risk decision model that considers the influence of the boundary input flow rate on sub-region state degradation. Then, based on the risk decision model, we propose a state transfer risk decision (STRD) perimeter control method for a large-scale traffic network with multiple sub-regions. This perimeter control method predicts the traffic demand of every sub-region, selects an acceptable traffic flow range at the sub-region boundary by risk interval, regards the maximum trip completion flow and the minimum total travel time as the decision-making objectives, and controls multiple sub-region boundaries. The simulation results show that compared with proportional integral (PI) control and no STRD (NSTRD) control, the STRD control scheme can effectively prevent the state transfer of all sub-regions, improve the trip completion flow, and decrease the travel delay of the network.

The area-based (i.e., non-lane based) heterogeneous traffic (as in developing countries) has significant differences from the lane-based homogeneous traffic (as in developed countries). In area-based traffic, drivers generally ignore the lane markings and perceive the entire road space while progressing longitudinally. The traditional car-following and lane changing models are not directly applicable to model such driving behaviour. This research aims to microscopically model the dynamics of the subject vehicle in area-based traffic flow. The modelling is conducted in two steps: area selection, and vehicle movement. In step-1, a multinomial logit (MNL) model is considered for selecting the area-based movement direction of the subject vehicle. The choice space of the subject vehicle is divided into numbers of realistic radial cones considering the possible moving directions of the subject vehicle in the next simulation time step. These radial cones form the alternatives for the decision of the subject vehicle. The attributes of the alternatives are defined in terms of angular deviation from the direction of the flow, spacing and relative speed. This model is calibrated and validated for both cars and motorcycles using real trajectory data, and the results demonstrate the good performance of the model. In step-2, a modified intelligent driving model (MIDM) is proposed to simulate the next position of the subject vehicle (along the selected direction as the outcome of the model in step-1). The parameters of the MIDM model are calibrated using real data for cars and motorcycles. The deterministic parameters for MNL model and the empirical distribution of the MIDM model parameters are utilized to stochastically simulate vehicle trajectories with the initial and boundary conditions determined by the real dataset. The comparison of the macroscopic properties between the simulated and real datasets provides promising results for the simulation applicability of the proposed modelling framework.

In the future, when traffic streams comprise a mix of conventional and automated vehicles (AVs), AVs may be employed as mobile actuators to regulate or manage traffic flow across an urban road network to enhance its performance. This paper develops a path-control scheme to achieve the system optimum (SO) of the network by controlling a portion of cooperative AVs (CAVs) as per the SO routing principle. A linear program is formulated to delineate the scheme and determine the minimum control ratio (MCR) of CAVs to achieve SO. The properties of the MCR are mathematically and numerically investigated. Numerical examples based on real-world networks reveal that the SO of most of the tested networks can be achieved with an MCR below 23%. Considering the low market penetration of AVs at early stages of their deployment, we further investigate a joint path-based control and pricing scheme to replicate SO. Numerical examples demonstrate the remarkable synergy of these combined instruments on reducing the MCR with little collected tolling revenue.

Presence of autonomous vehicles (AVs) affects traffic flow characteristics of a mixed traffic stream comprising human-driven vehicles. To model the impact of AVs on the saturation flow of arterials and highways, we propose analytical models to derive the expected value and variance of headway of a traffic stream with mixed AVs and conventional human-driven (or normal) vehicles (NVs), given the expected penetration rate of AVs. The proposed model considers the arrangement (order) of AVs and NVs in the mixed stream and the expected, the lowest, and the highest achievable headways and their variability are analytically derived. Moreover, the total delay of a two-lane road with interrupted traffic flow is analytically modeled for various possible lane allocation policies: (a) dedicated lanes, (b) mixed-mixed lanes, (c) mixed-NV lanes, and (d) mixed-AV lanes. Microsimulation experiments demonstrate the validity of the developed models for the average headway and its variability, as well as the delay formulas.

Use of shared autonomous vehicles (SAVs) for last-mile transportation can improve transit use and reduce road congestion. We investigate the problem of optimal integration of SAVs with transit. By optimal, we mean that transit should be used instead of or to complement SAVs only when it reduces the total system travel time. Although greater use of transit reduces congestion, transit often increases travel time due to frequent stops and transfers. By using a continuous approximation to passenger and vehicle movements, we formulate the problem as a linear program using the link transmission model for traffic flow. One of the main challenges is associating passenger movement with vehicles that can have different destinations (e.g. a transit stop). Although the mathematical program is linear, it nevertheless has a large number of variables. We find a suboptimal solution using a rolling horizon method, which greatly reduces the required computation time. We also demonstrate an example of a possibly unfair passenger-to-vehicle ordering, and propose a first-come-first-served greedy algorithm to match passengers. A suite of experimental results on the Sioux Falls network show that using transit decreases total system travel time, especially when SAV fleet sizes are small. Transit also decreases the time travelers spend waiting, but tends to increase in-vehicle travel time. The methodology could be useful both for future SAV operators and for planners seeking to predict the effects of SAVs on traffic congestion.

Vehicular traffic between different Origin-Destination (OD) pairs for a typical hourly period may statistically correlate with each other. The covariance mainly generated from the daily variation of travel patterns, network topology, and trip chaining activities of household members can be particularly high during the morning peak hour. With the increasing attention on the OD demand variance and covariance in stochastic road networks, a new criterion is proposed in this paper for measuring the estimation accuracy of OD demand covariance. The mathematical properties of this proposed criterion are analyzed to better understand the relationship between the estimation errors of mean and covariance of OD demands. This paper aims to investigate how to optimize the traffic count locations for minimizing the weighted maximum deviation of estimated mean and covariance of OD demands from the “true” values. To consider the effects of stochastic OD demands on the traffic count location problem in the proposed model, link choice proportions are regarded as stochastic variables and updated by an adapted traffic flow simulator in this study. Both the weighted-sum approach and bi-objective approach are examined with the adaption of the firefly algorithm (FA) to solve the single-objective and bi-objective problems. Numerical examples are presented to demonstrate the effects, with and without considering the covariance of the OD demands for the optimization of traffic count locations.

