Stephen G. Ritchie

Automated detection of lane-blocking freeway incidents using artificial neural networks

Abstract A major source of urban freeway delay in the U.S. is non-recurring congestion caused by incidents. The automated detection of incidents is an important function of a freeway traffic management center. A number of incident detection algorithms, using inductive loop data as input, have been developed over the past several decades, and a few of them are being deployed at urban freeway systems in major cities. These algorithms have shown varying degrees of success in their detection performance. In this paper, we present a new incident detection technique based on artificial neural networks (ANNs). Three types of neural network models, namely the multi-layer feedforward (MLF), the self-organizing feature map (SOFM) and adaptive resonance theory 2 (ART2), were developed to classify traffic surveillance data obtained from loop detectors, with the objective of using the classified output to detect lane-blocking freeway incidents. The models were developed with simulation data from a study site and tested with both simulation and field data at the same site. The MLF was found to have the highest potential, among the three ANNs, to achieve a better incident detection performance. The MLF was also tested with limited field data collected from three other freeway locations to explore its transferability. Our results and analyzes with data from the study site as well as the three test sites have shown that the MLF consistently detected most of the lane-blocking incidents and typically gave a false alarm rate lower than the California, McMaster and Minnesota algorithms currently in use.

STOCHASTIC MODELING AND REAL-TIME PREDICTION OF VEHICULAR LANE-CHANGING BEHAVIOR

Time-varying lane-changing fractions and queue lengths are important lane traffic characteristics which may exhibit significant changes in the presence of a lane-blocking incident. This paper describes a stochastic system modeling approach to estimate time-varying lane-changing fractions and queue lengths for real-time incident management on surface streets. A discrete-time nonlinear stochastic model, which consists of recursive equations, measurement equations, and boundary constraints, is proposed to characterize inter-lane and intra-lane traffic state variables during incidents. To estimate lane-changing fractions and other state variables of the model, a recursive estimation algorithm is developed which primarily involves an extended Kalman filter, truncation, normalization, and a queue-updating procedure. Lane traffic counts are the sole input data used in this method. These data can be readily collected from conventional point detectors. The proposed model was calibrated using video-based data, then tested using simulated data from the TRAF-NETSIM simulation model, Version 5.0, as well as real video-based data sets. Preliminary test results indicate the feasibility of employing the proposed approach to estimate time-varying mandatory lane-changing fractions as well as queue lengths during incidents. The estimated lane-changing fractions and queue lengths can be used not only in better understanding the phenomena of incident-related inter-lane and intra-lane traffic characteristics, but also in developing real-time incident management technologies. Moreover, it is hoped that the results of this study might contribute to future research in related areas such as incident traffic prediction, incident-responsive traffic control and management, and automatic road congestion warning systems for further use in advanced transportation management and information systems.

Use of vehicle signature analysis and lexicographic optimization for vehicle reidentification on freeways

Abstract The vehicle reidentification problem is the task of matching a vehicle detected at one location with the same vehicle detected at another location from a feasible set of candidate vehicles detected at the other location. This paper formulates and solves the vehicle reidentification problem as a lexicographic optimization problem. Lexicographic optimization is a preemptive multi-objective formulation, and this lexicographic optimization formulation combines lexicographic goal programming, classification, and Bayesian analysis techniques. The solution of the vehicle reidentification problem has the potential to yield reliable section measures such as travel times and densities, and enables the measurement of partial dynamic origin/destination demands. Implementation of this approach using conventional surveillance infrastructure permits the development of new algorithms for ATMIS (Advanced Transportation Management and Information Systems). Freeway inductive loop data from SR-24 in Lafayette, California, demonstrates that robust results can be obtained under different traffic flow conditions.

Enhancing the universality and transferability of freeway incident detection using a Bayesian-based neural network

Development of a universal freeway incident detection algorithm is a task that remains unfulfilled despite the promising approaches that have been recently explored. Incident detection researchers are realizing that an operationally successful detection framework needs to fulfill a full set of recognized needs. In this paper we attempt to define one possible set of universality requirements. Among the set of requirements, a freeway incident detection algorithm needs to be operationally accurate and transferable. Guided by the envisioned requirements, we introduce a new algorithm with potential for enhanced performance. The algorithm is a modified form of the Bayesian-based Probabilistic Neural Network (PNN) that utilizes the concept of statistical distance. The paper is divided into three main sections. The first section is a detailed definition of the attributes and capabilities that a potentially universal freeway incident detection framework should possess. The second section discusses the training and testing of the PNN. In the third section, we evaluate the PNN relative to the universality template previously defined. In addition to a large set of simulated incidents, we utilize a fairly large real incident databases from the I-880 freeway in California and the I-35W in Minnesota to comparatively evaluate the performance and transferability of different algorithms, including the PNN. Experimental results indicate that the new PNN-based algorithm is competitive with the Multi Layer Feed Forward (MLF) architecture, which was found in previous studies to yield superior incident detection performance, while being significantly faster to train. In addition, results also point to the possibility of utilizing the real-time learning capability of this new architecture to produce a transferable incident detection algorithm without the need for explicit off-line retraining in the new site. In this respect, and unlike existing algorithms, the PNN has been found to markedly improve in performance with time in service as it retrains itself on captured incident data, verified by the Traffic Management Center (TMC) operator. Moreover, the overall PNN-based framework promises potential enhancements towards the envisioned universality requirements.

