Download Free Traffic State Estimation And Prediction In Freeways And Urban Networks Book in PDF and EPUB Free Download. You can read online Traffic State Estimation And Prediction In Freeways And Urban Networks and write the review.

Centralization of work, population and economic growth alongside continued urbanization are the main causes of congestion. As cities strive to update or expand aging infrastructure, the application of big data, new models and analytics to better understand and help to combat traffic congestion is crucial to the health and development of our smart cities of XXI century. Traffic support tools specifically designed to detect, forecast and alert these conditions are highly requested nowadays.This dissertation is dedicated to study techniques that may help to estimate and forecast conditions about a traffic network. First, we consider the problem Dynamic Travel Time (DTT) short-term forecast based on data driven methods. We propose two fusion techniques to compute short-term forecasts from clustered time series. The first technique considers the error covariance matrix and uses its information to fuse individual forecasts based on best linear unbiased estimation principles. The second technique exploits similarity measurements between the signal to be predicted and clusters detected in historical data and it performs afusion as a weighted average of individual forecasts. Tests over real data were implemented in the study case of the Grenoble South Ring, it comprises a highway of 10.5Km monitored through the Grenoble Traffic Lab (GTL) a real time application was implemented and open to the public.Based on the previous study we consider then the problem of simultaneous density/flow reconstruction in urban networks based on heterogeneous sources of information. The traffic network is modeled within the framework of macroscopic traffic models, where we adopt Lighthill-Whitham-Richards (LWR) conservation equation and a piecewise linear fundamental diagram. The estimation problem considers two key principles. First, the error minimization between the measured and reconstructed flows and densities, and second the equilibrium state of the network which establishes flow propagation within the network. Both principles are integrated together with the traffic model constraints established by the supply/demand paradigm. Finally the problem is casted as a constrained quadratic optimization with equality constraints in order to shrink the feasible region of estimated variables. Some simulation scenarios based on synthetic data for a manhattan grid network are provided in order to validate the performance of the proposed algorithm.
The Network Fundamental Diagram (NFD) represents dynamics of traffic flow at the network level. It is exploited to design various network-wide traffic control and pricing strategies to improve mobility and mitigate congestion. This study presents a framework to estimate NFD and incorporates it for three specific applications in large-scale urban networks. Primarily, a resource allocation problem is formulated to find the optimal location of fixed measurement points and optimal sampling of probe trajectories to estimate NFD accounting for limited resources for data collection, network traffic heterogeneity and asymmetry in OD demand in a real-world network. Using a calibrated simulation-based dynamic traffic assignment model of Chicago downtown network, a successful application of the proposed model and solution algorithm to estimate NFD is presented. The proposed model, then, is extended to take into account the stochasticity of day-to-day fluctuations of OD demand in NFD estimation.Three main applications of NFD are also shown in this research: network-wide travel time reliability estimation, network-wide emission estimation, and real-time traffic state estimation for heterogenous networks experiencing inclement weather impact. The main objective of the travel time reliability estimation application is to improve estimation of this network-wide measure of effectiveness using network partitioning. To this end, a heterogeneous large-scale network is partitioned into homogeneous regions (clusters) with well-defined NFDs using directional and non-directional partitioning approaches. To estimate the network travel time reliability, a linear relationship is estimated that relates the mean travel time with the standard deviation of travel time per unit of distance at the network level. Partitioning and travel time reliability estimation are conducted for both morning and afternoon peak periods to demonstrate the impacts of travel demand pattern variations.This study also proposes a network-level emission modeling framework via integrating NFD properties with an existing microscopic emission model. The NFDs and microscopic emission models are estimated using microscopic and mesoscopic traffic simulation tools at different scales for various traffic compositions. The major contribution is to consider heterogenous vehicle types with different emission generation rates in the network-level model. Non-linear and support vector regression models are developed using simulated trajectory data of thirteen simulated scenarios. The results show a satisfactory calibration and successful validation with acceptable deviations from underlying microscopic emission model, regardless of the simulation tool that is used to calibrate the network-level emission model.Finally, the NFD application for real-time traffic state estimation in a network experiencing inclement weather conditions is explored. To this end, the impacts of weather conditions on the NFD and travel time reliability relation are illustrated through a scenario-based analysis using traffic simulation. Then, the real-time traffic state prediction framework in the literature is adjusted to capture weather conditions as a key parameter. The extended Kalman filter algorithm is employed as an estimation engine to predict the real-time traffic state. The results highlight the importance of considering weather conditions in the traffic state prediction model.
