Traffic State Prediction Research Articles

A Multi-Parameter Chaotic Fusion Approach for Traffic Flow Forecasting

Since traffic is a time-dependent and complicated non-linear system. Chaos Theory has been specifically designed to identify chaotic behaviour and properties for such systems. Previous researches have been restricted to single traffic parameter, which fails to get actual behaviour of the traffic systems. In this work, we have devised the multi-parameters chaos prediction method to describe the tendency of traffic from different aspects for prediction of single parameter chaotic time series. The fusion method is considered the relationship of different traffic parameters according to the phase space reconstruction. Taking the exit ramp of Xishanping expressway in Beibei district, Chongqing Municipality as the example, the feasibility and reliability of traffic state prediction were tested. The trail results indicates that more features of realistic traffic conditions are reflected by fusing multiple traffic parameters, and gain real accuracy improvements in traffic prediction. Compared with three single-parameter time series prediction methods, the mean absolute relative error of the multi-parameter prediction method decreased by 2.42, 2.39 and 0.8 respectively, the mean absolute relative error decreased by 2.33, 3.25 and 1.27, and the equal coefficient reached 0.9528 with a slight increase. Proposed method will definitely generates the opportunity for next generation of traffic control that are better able to detect the dynamic states of traffic, and therefore more effectively prevent the traffic congestion and pollution in the urban areas worldwide.

Integrated Variable Speed Limits and Lane-Changing Control for Freeway Lane-Drop Bottlenecks

There are generally two kinds of traffic control strategies to relieve traffic congestion in lane-drop bottlenecks: variable speed limits (VSL) control and lane-changing (LC) control. However, VSL has limited or even no effect due to many mandatory LC maneuvers near bottlenecks, while LC fails to reduce traffic congestion when traffic demand is high. Although a few control methods combine VSL and LC, they do not consider the interaction between VSL and LC, which rules out many potentially good alternatives. We instead propose an integrated VSL and LC control method under a connected and automated vehicle (CAV) environment, which can consider the interaction and simultaneously find the values of LC numbers and speed limits to maximize traffic efficiency. Our control is in the framework of the model predictive control (MPC), which consists of prediction, optimization, and implementation. We adopt an improved multi-class cell transmission model (CTM) for traffic state prediction, then use the genetic algorithm (GA) for optimization which optimizes traffic network performance, and implement our control method in the SUMO platform. Simulation results demonstrate that our control method greatly improves the capacity of the road and is robust to different traffic demands and scenarios. Our control outperforms no control and VSL-only control in travel time and exhaust emissions, which reduces total travel time by 23.86% to 44.62% and exhaust emissions by 10.29% to 48.19%.

Short-Term Traffic Flow Prediction Using the Modified Elman Recurrent Neural Network Optimized Through a Genetic Algorithm

Traffic stream determining is an essential part of the intelligent transportation management system. Precise prediction of traffic flow provides a basis for other tasks, like forecasting travel time. While traditional methods have some merits for improving traffic prediction precision in some ways, high precision, considering different circumstances, is still difficult to achieve. This article presents a short-term traffic flow prediction model based on the Modified Elman Recurrent Neural Network model (GA-MENN) to deal with this practical problem. In GA-MENN, the algorithm of Elman Recurrent Neural Network is modified, optimized through the Genetic Algorithm (GA) and considered weather conditions, weekday, hour and day’s classification to forecast the vehicle velocity in Tehran streets and highways. The traffic data were collected from the online Google Map API service for 139 routs in 7 districts in Tehran. The method improves prediction precision and also lowers the prediction error rate, according to experimental results. Exploratory outcomes verify the superior performance of the proposed traffic condition prediction model over Regression Multi-layer Perceptron, Linear Regression, Logistic Regression, Probabilistic Neural Network, Regression Generalized Feedforward, Time-lag Recurrent Network, Support Vector Machine model, Elman neural network, K- NN model, ARIMA, Kalman filter model, Convolutional Neural Networks (CNNs), SARIMA, and Long Short-Term Memory (LSTM) model. To the best of our knowledge, this is the first occasion when that traffic stream is gauged in urban roads and avenues in this specific way.

Short-term traffic state prediction from latent structures: Accuracy vs. efficiency

Recently, deep learning models have shown promising performances in many research areas, including traffic states prediction, due to their ability to model complex nonlinear relationships. However, deep learning models also have drawbacks that make them less preferable for certain short-term traffic prediction applications. For example, they require a large amount of data for model training, which is also computationally expensive. Moreover, deep learning models lack interpretability of the results. This paper develops a short-term traffic states forecasting algorithm based on partial least square (PLS) to help enhance real-time decision-making and build better insights into traffic data. The proposed model is capable of predicting short-term traffic states accurately and efficiently by capturing dominant spatiotemporal features and day-to-day variations from collinear and correlated traffic data. Three case studies are developed to demonstrate the proposed model in short-term traffic prediction applications.