This study deals with urban transport network flow estimation based on Cellphone Location (CL) and License Plate Recognition (LPR) data. We first propose two methods to filter CL data and extract the spatio-temporal traffic features for a specific road segment. A tailored machine learning approach is developed, including two components: a tangible multi-grained scanning ensemble learning model and a novel two-stage zero-shot learner. The former aims to estimate traffic flow on a single link with both filtered CL data, extracted spatio-temporal traffic features, and LPR data by incorporating the unique merits thereof. The latter is capable of estimating traffic flow on those links with only CL data by considering the spatial features of these links and relevant land-use information. Finally, case studies are analysed to demonstrate the impressive performance of the tailored machine learning approach.

Due to its importance in Intelligent Transport Systems (ITS), traffic flow prediction has been the focus of many studies in the last few decades. Existing traffic flow prediction models mainly extract static spatial-temporal correlations, although these correlations are known to be dynamic in traffic networks. Attention-based models have emerged in recent years, mostly in the field of natural language processing, and have resulted in major progresses in terms of both accuracy and interpretability. This inspires us to introduce the application of attentions for traffic flow prediction. In this study, a deep learning based traffic flow predictor with spatial and temporal attentions (STANN) is proposed. The spatial and temporal attentions are used to exploit the spatial dependencies between road segments and temporal dependencies between time steps respectively. Experiment results with a real-world traffic dataset demonstrate the superior performance of the proposed model. The results also show that the utilization of multiple data resolutions could help improve prediction accuracy. Furthermore, the proposed model is demonstrated to have potential for improving the understanding of spatial-temporal correlations in a traffic network.

It is becoming common to use instrumented probe vehicles (IPV) equipped with perception sensors (e.g., LIDAR or radar) to monitor ambient traffic for empirical studies of traffic dynamics and driver behavior, while the rapidly advancing field of autonomous vehicles (AV) uses perception sensors to identify and mitigate hazards. Yet the perception sensors are far from perfect, so targets of interest go undetected. The “unseen” targets might lead to inaccurate theories of traffic flow or worse, autonomous vehicle crashes. The challenge with the unseen targets is that it is difficult to detect the absence. There is little study of how vehicle-mounted perception sensors perform “in the wild” because often the perception sensors themselves usually provide the best measurements available. The focus of this work is IPV data for offline study, but the general findings are also relevant for perception sensors used in AV applications. While there is a growing number of IPV data sets, to date most suffer from the lack of independent validation of the perception sensors. This paper fuses individual vehicle actuations from loop detectors and concurrent ambient vehicle trajectories collected from an IPV to provide a rare opportunity to assess the perception sensor performance in situ, using the loop detector data to “see” what is missed by the perception sensors. While the specific results are unique to the IPV, one should expect similar biases from other perception sensor data; though without an independent measure like the loop detectors in this study, it will be difficult to quantify what goes unseen in those data sets. These findings also highlight the need to develop means to actively calibrate mobile perception sensors for AV's. It is important to capture the behavior of these systems in situ, as maintained by private owners, though we envision in the case of an AV the role of the loop detectors might be replaced with localization information from other AV's. The data fusion has added benefits for the study of traffic dynamics, with the loop detectors providing continuous monitoring at the detector locations and the IPV providing spatially rich information about the ambient traffic as it travels between detector stations. This fused data set will be available upon publication.

A deep learning model is adopted for predicting block-level parking occupancy in real time. The model leverages Graph-Convolutional Neural Networks (GCNN) to extract the spatial relations of traffic flow in large-scale networks, and utilizes Recurrent Neural Networks (RNN) with Long-Short Term Memory (LSTM) to capture the temporal features. In addition, the model is capable of taking multiple heterogeneously structured traffic data sources as input, such as parking meter transactions, traffic speed, and weather conditions. The model performance is evaluated through a case study in Pittsburgh downtown area. The GCNN-based model outperforms other baseline methods including multi-layer LSTM and LASSO with an average testing MAPE of 10.6% when predicting block-level parking occupancies 30 min in advance. The case study also shows that, in generally, the prediction model works better for business areas than for recreational locations. We found that incorporating traffic speed and weather information can significantly improve the prediction performance. Weather data is particularly useful for improving predicting accuracy in recreational areas.

Ride-sourcing companies such as Uber and Lyft represent a popular and growing mode of transit in cites worldwide. These companies employ surge pricing in real time to balance the needs of both drivers and riders. The prediction of surge prices in the next few minutes to hours encapsulates the complex evolution of service fleets and service demand in the short term. Surge pricing, if effectively predicted and disseminated to both drivers and riders, can be used to more efficiently allocate vehicles, save users money and time, and provide profitable insight to drivers, which ultimately helps the efficiency and reliability of transportation networks. This paper explores the spatio-temporal correlations between the urban environment, traffic flow characteristics, and surge multipliers. We propose a general framework for predicting the short-term evolution of surge multipliers in real-time using a log-linear model with L 1 regularization, coupled with pattern clustering. This model is able to predict Uber surge multipliers in Pittsburgh up to two hours in advance using data from the previous hour out-performing the overall mean and the historical average in all but 3 of the 49 locations in Pittsburgh and outperforming three non-linear methods in 28 of the 49 locations. The model is able to out-perform the overall mean, historical mean, and non-linear methods on Lyft surge multipliers in Pittsburgh up to 20 min in advance. Cross-correlation of Uber and Lyft surge multipliers is also explored.