The past decade witnessed rapid developments in traffic data sensing technologies in the form of roadside detector hardware, vehicle on-board units, and pedestrian wearable devices. The growing magnitude and complexity of the available traffic data has fueled the demand for data-driven models that can handle large scale inputs. In the recent past, deep-learning-powered algorithms have become the state-of-the-art for various data-driven applications. In this research, three applications of deep learning algorithms for traffic state estimation were investigated. Firstly, network-wide traffic parameters estimation was explored. An attention-based multi-encoder-decoder (Att-MED) neural network architecture was proposed and trained to predict freeway traffic speed up to 60 minutes ahead. Att-MED was designed to encode multiple traffic input sequences: short-term, daily, and weekly cyclic behavior. The proposed network produced an average prediction accuracy of 97.5%, which was superior to the compared baseline models. In addition to improving the output performance, the model’s attention weights enhanced the model interpretability. This research additionally explored the utility of low-penetration connected probe-vehicle data for network-wide traffic parameters estimation and prediction on freeways. A novel sequence-to-sequence recurrent graph networks (Seq2Se2 GCN-LSTM) was designed. It was then trained to estimate and predict traffic volume and speed for a 60-minute future time horizon. The proposed methodology generated volume and speed predictions with an average accuracy of 90.5% and 96.6%, respectively, outperforming the investigated baseline models. The proposed method demonstrated robustness against perturbations caused by the probe vehicle fleet’s low penetration rate. Secondly, the application of deep learning for road weather detection using roadside CCTVs were investigated. A Vision Transformer (ViT) was trained for simultaneous rain and road surface condition classification. Next, a Spatial Self-Attention (SSA) network was designed to consume the individual detection results, interpret the spatial context, and modify the collective detection output accordingly. The sequential module improved the accuracy of the stand-alone Vision Transformer as measured by the F1-score, raising the total accuracy for both tasks to 96.71% and 98.07%, respectively. Thirdly, a real-time video-based traffic incident detection algorithm was developed to enhance the utilization of the existing roadside CCTV network. The methodology automatically identified the main road regions in video scenes and investigated static vehicles around those areas. The developed algorithm was evaluated using a dataset of roadside videos. The incidents were detected with 85.71% sensitivity and 11.10% false alarm rate with an average delay of 27.53 seconds. In general, the research proposed in this dissertation maximizes the utility of pre-existing traffic infrastructure and emerging probe traffic data. It additionally demonstrated deep learning algorithms’ capability of modeling complex spatiotemporal traffic data. This research illustrates that advances in the deep learning field continue to have a high applicability potential in the traffic state estimation domain.
This book emphasizes the increasingly important role that Computational Intelligence (CI) methods are playing in solving a myriad of entangled Wireless Sensor Networks (WSN) related problems. The book serves as a guide for surveying several state-of-the-art WSN scenarios in which CI approaches have been employed. The reader finds in this book how CI has contributed to solve a wide range of challenging problems, ranging from balancing the cost and accuracy of heterogeneous sensor deployments to recovering from real-time sensor failures to detecting attacks launched by malicious sensor nodes and enacting CI-based security schemes. Network managers, industry experts, academicians and practitioners alike (mostly in computer engineering, computer science or applied mathematics) benefit from th e spectrum of successful applications reported in this book. Senior undergraduate or graduate students may discover in this book some problems well suited for their own research endeavors.