Road traffic state prediction based on a graph embedding recurrent neural network under the SCATS.

Road traffic state prediction is one of the essential and vital issues in intelligence transportation system, but it is difficult to get high accuracy due to the complicated spatiotemporal characteristics of traffic flow data, especially under the Sydney coordinated adaptive traffic system. In this work, we represent the traffic road network as a graph and propose a novel traffic flow prediction framework named the graph embedding recurrent neural network (GERNN). It could tackle the difficulty in the road traffic state prediction. We conduct numerical tests to compare GERNN with other existing methods using a real-world dataset.

Extended Observer for Urban Traffic Control Based on Limited Measurements From Connected Vehicles

Current Urban Traffic Control (UTC) systems rely heavily on inductive loop detectors. The emergence of Connected Vehicles (CV) opens new possibilities for improving signal control, while reducing the need for loop detectors. However, for low penetration rates, the CV measurements are sporadic and thus difficult to exploit by existing UTC systems. In this paper, a methodology that enables cycle-to-cycle traffic state estimation and prediction based on limited CV measurements is presented. Furthermore, the proposed formulation enables fusion of CV with other data sources and their integration in any UTC system. The developed Extended Observer (EO) is a discrete-time, variable-dimension implementation of the Extended Kalman Filter. It does not require loop detectors and is independent of the type of signal control. The evaluation focuses on the queue length estimation. Especially in oversaturation, the EO outperforms the CV measurements for all examined penetration rates. Moreover, the highest benefit is observed for the lowest penetration rates. The results show that for oversaturation and low penetration rates the EO improves the CV measurements 13-31%. In addition, the EO is tested with an adaptive UTC system by feeding the fused estimation from CV and camera measurements. The error in queue length estimation from the EO is significantly lower than the error based on stochastic arrivals. Additionally, the results show a reduction in delays at the examined signal and the complete intersection. Overall, this paper sheds light on the potential benefits from enhancing limited CV measurements in order to contribute immediately to current UTC systems.

A New Switched State Jump Observer for Traffic Density Estimation in Expressways Based on Hybrid-Dynamic-Traffic-Network-Model.

When faced with problems such as traffic state estimation, state prediction, and congestion identification for the expressway network, a novel switched observer design strategy with jump states is required to reconstruct the traffic scene more realistically. In this study, the expressway network is firstly modeled as the special discrete switched system, which is called the piecewise affine system model, a partition of state subspace is introduced, and the convex polytopes are utilized to describe the combination modes of cells. Secondly, based on the hybrid dynamic traffic network model, the corresponding switched observer (including state jumps) is designed. Furthermore, by applying multiple Lyapunov functions and S-procedure theory, the observer design problem can be converted into the existence issue of the solutions to the linear matrix inequality. As a result, a set of gain matrices can be obtained. The estimated states start to jump when the mode changes occur, and the updated value of the estimated state mainly depends on the estimated and the measured values at the previous time. Lastly, the designed state jump observer is applied to the Beijing Jingkai expressway, and the superiority and the feasibility are demonstrated in the application results.

A spatio‐temporal ensemble method for large‐scale traffic state prediction

AbstractHow to effectively ensemble multiple models while leveraging the spatio‐temporal information is a challenging but practical problem. However, there is no existing ensemble method explicitly designed for spatio‐temporal data. In this paper, a fully convolutional model based on semantic segmentation technology is proposed, termed as spatio‐temporal ensemble net. The proposed method is suitable for grid‐based spatio‐temporal prediction in dense urban areas. Experiments demonstrate that through spatio‐temporal ensemble net, multiple traffic state prediction base models can be combined to improve the prediction accuracy.

Multistep speed prediction on traffic networks: A deep learning approach considering spatio-temporal dependencies

Multistep traffic forecasting on road networks is a crucial task in successful intelligent transportation system applications. To capture the complex non-stationary temporal dynamics and spatial correlations in multistep traffic-condition prediction, we propose a novel deep learning framework named attention graph convolutional sequence-to-sequence model (AGC-Seq2Seq). In the proposed deep learning framework, spatial and temporal dependencies are modeled through the Seq2Seq model and graph convolution network separately, and the attention mechanism along with a newly designed training method based on the Seq2Seq architecture is proposed to overcome the difficulty in multistep prediction and further capture the temporal heterogeneity of traffic pattern. We conduct numerical tests to compare AGC-Seq2Seq with other benchmark models using two real-world datasets. The results indicate that our model yields the best prediction performance in terms of various prediction error measures. Furthermore, the variations of spatio-temporal correlations of traffic conditions under different perdition steps and road segments are revealed.