Railway operations in dense traffic areas are very sensitive even to small disturbances, and thus require careful planning and real-time management. Dwell times in stations are in particular subject to a high variability and are hard to predict; this is mostly due to the interactions between passengers and the railway system during the dwelling process. This paper presents a data-driven approach for assessing the influence of the numbers of alighting, boarding and on board passengers on the dwell time. We propose to split the dwell time into a deterministic component depending on the passenger flow, called the Minimum Dwell Time, and a random component. A method for estimating the minimum dwell time is provided. Based on the knowledge of this value, observations can be classified according to the main determinant of dwell time, namely timetable constraints or passenger exchange. The latter observations are used for estimating the conditional distribution of dwell time given passenger flows. Numerical experiments are carried out on stations located inside the dense traffic area of Paris suburban network. The obtained results indicate that the presented method can be used for a variety of applications, such as capacity assessment or stochastic simulation.

Signal optimization has received significant attentions from the research community. However, during peak hours, the performances of existing models still have room for improvements, especially when channelized section spillover (CSS) occurs. The evolution of CSS is dynamic in nature, not only due to the interactions between traffic flow of different movements, but also because existing CSS influences the queuing process in a channelized section and contributes to new CSS formation at the next cycle. Neglecting such spatial-temporal interaction in current traffic signal optimization procedures may lead to suboptimal results. A Monte Carlo Tree Search-based model is proposed to solve the intersection optimization problem (named MCTS-IO) with explicit modeling of CSS dynamic evolution. The model works in a rolling horizon way. At each decision point, MCTS-IO simulates the intersection by selecting a sequence of phases, and progressively updates the relative preferences of the phases. After all simulations are performed, the phase with the best policy PI is selected. Both the PI and decision space can be customized, which ensures algorithm flexibility. The method is tested against Synchro results with both stable and variable demand, which demonstrates the proposed model is always able to find a solution better than Synchro.

With automated and cooperative driving making its breakthrough, and related systems in fast development, their future influence and impact on roads and traffic may be extensive. Truck platooning is such an application that relies on the development of Cooperative Adaptive Cruise Control (CACC) and is said to be practice ready. While the main advantages of truck platooning lie in emission and energy reduction, claims are also being made for the influence on traffic flow. In this paper, we pose hypotheses based on some of the main claims. We also attempt to substantiate and give quantitative proof of the potential effects of truck platooning on traffic flow performance. The simulation model LMRS-IDM+ is extended to encompass the main influencing dynamics related to potential effects of truck platooning, based on empirical findings. The effects of truck platooning were tested for the influence of traffic states, truck gap settings, platoon sizes, and the share of equipped trucks. This resulted in outcomes regarding the total traffic performance, the performance of traffic at ramps, and the ability of a platoon to remain platooning as part of a case experiment performed on a part of the Trans-European ITS Corridor. The results showed that truck platooning may have a small negative effect on the total non-saturated traffic flow, however with a much larger negative effect on saturated traffic flow. However, drivers may be reluctant to platoon in saturated traffic in any case. The ability of inflowing traffic to merge at on-ramps was found to be affected by truck platoons, with platoon disengagements occurring under various conditions. The applied gap settings for platooning trucks did not significantly affect the merge time, while a higher gap did lead to a higher number of disengagements. The ability of trucks to platoon was positively affected by a greater percentage of equipped trucks and by larger platoon sizes. Shorter gap times also slightly improved the ability of trucks to remain in platooning formation. Finally, recommendations are given to improve platoon strategies and for policymakers to only allow truck platooning outside of busy (near-) saturated traffic, even though drivers may be reluctant to use the system in these conditions. Also, recommendations are made to investigate potential differences in the effects between the European and American contexts for truck-platooning.

Vehicles are extremely vulnerable to single-vehicle accidents under some hazardous driving conditions (i.e. strong wind, icy or snowy road surface). An integrated framework is proposed to assess single-vehicle traffic safety performance of stochastic traffic flow under hazardous driving conditions. Different from most existing studies focusing on a single vehicle moving at a constant speed, for the first time, the proposed work evaluates individual vehicle safety performance based on the time-dependent simulation results of stochastic traffic flow, including instantaneous speeds and positions of each vehicle as a part of simulated traffic flow. Simultaneously, complex geometric and other environmental conditions of the highway system are also considered realistically, not only during the safety assessment process, but also in quantifying the wind loads applied on the vehicles. Finally, with the safety information of each individual vehicle, an overall safety performance index of the whole traffic flow on the highway system is further introduced, which serves as a potential traffic safety performance measure and resilience indicator of transportation infrastructure systems under various hazards. This study has potential applications to not only regular vehicles, but also advanced traffic management and control algorithms for connected and autonomous vehicles in hazardous driving environments.

Most large scale traffic information systems rely on a combination of fixed sensors (e.g. loop detectors, cameras) and user generated data, the latter in the form of probe traces sent by smartphones or GPS devices onboard vehicles. While this type of data is relatively inexpensive to gather, it can pose multiple privacy risks, even if the location tracks are anonymous. In particular, an issue could be the possibility for an attacker to infer user location tracks from anonymous location data, which affects users privacy. In this article, we propose a new framework for analyzing a variety of privacy problems arising in transportation systems. The state of traffic is modeled by the Lighthill–Whitham–Richards traffic flow model, which is a first order scalar conservation law with a concave flux function. Given a set of traffic flow data, we show that the constraints resulting from this partial differential equation are mixed integer linear inequalities for some decision variable. These constraints allow us to determine the likelihood of two distinct location tracks being generated by the same vehicle. We then use these model-based reidentification metrics to train an artificial neural network classifier. Numerical implementations are performed on experimental data from the Mobile Century experiment, and show that this framework significantly outperforms naive reidentification techniques.

Car-following (CF) and lane-changing (LC) behaviors are two basic movements in traffic flow which are generally modeled separately in the literature, and thus the interaction between the two behaviors may be easily ignored in separated models and lead to unrealistic traffic flow description. In this paper, we adopt a deep learning model, long short-term memory (LSTM) neural networks, to model the two basic behaviors simultaneously. By only observing the position information of the six vehicles surrounding the subject vehicle, the LSTM can extract the significant features that influence the CF and LC behaviors automatically and predict the vehicles behaviors with time-series data and memory effects. In addition, we propose a hybrid retraining constrained (HRC) training method to further optimize the LSTM model. With the I-80 trajectory data of NGSIM dataset we train and test the HRC LSTM model, while the results show that HRC LSTM model can accurately estimate CF and LC behaviors simultaneously with low longitudinal trajectories error and high LC prediction accuracy compared with the classical models. We also evaluate the transferability of the proposed model with the US101 dataset and a good transferability result is obtained as well.