Travel time is one of the important transportation performance measures, and represents the quality of service for most of the facilities. In other words, one of the essential goals of any traffic treatment is to reduce the average travel time. Therefore, extensive work has been done to measure, estimate, and predict travel time. Using historical observations, traditional traffic analysis methods try to calibrate empirical models to estimate the average travel time of different transportation facilities. However, real-time traffic responsive management strategies require that estimates of travel time also be available in real-time. As a result, real time estimation and prediction of travel time has attracted increasing attention. Various factors influence the travel time of a road segment including: road geometry, traffic demand, traffic control devices, weather conditions, driving behaviors, and incidents. Consequently, the travel time of a road segment varies as a result of the variation of the influencing factors. Predicting near-future freeway traffic conditions is challenging, especially when traffic conditions are in transition from one state to another (e.g. changing from free flow conditions to congestion and vice versa). This research aims to develop a method to perform real-time prediction of near-future freeway traffic stream characteristics (namely speed) and that relies on spot speed, volume, and occupancy measurements commonly available from loop detectors or other similar traffic sensors. The framework of this study consists of a set of individual modules. The first module is called the Base Predictor. This module provides prediction for traffic variables while state of the traffic remains constant i.e free flow or congested. The Congestion Detection Module monitors the traffic state at each detector station of the study route to identify whether traffic conditions are congested or uncongested. When a congestion condition is detected, the Traffic Propagation Module is activated to update the prediction results of the Steady-State Module. The Traffic Propagation Module consists of two separate components: Congestion Spillback activates when traffic enters a congested state; Congestion Dissipation is activated when a congested state enters a recovery phase. The proposed framework of this study is calibrated and evaluated using data from an urban expressway in the City of Toronto, Canada. Data were obtained from the westbound direction of the Gardiner Expressway which has a fixed posted speed limit of 90 km/hr. This expressway is equipped with mainline dual loop detector stations. Traffic volume, occupancy and speed are collected every 20 seconds for each lane at all the stations. The data set used in this study was collected over the period from January 2009 to December 2011. For the Steady-State Module, a model based on Kalman filter was developed to predict the near future traffic conditions (speed, flow, and occupancy) at the location of fixed point detectors (i.e. loop detector in this study). Traffic propagation was proposed to be predicted based on either a static or dynamic traffic pattern. In the static pattern it was assumed that traffic conditions can be attributed based on the observed conditions in the same time of day; however, in the dynamic pattern, expected traffic conditions are estimated based on the current measurements of upstream and downstream detectors. The prediction results were compared to a naïve method, and it was shown that the average prediction error during the “change period” when traffic conditions are changing from free flow to congestion and vice versa is significantly lower when compared to the naïve method for the sample locations (approximately 25% improvement) For the Traffic Propagation Module, a model has been developed to predict the speed of backward forming and forward recovery shockwaves. Unlike classic shockwave theory which is deterministic, the proposed model expresses the spillback and recovery as a stochastic process. The transition probability matrix is defined as a function of traffic occupancy on upstream and downstream stations in a Markov framework. Then, the probability of spillback and recovery was computed given the traffic conditions. An evaluation using field data has shown that the proposed stochastic model performs better than a classical shockwave model in term of correctly predicting the occurrence of backward forming and forward recovery shockwaves on the field data from the urban expressway. A procedure has been proposed to improve the prediction error of a time series model (Steady-State Module) by using the results of the proposed Markov model. It has been shown that the combined procedure significantly reduces the prediction error of the time series model. For the real-time application of the proposed shockwave model, a module (Congestion Detection Module) is required to simultaneously work with the shockwave model, and identify the state of the traffic based on the available measurements. A model based on Support Vector Machine (SVM) was developed to estimate the current traffic state based on the available information from a fixed point detector. A binary model for the traffic state was considered i.e. free follow versus congested conditions. The model was shown to perform better compared to a Naïve model. The classification model was utilized to inform the Traffic Propagation Module. The combined model showed significant improvement in prediction error of traffic speed during the “Change Period” when traffic conditions are changing from free flow to congestion and vice versa. Variability of travel speed in the near future was also investigated in this research. A continuous-time Markov model has been developed to predict the state of the traffic for the near future. Four traffic states were considered to characterize the state of traffic: two free flow states, one transition state, and one congested state. Using the proposed model, we are able to predict the probability of the traffic being in each of the possible states in the near future based on the current traffic conditions. The predicted probabilities then were utilized to characterize the expected distribution of traffic speed. Based on historical observations, the distribution of traffic speed was characterized for each traffic state separately. Given these empirical distributions and the predicted probabilities, distribution of traffic speed was predicted for the near future. The distribution of traffic speed then was used to predict a confidence interval for the near future. The confidence interval can be used to identify the expected range of future speeds at a given confidence level.