Connected Vehicle-Based Lane Selection Assistance Application

Connected vehicle (CV) technology has great potential to improve the performance of today’s advanced driver assistance systems in terms of safety, energy efficiency, and driving comfort. The aim of this paper is to develop a specific CV application that assists with lane selection, i.e., finding the best travel lane in terms of travel time based on predicted lane-level traffic states. In this paper, a spatial-temporal model (ST-model) was developed, which utilizes spatial and temporal information of road cells to predict future traffic states. This information was used by the proposed lane selection assistance application to select an optimal lane sequence for the application-equipped vehicle. A comprehensive simulation-based evaluation was then conducted under various scenarios, e.g., with different traffic volumes, penetration rates of communication-capable vehicles, and information update cycles. The evaluation results reveal several interesting findings, including: 1) the proposed ST-model outperforms the basic estimation model in terms of traffic state prediction accuracy; 2) travel times of application-equipped vehicles can be reduced by up to 8% with the use of the proposed lane selection assistance application when compared with the baseline, under various traffic scenarios; 3) the application can be effective in the early deployment stage of CV technology, where the penetration rate of communication-capable vehicles is still low; and 4) the potential conflict risk of application-equipped vehicles is reduced, although the application is mainly designed for mobility benefits, due to the more strategic and informed lane changes suggested by the proposed application.

Lane-based short-term urban traffic parameters forecasting using multivariate artificial neural network and locally weighted regression models: A genetic approach

Short-term prediction of traffic conditions on urban arterials has recently become increasingly important because of its vital role in the basic traffic management functions and trip decision-making processes. Such information is useful for optimal infrastructure operation, routing, and trip scheduling. However, forecasting models offering a high accuracy at a fine temporal resolution (e.g., 1 or 5 min) and, especially, lane-based are still rare and need special attention. Given the dynamic and stochastic nature of traffic, this study proposes a genetically optimized artificial neural network (GA-ANN) and locally weighted regression (GA-LWR) multivariate models, for short-term traffic prediction using a combination of multiple traffic variables such as volume, occupancy, and speed, during peak and off-peak periods. The proposed 5-min GA-ANN and GA-LWR disaggregate multivariate models show lower average and 95th percentile (P95) errors, when compared to those reported in the literature. In particular, for peak and off-peak time prediction, the GA-ANN disaggregate multivariate models result in most of the average errors being from 2% to 5% and the 95th percentile errors being from 9% to 10%. On the other hand, for peak and off-peak time traffic prediction, the GA-LWR disaggregate multivariate models show that most of the average errors are lower than 5% and the 95th percentile errors are lower than 10%. Meanwhile, for peak and off-peak time prediction, both GA-ANN and GA-LWR disaggregates models show lower MSE of 0.11–1.84. Hence, such techniques are believed useful for developing a robust urban traffic forecasting system.

Real-Time Dynamic Traffic Control Based on Traffic-State Estimation

The accurate depiction of the existing traffic state on a road network is essential in reducing congestion and delays at signalized intersections. The existing literature in the optimization of signal timings either utilizes prediction of traffic state from traffic flow models or limited real-time measurements available from sensors. Prediction of traffic state based on historic data cannot represent the dynamics of change in traffic demand or network capacity. Similarly, data obtained from limited point sensors in a network provides estimates which contain errors. A reliable estimate of existing traffic state is, therefore, necessary to obtain signal timings which are based on the existing condition of traffic on the network. This research proposes a framework which utilizes estimates of traffic flows and travel times based on real-time estimated traffic state for obtaining optimal signal timings. The prediction of traffic state from the cell transmission model (CTM) and measurements from traffic sensors are combined in the recursive algorithm of extended Kalman filter (EKF) to obtain a reliable estimate of existing traffic state. The estimate of traffic state obtained from the CTM-EKF model is utilized in the optimization of signal timings using genetic algorithm (GA) in the proposed CTM-EKF-GA framework. The proposed framework is applied to a synthetic signalized intersection and the results are compared with a model-based optimal solution and simulated reality. The optimal delay estimated by CTM-EKF-GA framework is only 0.6% higher than the perfect solution, whereas the delay estimated by CTM-GA model is 12.9% higher than the perfect solution.