This study examines the spatiotemporal dependency between traffic links. We model the traffic flow of 140 traffic links in a sub-network of the Minneapolis-St. Paul highway system for both rush hour and non-rush hour time intervals, and validate the extracted network weight matrix. The results of the modeling indicate: (1) the spatial weight matrix is unstable over time-of-day, while the network weight matrix is robust in all cases and (2) the performance of the network weight matrix in non-rush hour traffic regimes is significantly better than rush hour traffic regimes. The results of the validation show the network weight matrix outperforms the traditional way of capturing spatial dependency between traffic links. Averaging over all traffic links and time, this superiority is about 13.2% in rush hour and 15.3% in non-rush hour, when only the first-order neighboring links are embedded in modeling. In addition, this study proposes a two-step algorithm to search and identify the best look-back time window for upstream links. We indicate the best look-back time window depends on the travel time between two study detectors, and it varies by time-of-day and traffic link.

This paper proposes a framework to combine shockwave analysis (SA) with Bayesian Network (BN) for traffic flow parameters estimation at signalized intersections using vehicle trajectory data. Traffic SA is applied as the basis to derive the analytical expression of individual vehicle trajectory given the fundamental diagram (FD) parameters (e.g., capacity, jam density, free flow speed) as well as the traffic volume. According to SA, the analytical probability distribution of each vehicle’s travel time (defined as the time of a vehicle travelling between an upstream and a downstream location at a signalized intersection) is derived. This probability distribution is parameterized by traffic volume and FD parameters. We then apply a three-layer BN model to construct the relationship between traffic volume, FD parameters, sampled vehicles’ arrival times and intersection travel times. Since traffic volume cannot be measured directly from sampled vehicle trajectories, an expectation maximization (EM) algorithm is introduced to perform learning for estimating the cycle-by-cycle volume (or arrivals) and FD parameters. Such estimated parameters are then used, via SA, to estimate vehicles’ intersection travel times. The proposed combined SA-BN model is evaluated using microscopic traffic simulation data and the NGSIM dataset. Sensitivity analysis of the estimation performance with respect to penetration rates is conducted. In general, the proposed method yields promising estimation accuracy, with the mean absolute percentage error (MAPE) of the estimation bounded by 15%, generally around 5%, as compared with the true value, even under low penetration rates.

Locating the surrounding vehicles is an important environment perception task for autonomous vehicles and advanced driver assistance systems. This task is usually explored based on the sensors’ pre-calibration (e.g. height or pitch angle), but can be challenging when the calibration fails (e.g. on the sloping and uneven roads). In this work, we propose a calibrated feature-point based (CFPB) method to estimate the car-following distance adaptive to rough roads, using a single camera. Instead of using the pre-calibrated parameters, CFPB method is based on the surrounding vehicles’ feature points. It benefits from the fixed coordinate relations among these points, which enables the algorithm to be adaptive to rough roads. These fixed coordinate relations can be calibrated during driving. Namely, after a few seconds of observation to a surrounding vehicle, the CFPB method can start working for more accurate estimation. Furthermore, the proposed algorithm takes the perspective-n-point method as the framework. YOLO V3 and scale-invariant feature transform are applied as the vehicle detector and feature point extractor. The feature point calibration is dynamically updated and the results are smoothened by a Kalman filter. The camera is chosen because of the good performance on objects detection and feature extraction. The proposed algorithm is evaluated on a real-world road with dynamic traffic flow. Mobileye, a widely used car-following distance estimator on AV, is installed during the tests as the benchmark. The results indicate that the proposed method achieves decimeter-level accuracy and outperforms the Mobileye system in cases where the road slope changes significantly.

This study presents a method of mathematical programs with equilibrium constraints (MPEC) to determine the optimal sites of street parking facilities in a working area where drivers travel through or cruise for parking spaces in the morning commute. The optimal sites of street parking facilities incur shortest queuing delay in the capacitated network. We first propose a variational inequality to formulate the heterogeneous quantal response equilibrium (HQRE) for parkers in non-cooperative parking game and the logit-based stochastic user equilibrium (SUE) for travelers traversing the area. The HQRE is developed under the “infinite cyclic flows” assumption, allowing all possible routes, containing the cyclic routes of all times including the infinite. Therefore boundedly rational parkers who search the same curb for more than once via cyclic cruising strategies can be modeled in the parking game. Then based on the equilibrium conditions, the MPEC was developed to minimize the total queuing delay for all drivers in the area. The radial basis function (RBF) model algorithm, one of the surrogate model-based algorithms, was proposed to iteratively solve the MPEC problem. A numerical example from a real working area is presented to illustrate the proposed MPEC problem. The numerical results show that the traffic flow increase caused by cruising for parking can be presented by our model. In addition, the appropriate allocation should avoid parking spaces at the curbs of arterial where travelers drive for traversing the area and the streets intersecting the arterial.

Advanced autonomous vehicle technology is suggested to offer a unique solution to many of the current problems in road transport. This paper studies the impact of the transition in automated driving capabilities (driving mode choice) on road congestion pricing and vice versa, accounting for the interdependencies between traffic flow, the choice level of autonomous driving, effective road capacity and marginal travel cost. It is shown that (in a more or less distant future) when drivers can choose among various levels of car automation during a trip and when shortcomings associated with autonomous driving – e.g. adverse effects regarding privacy, software hacking, liability, safety, driving pleasure, manual control recovery – are non-negligible, their generalized user cost curve is the lower envelope of a swarm of traffic flow dependent individual cost curves whose slope is determined by the level of autonomous driving. This has several implications for our understanding of Pigouvian road congestion pricing. Most importantly, marginal social trip costs are then no longer convex and strictly increasing in traffic flow even when traffic is only ‘normally congested’, thereby introducing the possibility of multiple congestion pricing equilibria even in the absence of what is usually called ‘hypercongestion’. When inconveniences related to autonomous driving are sufficiently high, the imposition of congestion tolls may lead to a situation where drivers abandon autonomous technologies entirely and opt instead for fully manual driving. Our findings suggest that the emergence of advanced autonomous driving capabilities and the transition from manual to automated control on the road will probably make congestion pricing more, not less, complex for researchers and policymakers.

Inefficient traffic control is pervasive in modern urban areas, which would exaggerate traffic congestion as well as deteriorate mobility, fuel economy and safety. In this paper, we systematically review the potential solutions that take advantage of connected and automated vehicles (CAVs) to improve the control performances of urban signalized intersections. We review the methods and models to estimate traffic flow states and optimize traffic signal timing plans based on CAVs. We summarize six types of CAV-based traffic control methods and propose a conceptual mathematical framework that can be specified to each of six three types of methods by selecting different state variables, control inputs, and environment inputs. The benefits and drawbacks of various CAV-based control methods are explained, and future research directions are discussed. We hope that this review could provide readers with a helpful roadmap for future research on CAV-based urban traffic control and draw their attention to the most challenging problems in this important and promising field.

With increasing number of vehicles registered on road the CO2 emissions and the amount of fuel wasted because of congestion has been rising. V2X and Connected and Automated Vehicle (CAV) technology allows vehicles and traffic infrastructure to communicate with each other, and could facilitate better use of existing resources by providing vehicles information about their surroundings and traffic signals. The information regarding the phase of traffic signal, vehicles’ position and vehicles’ speed can be used by drivers and autonomous vehicle control algorithms to make informed decisions as they approach traffic signals. This research proposes a coordination heuristic to improve traffic flow of CAV vehicles through traffic signals and reduce system wide emissions. A simulation was used to compare the system level impacts to emissions, travel time, and wait time of the proposed coordination heuristic to a single vehicle optimization strategy and baseline driver behavior. The results of this research suggest that traffic density and CAV market penetration impact the potential benefits of CAV control strategies. At low CAV penetration rates both the single vehicle and coordination heuristic produced increased emissions primarily driven by driving behavior of non-V2X enabled vehicles. At higher CAV penetration rates the coordination heuristic significantly outperformed the single vehicle control strategy and baseline driver behavior with regards to emissions reduction. The analysis indicates that at 900 vehicles per hour for either of the two driving strategies: coordination heuristic or single-vehicle optimization, to be more preferred over baseline driver behavior, at least 50% of the vehicles should be CAV. Once a threshold penetration rate of CAV vehicles is achieved, vehicles following the coordination heuristic generate nearly 10% fewer CO2 emissions than vehicles following baseline driver behavior, a 30% improvement over the reduction in CO2 emissions obtained using single-vehicle optimization. The vehicles following the coordination heuristic also have less travel time than vehicles following single-vehicle optimization, and less wait times than vehicles following baseline driver behavior.

Traffic prediction, as an important part of intelligent transportation systems, plays a critical role in traffic state monitoring. While many studies accomplished traffic forecasting task with deep learning models, there is still an open issue of exploiting spatial-temporal traffic state features for better prediction performance, and the model interpretability has not been taken serious. In this study, we propose a path based deep learning framework which can produce better traffic speed prediction at a city wide scale, furthermore, the model is both rational and interpretable in the context of urban transportation. Specifically, we divide the road network into critical paths, which is helpful to mine the traffic flow mechanism. Then, each critical path is modeled through the bidirectional long short-term memory neural network (Bi-LSTM NN), and multiple Bi-LSTM layers are stacked to incorporate temporal information. At the stage of traffic prediction, the spatial-temporal features captured from these processes are fed into a fully-connected layer. Finally, results for each path are ensembled for network-wise traffic speed prediction. In the empirical studies, we compare the proposed model with multiple benchmark methods. Under a series of prediction scenarios (i.e., different input and prediction horizons), the superior performance of the proposed framework is validated. Moreover, by analyzing feature from hidden-layer output, the study explains the physical meaning of the hidden feature and illustrate model’s interpretability.

Future Air Traffic Management systems can benefit from innovative approaches that leverage the increasing availability of operational data to facilitate the development of new performance assessment and decision-support capabilities. This paper presents a data analytics framework for high-fidelity characterization of air traffic flows from large-scale flight tracking data. Machine learning methods are used to exploit spatiotemporal patterns in aircraft movement towards the identification of trajectory patterns and traffic flow patterns. The outcomes and potential impacts of this framework are demonstrated with a comparative analysis of terminal area operations in three representative multi-airport (metroplex) systems of the global air transportation system: New York, Hong Kong and Sao Paulo. As a descriptive tool for systematic analysis of the flow behavior, the framework allows for cross-metroplex comparisons of terminal airspace design, utilization and traffic performance. Novel quantitative metrics are created to summarize metroplex efficiency, capacity and predictability. The results reveal several structural, operational and performance differences between the multi-airport systems analyzed. Our findings show that New York presents the most complex airspace design, with considerably higher number of routes and interactions between them, as well as more dynamic changes in the terminal area flow structure during the day, in part driven by the presence of flow dependencies. Interestingly, it exhibits the best levels of traffic flow efficiency on average, both spatially and temporally, yet the highest variability in metroplex configuration performance, with more pronounced performance degradation during inclement weather.