Overall Traffic Mode Prediction by VOMM Approach and AR Mining Algorithm With Large-Scale Data

Traffic state prediction has been a popular topic, since traffic congestion occurs in most cities and creates inconvenience to human daily life. In this paper, we propose a predicting method for a city’s overall traffic state, in order to help people avoid possible future congestion. Based on the variable-order Markov model theory and probability suffix tree, the proposed method makes use of the association rules to improve forecasting performance. Since the association rules are extracted from the historical traffic data and describe the traffic state relations among different regions, the proposed method can improve the predictive accuracy. The traffic system in Shanghai is considered as our experimental case because of its complicated and gigantic coupling transport network. The experimental results indicate more accuracy compared with other methods in long-term traffic status prediction.

Adaptive Rolling Smoothing With Heterogeneous Data for Traffic State Estimation and Prediction

Spatial–temporal traffic state estimation and the prediction of urban expressways is a vital component of traffic management and information systems. The adaptive smoothing method is one of the most frequently used approaches to estimate traffic states. However, the fixed filter parameters used in existing approaches sometimes fail to characterize traffic dynamics well. To better capture generation, propagation, and mitigation dynamics of traffic congestion, we propose an adaptive rolling smoothing (ARS) approach by dynamically tuning the filter parameters in a rolling horizon scheme for online applications. The fusion of heterogeneous traffic data combines aggregate traffic measurements (e.g., traffic flow rate, time occupancy, and speed collected by remote microwave sensors) and disaggregate information (e.g., timestamps of individual vehicles detected by license plate recognition cameras). A nonlinear traffic flow filter based on the virtual trajectory algorithm is established to reconstruct the spatial–temporal traffic state and estimate experienced travel times of individual vehicles. The results demonstrate the capability and effectiveness of the proposed ARS approach in the historical traffic state estimation and short-term traffic flow prediction. Complicated traffic states of weaving, merging, and diverging segments can be well distinguished by reconstructing time-space speed diagrams. The proposed approach can be extended to develop efficient missing data imputation algorithms and hierarchical control strategies for heterogeneously congested urban expressways.

Real-Time Prediction of Fuel Consumption Based on Digital Map API

At present, digital maps can estimate the travel time of each trip’s route but cannot offer a fuel consumption estimation at the same time. In this paper, we develop a fuel consumption model based on the Vehicle Specific Power (VSP) distribution, which can connect the traffic condition prediction with the fuel consumption model to predict fuel consumption. First, the traffic condition forecasting and the trip time of each route can be obtained through the digital map Application Programming Interface (API). Secondly, the users need to provide the engine displacement of their vehicles to match the fuel consumption model. Then, the fuel consumption prediction application based on Android is developed to forecast the fuel consumption by using traffic prediction data. Finally, the fuel consumption provided by the On-Board Diagnostic (OBD) data is used to verify the proposed application, and the forecasting error is less than 20%.

Dynamic identification of urban traffic congestion warning communities in heterogeneous networks

Network-wide traffic control strategies (e.g. perimeter control and route guidance) in urban networks have recently been mainly studied to relieve or postpone congestion based on the theory of macroscopic fundamental diagram (MFD). Nevertheless, these studies are mostly applied to the statically partitioned networks or the dynamic networks that fail to fully consider traffic state prediction, conflicting with strongly spatiotemporal variability of traffic congestion or objective of active traffic management (ATM). This paper proposes a methodology to dynamically identify critical congestion warning areas from heterogeneous urban road networks, which aids to design efficient perimeter control approaches. In the methodology, a dynamic directed weighted network is built on the base of the link connectivity and the real-time traffic loads, and a link travel time prediction method based on Kalman filter is developed to calibrate directed weight values and undirected input values for links. With the undirected link input information, a dynamic congestion warning community detection method which consists of three consecutive steps is developed. Firstly, it could capture emergence of new congestion areas based on the definition of congestion seed intersection. Secondly, expansion and regression of each congested area could be achieved with the objectives of spatial compactness and traffic condition homogeneity. Thirdly, a two-level merging algorithm of adjacent different congestion areas is designed utilizing modularity model in community detection of complex networks. The proposed methodology is validated using ground truth data from downtown network of Jinan City in China. The results show that the proposed algorithms can efficiently track congestion evolutionary processes and effectively detect congestion warning areas from the test real network.