Efficient planning of Airport Acceptance Rates (AARs) is key for the overall efficiency of Traffic Management Initiatives such as Ground Delay Programs (GDPs). Yet, precisely estimating future flow rates is a challenge for traffic managers during daily operations as capacity depends on a number of factors/decisions with very dynamic and uncertain profiles. This paper presents a data-driven framework for AAR prediction and planning towards improved traffic flow management decision support. A unique feature of this framework is to account for operational interdependency aspects that exist in metroplex systems and affect throughput performance. Gaussian Process regression is used to create an airport capacity prediction model capable of translating weather and metroplex configuration forecasts into probabilistic arrival capacity forecasts for strategic time horizons. To process the capacity forecasts and assist the design of traffic flow management strategies, an optimization model for capacity allocation is developed. The proposed models are found to outperform currently used methods in predicting throughput performance at the New York airports. Moreover, when used to prescribe optimal AARs in GDPs, an overall delay reduction of up to 9.7% is achieved. The results also reveal that incorporating robustness in the design of the traffic flow management plan can contribute to decrease delay costs while increasing predictability.

Cooperative Adaptive Cruise Control (CACC) systems have the potential to increase roadway capacity and mitigate traffic congestion thanks to the short following distance enabled by inter-vehicle communication. However, due to limitations in acceleration and deceleration capabilities of CACC systems, deactivation and switch to ACC or human-driven mode will take place when conditions are outside the operational design domain. Given the lack of elaborate models on this interaction, existing CACC traffic flow models have not yet been able to reproduce realistic CACC vehicle behaviour and pay little attention to the influence of system deactivation on traffic flow at bottlenecks. This study aims to gain insights into the influence of CACC on highway operations at merging bottlenecks by using a realistic CACC model that captures driver-system interactions and string length limits. We conduct systematic traffic simulations for various CACC market penetration rates (MPR) to derive free-flow capacity and queue discharge rate of the merging section and compare these to the capacity of a homogeneous pipeline section. The results show that an increased CACC MPR can indeed increase the roadway capacity. However, the resulting capacity in the merging bottleneck is much lower than the pipeline capacity and capacity drop persists in bottleneck scenarios at all CACC MPR levels. It is also found that CACC increases flow heterogeneity due to the switch among different operation modes. A microscopic investigation of the CACC operational mode and trajectories reveals a close relation between CACC deactivation, traffic congestion and flow heterogeneity.

Short-term traffic flow prediction is an integral part in most of Intelligent Transportation Systems (ITS) research and applications. Many researchers have already developed various methods that predict the future traffic condition from the historical database. Nevertheless, there has not been sufficient effort made to study how to identify and utilize the different factors that affect the traffic flow. In order to improve the performance of short-term traffic flow prediction, it is necessary to consider sufficient information related to the road section to be predicted. In this paper, we propose a method of constructing traffic state vectors by using mutual information (MI). First, the variables with different time delays are generated from the historical traffic time series, and the spatio-temporal correlations between the road sections in urban road network are evaluated by the MI. Then, the variables with the highest correlation related to the target traffic flow are selected by using a greedy search algorithm to construct the traffic state vector. The K-Nearest Neighbor (KNN) model is adapted for the application of the proposed state vector. Experimental results on real-world traffic data show that the proposed method of constructing traffic state vector provides good prediction accuracy in short-term traffic prediction.

In this paper, we consider connected cruise control design in mixed traffic flow where most vehicles are human-driven. We first propose a sweeping least square method to estimate in real time feedback gains and driver reaction time of human-driven vehicles around the connected automated vehicle. Then we propose an optimal connected cruise controller based on the mean dynamics of human driving behavior. We test the performance of both the estimation algorithm and the connected cruise control algorithm using experimental data. We demonstrate that by combining the proposed estimation algorithm and the optimal controller, the connected automated vehicle has significantly improved performance compared to a human-driven vehicle.

Major steps towards implementation of autonomous and connected transport are being taken nowadays. The trend of automation technology being used in vehicles by the most important vehicle manufacturing industries is expected to move closer to high or fully Autonomous Vehicles (AVs) through technological advancements in sectors of robotics and artificial intelligence. Vehicles with autonomous driving capabilities are planning to be available on market, in full scale, in the next years. In the longer term substantial benefits are mainly expected for accessibility to transport, safety, traffic flow, emissions, fuel use and comfort. All these potential societal benefits will not be achieved unless AVs are accepted and used by a critical mass of people. Addressing these challenges, this paper: (a) proposes a technology acceptance modelling process by extending the original Technology Acceptance Model (TAM) to explain and predict consumers’ intensions towards AVs, (b) based on the proposed TAM-extended framework, a 30-question survey was conducted in order to investigate the factors influencing consumers’ intensions to use and accept AVs. Results show that the constructs of perceived usefulness, perceived ease to use, perceived trust and social influence, are all useful predictors of behavioral intentions to have or use AVs, with perceived usefulness having the strongest impact. The insights derived from this study could significantly contribute to ongoing research related to technology acceptance of AVs and are expected to allow automobile industries to improve their design and technology.

Modeling impacts of Cooperative Adaptive Cruise Control (CACC) on multi-lane freeway traffic can be challenging. It requires accurate description of the formation and disengagement of CACC vehicle strings when CACC vehicles are mixed with manually driven vehicles in the traffic stream. It also needs to depict the behaviors of CACC vehicles under the influence of CACC operation strategies such as the CACC vehicle managed lane (ML) and implementing the Vehicle Awareness Devices (VAD), which are intended to enhance the CACC string operations. To address these challenges, we extended a state of the art CACC modeling framework to incorporate new algorithms that are essential to describe the interactions among the CACC vehicles and manually driven vehicles in mixed traffic. The updated modeling framework adopts a new vehicle dispatching model to generate the high-volume traffic flow expected to exist due to the CACC string operation. The framework also includes new lane changing rules and automated speed control algorithms that ensure realistic CACC vehicle behaviors at freeway on/off-ramp areas where traffic disturbances might frequently interrupt the CACC string operations. With the model updates, we can further reproduce traffic flow dynamics under the influence of the CACC operation strategies. The modeling capability of the presented framework has been verified via case studies on a simple 4-lane freeway segment with an on-ramp and an off-ramp and a complex 18-kilometer freeway corridor. The case study results indicate that the presented modeling framework not only quantifies the mobility improvements for the study sites under different CACC market penetrations and CACC operation strategies, but also discloses the mechanism that governs the improvement. This study creates a methodology that can estimate detailed kinematics of connected automated vehicles under realistic traffic environments. Findings produced by the methodology are helpful for the future development, implementation and management of the advanced transportation technologies.

Asymmetric driving behavior is a critical characteristic of human driving behaviors and has a significant impact on traffic flow. In consideration of the asymmetric driving behavior, this paper proposes a long short-term memory (LSTM) neural networks (NN) based car-following (CF) model to capture realistic traffic flow characteristics by incorporating the driving memory. The NGSIM data are used to calibrate and validate the proposed CF model. Meanwhile, three characteristics closely related to the asymmetric driving behavior are investigated: hysteresis, discrete driving, and intensity difference. The simulation results show the good performance of the proposed CF model on reproducing realistic traffic flow features. Moreover, to further demonstrate the superiority of the proposed CF model, two other CF models including recurrent neural network based CF model and asymmetric full velocity difference model, are compared with LSTM-NN model. The results reveal that LSTM-NN model can capture the asymmetric driving behavior well and outperforms other models.

Traffic instability is an important but undesirable feature of traffic flow. This paper reports our experimental and empirical studies on traffic flow instability. We have carried out a large scale experiment to study the car-following behavior in a 51-car-platoon. The experiment has reproduced the phenomena and confirmed the findings in our previous 25-car-platoon experiment, i.e., standard deviation of vehicle speeds increases in a concave way along the platoon. Based on our experimental results, we argue that traffic speed rather than vehicle spacing (or density) might be a better indicator of traffic instability, because vehicles can have different spacing under the same speed. For these drivers, there exists a critical speed between 30 km/h and 40 km/h, above which the standard deviation of car velocity is almost saturated (flat) along the 51-car-platoon, indicating that the traffic flow is likely to be stable. In contrast, below this critical speed, traffic flow is unstable and can lead to the formation of traffic jams. Traffic data from the Nanjing Airport Highway support the experimental observation of existence of a critical speed. Based on these findings, we propose an alternative mechanism of traffic instability: the competition between stochastic factors and the so-called speed adaptation effect, which can better explain the concave growth of speed standard deviation in traffic flow.

This study is the first in the literature to model the joint equilibrium of departure time and parking location choices when commuters travel with autonomous vehicles (AVs). With AVs, walking from parking spaces to the work location is not needed. Instead, AVs will drop off the commuters at the workplace and then drive themselves to the parking spaces. In this context, the equilibrium departure/arrival profile is different from the literature with non-autonomous vehicles (non-AVs). Besides modeling the commuting equilibrium, this study further develops the first-best time-dependent congestion tolling scheme to achieve the system optimum. Also, a location-dependent parking pricing scheme is developed to replace the tolling scheme. Furthermore, this study discusses the optimal parking supply to minimize the total system cost (including both the travel cost and the social cost of parking supply) under either user equilibrium or system optimum traffic flow pattern. It is found that the optimal planning of parking can be different from the non-AV situation, since the vehicles can drive themselves to parking spaces that are further away from the city center and walking of commuters is avoided. This paper sheds light on future parking supply planning and traffic management.

Fully automated vehicles could have a significant share of the road network traffic in the near future. Several commercial vehicles with full-range Adaptive Cruise Control (ACC) systems or semi-autonomous functionalities are already available on the market. Many research studies aim at leveraging the potential of automated driving in order to improve the fuel efficiency of vehicles. However, in the vast majority of those, fuel efficiency is isolated to the driving dynamics between a single follower-leader pair, hence overlooking the complex nature of traffic. Consequently fuel efficiency and the efficient use of the roadway capacity are framed as conflicting objectives, leading to fuel-economy control models that adopt highly conservative driving styles.This formulation of the problem could be seen as a user-optimal approach, where in spite of delivering savings for individual vehicles, there is the side-effect of the deterioration of traffic flow. An important point that is overlooked is that the inefficient use of roadway capacity gives rise to congested traffic and traffic breakdowns, which in return increases energy costs within the system. The optimisation methods used in these studies entail high computational costs and, therefore, impose a strict constraint on the scope of problem.In this study, the use of car-following models and the limitation of the search space of optimal strategies to the parameter space of these is proposed. The proposed framework enables performing much more comprehensive optimisations and conducting more extensive tests on the collective impacts of fuel-economy driving strategies. The results show that, as conjectured, a “short-sighted” user-optimal approach is unable to deliver overall fuel efficiency. Conversely, a system-optimal formulation for fuel efficient driving is presented, and it is shown that the objectives of fuel efficiency and traffic flow are in fact not only non-conflicting, but also that they could be viewed as one when the global benefits to the network are considered.

Deep neural networks (DNNs) have recently demonstrated the capability to predict traffic flow with big data. While existing DNN models can provide better performance than shallow models, it is still an open issue of making full use of spatial-temporal characteristics of the traffic flow to improve their performance. In addition, our understanding of them on traffic data remains limited. This paper proposes a DNN based traffic flow prediction model (DNN-BTF) to improve the prediction accuracy. The DNN-BTF model makes full use of weekly/daily periodicity and spatial-temporal characteristics of traffic flow. Inspired by recent work in machine learning, an attention based model was introduced that automatically learns to determine the importance of past traffic flow. The convolutional neural network was also used to mine the spatial features and the recurrent neural network to mine the temporal features of traffic flow. We also showed through visualization how DNN-BTF model understands traffic flow data and presents a challenge to conventional thinking about neural networks in the transportation field that neural networks is purely a “black-box” model. Data from open-access database PeMS was used to validate the proposed DNN-BTF model on a long-term horizon prediction task. Experimental results demonstrated that our method outperforms the state-of-the-art approaches.