Deep reinforcement learning enabled self-learning control for energy efficient driving

To address the air pollution problems and reduce greenhouse gas emissions (GHG), plug-in hybrid electric vehicles (PHEV) have been developed to achieve higher fuel efficiency. The Energy Management System (EMS) is a very important component of a PHEV in achieving better fuel economy and it is a very active research area. So far, most of the existing EMS strategies are either just simply following predefined rules that are not adaptive to changing driving conditions; or heavily relying on accurate prediction of future traffic conditions. Deep learning algorithms have been successfully applied to many complex problems and proved to even outperform human beings in some tasks (e.g., play chess) in recent years, which shows the great potential of such methods in practical engineering problems. In this study, a deep reinforcement learning (Deep Q-network or DQN) based PHEV energy management system is designed to autonomously learn the optimal fuel/electricity splits from interactions between the vehicle and the traffic environment. It is a fully data-driven and self-learning model that does not rely on any prediction, predefined rules or even prior human knowledge. The experiment results show that the proposed model is capable of achieving 16.3% energy savings (with the designed PHEV simulation model) on a typical commute trip, compared to conventional binary control strategies. In addition, a dueling Deep Q-network with dueling structure (DDQN) is also implemented and compared with single DQN in particular with respect to the convergence rate in the training process.

An Infrastructure-Assisted Crowdsensing Approach for On-Demand Traffic Condition Estimation

With the widespread use of smartphones and the continuous increase of their capabilities, a new sensing paradigm has emerged: mobile crowdsensing. The concept of crowdsensing implies the reliance on the crowd to perform sensing tasks and collect data about a phenomena of interest. Due to the benefits it offers in terms of time and cost savings in terms of sensors' deployment and maintenance, the concept of mobile crowdsensing is now being adopted in the area of intelligent transportation. In this context, drivers or pedestrians equipped with sensor enabled smartphones collaborate to collect information about roads and traffic. However, the current solutions proposed for the use of crowdsensing for the collection of traffic related data adopt an opportunistic continuous sensing approach, which entails high resource consumption on the server and mobile device side, a high communication overhead, while offering little control of the users over the sensing activity. In this paper, we address these limitations by proposing an infrastructure-assisted on-demand crowdsensing approach for the real time detection and prediction of traffic conditions in an area of interest. Our approach combines the strengths of mobile crowdsensing, with the support of the mobile infrastructure, a multi-criteria algorithm for the participants' selection, and a deductive rule-based model for traffic condition estimation. The proposed solution was validated through a combination of prototyping and simulated traffic traces, and the results show a significant reduction in terms of resources' consumption and network overhead, while reaching high accuracy for the traffic condition estimation.

Multi-source traffic data reconstruction using joint low-rank and fundamental diagram constraints

Accurate and complete traffic information plays an important role in traffic monitoring, traffic state estimation and prediction in all intelligent transportation systems. However, in practice, data provided from various types of traffic sensors may suffer from many issues, such as heterogeneity, missing data and bias. Therefore, extracting unbiased, accurate and complete data from multiple traffic data sources, referred to as multi-source traffic data reconstruction, is critical before data can be utilized. There are many existing methods for multi-source traffic data reconstruction, but two issues remain challenging and unaddressed: the effectiveness of reconstruction is fairly limited to attributes of available data, especially data loss ratio and reliability; the probability distribution of data assumed, which brings in bias prior to data analytics. This paper proposes a novel framework that integrates low-rank representation and fundamental traffic flow models into multi-source traffic data reconstruction. It makes full use of multi-source data, and can simultaneously impute missing data and eliminate outliers/errors. Multiple parameters in the traffic flow models are learned from large-scale spatio-temporal data, which ensures the extracted and imputed data are consistent with traffic flow physics. Experimental results on real-world traffic data demonstrate the superiority of this method over existing methods.

Short‐Term Traffic Volume Forecasting with Asymmetric Loss Based on Enhanced KNN Method

Short‐term traffic volume forecasting is one of the most essential elements in Intelligent Transportation System (ITS) by providing prediction of traffic condition for traffic management and control applications. Among previous substantial forecasting approaches, K nearest neighbor (KNN) is a nonparametric and data‐driven method popular for conciseness, interpretability, and real‐time performance. However, in previous related researches, the limitations of Euclidean distance and forecasting with asymmetric loss have rarely been focused on. This research aims to fill up these gaps. This paper reconstructs Euclidean distance to overcome its limitation and proposes a KNN forecasting algorithm with asymmetric loss. Correspondingly, an asymmetric loss index, Imbalanced Mean Squared Error (IMSE), has also been proposed to test the effectiveness of newly designed algorithm. Moreover, the effect of Loess technique and suitable parameter value of dynamic KNN method have also been tested. In contrast to the traditional KNN algorithm, the proposed algorithm reduces the IMSE index by more than 10%, which shows its effectiveness when the cost of forecasting residual direction is notably different. This research expands the applicability of KNN method in short‐term traffic volume forecasting and provides an available approach to forecast with asymmetric loss.