International audience; Traffic waves are phenomena that emerge when the vehicular density exceeds a critical threshold. Considering the presence of increasingly automated vehicles in the traffic stream, a number of research activities have focused on the influence of automated vehicles on the bulk traffic flow. In the present article, we demonstrate experimentally that intelligent control of an autonomous vehicle is able to dampen stop-and-go waves that can arise even in the absence of geometric or lane changing triggers. Precisely, our experiments on a circular track with more than 20 vehicles show that traffic waves emerge consistently, and that they can be dampened by controlling the velocity of a single vehicle in the flow. We compare metrics for velocity, braking events, and fuel economy across experiments. These experimental findings suggest a paradigm shift in traffic management: flow control will be possible via a few mobile actuators (less than 5%) long before a majority of vehicles have autonomous capabilities.

This study aims to propose a new algorithm to evaluate the rear-end collision risk under fog conditions considering reduced visibility. The proposed algorithm compares the safe stopping distance of the leading and following vehicles. According to the relationship between clearance distance between the two consecutive vehicles and visibility distance, the car-following maneuver is divided into different situations and the algorithms to calculate the safe stopping distances are suggested correspondingly. The visibility distance is collected by a new visibility detection system with adaptive learning modules and the clearance distance is obtained from a vehicle-based detector. By comparing the safe stopping distances of the following and leading vehicles, the potential rear-end collision can be identified. Subsequently, statistical tests are conducted to analyze rear-end collision risk and compare the different impact of reduced visibility on the collision risk for different vehicle types and lanes. Furthermore, random parameters logistic and negative binomial models are estimated by using individual vehicle data and aggregated traffic flow data, respectively, in order to explore the relationship between the potential rear-end crash and the reduced visibility together with other traffic parameters. The results suggest that the proposed algorithm works well in evaluating rear-end collision risk under fog conditions. It is found that reduced visibility has significant impact on the rear-end collision risk and the impact vary by the different vehicle types and by lane. Further, it is concluded that the driving maneuver of the leading and following vehicles can affect the rear-end collision risk. It is expected that the proposed algorithm can be implemented in a Traffic Management context to improve road safety under fog conditions. Specifically, it is suggested to implement the proposed algorithms in real-time and integrate it with ITS technologies such as Variable Speed Limit (VSL) and Dynamic Message Signs (DMS) to enhance traffic safety when the visibility declines. This car following algorithm could also be extended to adapt for the advent of Connected Vehicles in Fog conditions.

This paper proposes a cooperative strategy of connected and automated vehicles (CAVs) longitudinal control for a mixed connected and automated traffic environment based on deep reinforcement learning (DRL) algorithm, which enhances the string stability of mixed traffic, car following efficiency, and energy efficiency. Since the sequences of mixed traffic are combinatory, to reduce the training dimension and alleviate communication burdens, we decomposed mixed traffic into multiple subsystems where each subsystem is comprised of human-driven vehicles (HDV) followed by cooperative CAVs. Based on that, a cooperative CAV control strategy is developed based on a deep reinforcement learning algorithm, enabling CAVs to learn the leading HDV’s characteristics and make longitudinal control decisions cooperatively to improve the performance of each subsystem locally and consequently enhance performance for the whole mixed traffic flow. For training, a distributed proximal policy optimization is applied to ensure the training convergence of the proposed DRL. To verify the effectiveness of the proposed method, simulated experiments are conducted, which shows the performance of our proposed model has a great generalization capability of dampening oscillations, fulfilling the car following and energy-saving tasks efficiently under different penetration rates and various leading HDVs behaviors.

Connected Automated Vehicles (CAVs) bring promise of increasing traffic capacity and energy efficiency by forming platoons with short headways on the road. However at low CAV penetration, the capacity gain will be small because the CAVs that randomly enter the road will be sparsely distributed, diminishing the probability of forming long platoons. Many researchers propose to solve this issue by platoon organization strategies, where the CAVs search for other CAVs on the road and change lanes if necessary to form longer platoons. However, the current literature does not analyze a potential risk of platoon organization in disrupting the flow and reducing the capacity by inducing more lane changes. In this research, we use driving model of Cooperative Adaptive Cruise Control (CACC) vehicles and human-driven vehicles that are validated with field experiments and find that platoon organization can indeed drop the capacity with more lane changes. But when the traffic demand is well below capacity, platoon organization forms longer CAV platoons without reducing the flow. Based on this finding, we develop the Flow-Aware platoon organization strategy, where the CAVs perform platoon organization conditionally on the local traffic state, i.e., a low flow and a high speed. We simulate the Flow-Aware platoon organization on a realistic freeway network and show that the CAVs successfully form longer platoons, while ensuring a maximal traffic flow.

Traffic flow measurement is very important for traffic management systems. However, the existing traditional measurement approaches are highly time-consuming and expensive to continuously gather the required data and to maintain the corresponding equipment, such as loop detectors and video cameras. On the other hand, many services on the web propose to estimate automobile travel time taking into account traffic conditions thanks to crowd sourced data (Floating Car Data). This work proposes to reconstruct, from estimated travel time, traffic flows using machine learning method. In particular, we evaluate the capacity of Gaussian Process Regressor (GPR) to address this issue. After obtaining estimated travel time on a given route, a clustering process shows that travel duration profiles in each day can be associated to different “types of day”. Then, different regressors are trained in order to estimate traffic flows from travel duration. In the “multi-model” variant, we trained a Regressor for each type of day. Conversely, in the “single model” variant, only one Regressor is trained (the type of day is not taken into account). This is an innovative work to estimate and reconstruct the traffic flow in transportation networks with machine learning method from aggregated Floating Car Data (FCD). A series of experiments are conducted to compare the estimated traffic flows, obtained by the proposed single model and multi-model, and the real ones from actual sensors. The obtained results show that both single model and multi-models can capture the tendency of real traffic flows. Furthermore, the performance can be improved by regulating parameters in GPR machine learning model, such as half width of sample window and sample size (a whole week or only weekdays), and multi-models can highly increase the performance compared with the single model. Therefore, the proposed GPR machine learning and FCD based new method can replace those traditional loop detectors for the measurement of traffic flow.