Some general results on the optimal ramp control problem

In an effort to relieve peak hour congestion on freeways, various ramp metering algorithms have been employed to regulate the inputs to freeways from entry ramps. In this paper, we consider a freeway system comprised of a freeway section and its entry/exit ramps, and formulate the ramp control problem as a dynamic optimal process to minimize the total time spent in this system. Within this framework, we are able to show when ramp metering is beneficial to the system in terms of total time savings, and when it is not, under the restriction that the controlled freeway has to serve all of its ramp demand, and the traffic flow process follows the rules prescribed by the LWR theory with a triangular flow-density relationship. We also provide solution techniques to the problem and present some preliminary numerical results and empirical validation.

FREEWAY RAMP METERING USING ARTIFICIAL NEURAL NETWORKS

This paper proposes a nonlinear approach for designing local traffic-responsive ramp controls using artificial neural networks. The problem is formulated as a nonlinear feedback control problem, where the system model is the well known hydrodynamic model developed by Lighthill and Whitham (1955), and Richards (1956), the models flow-density relationship is nonlinear, and the feedback nonlinear controllers are composed of one or a number of feed-forward neural networks. These neural network controllers are of integral (I) or proportional-plus-integral (PI) type, and can be tuned on-line to achieve prescribed performance. Initial simulation results show that such an approach is promising.

SIMULATION OF FREEWAY INCIDENT DETECTION USING ARTIFICIAL NEURAL NETWORKS

Abstract A major source of traffic delay in many large urban areas in the United States is non-recurring congestion caused by incidents. In the last several decades, a number of incident detection algorithms have been developed for freeway surveillance and control systems. However, conventional algorithms have generally met with mixed success in terms of performance criteria, such as detection rate, false alarm rate, and the mean time to detect incidents. The need for improved techniques is pressing, particularly with the advent of intelligent vehicle-highway system concepts. These systems will rely heavily on the ability to detect non-recurring traffic congestion automatically. In this paper, we hypothesize that spatial and temporal traffic patterns can be recognized and classified by an artificial neural network, and we present an investigation of such models for the automated detection of lane blocking incidents in a one-mile section of urban freeway. The artificial neural network was trained with data obtained from a microscopic freeway traffic simulation model that was specially calibrated for the actual freeway test section. The neural network first classifies the traffic state of the freeway section into either “incident-free” or “incident” conditions in every 30-second interval. The change in traffic state from incident-free to incident conditions is then used to trigger an incident alarm. Based on the results of an off-line test using simulated data, and comparisons with the well known California incident detection algorithm and the recently developed modified McMaster algorithm, the results suggest that neural network models have the potential to achieve significant improvements in incident-detection performance.

A NEURAL NETWORK-BASED METHODOLOGY FOR PAVEMENT CRACK DETECTION AND CLASSIFICATION

Abstract This paper presents a methodology for automating the processingof highway pavement video images using an integration of artificial neural network models with conventional image-processing techniques. The methodology developed is able to classify pavement surface cracking by the type, severity, and extent of cracks detected in video images. The approach is divided into five major steps: (1) image segmentation, which involves reduction of a raw gray-scale pavement image into a binary image, (2) feature extraction, (3) decomposition of the image into tiles and identification of tiles with cracking, (4) integration of the results from step (3) and classification of the type of cracking in each image, and (5) computation of the severities and extents of cracking detected in each image. In this methodology, artificial neural network models are used in automatic thresholding of the images in stage (1) and in the classification stages (3) and (4). The results obtained in each stage of the process are presented and discussed in this paper. The research results demonstrate the feasibility of this new approach for the detection, classification, and quantification of highway pavement surface cracking.

COORDINATED TRAFFIC-RESPONSIVE RAMP CONTROL VIA NONLINEAR STATE FEEDBACK

In this paper, we develop a coordinated traffic responsive ramp control strategy based on feedback control and artificial neural networks. The proposed feedback control law is nonlinear and realized by a series of neural networks. The parameters of the neural networks are obtained through a nonlinear optimization procedure. Traffic simulations show that the proposed nonlinear ramp control strategy compares favorably against the well-known linear quadratic (LQ) control strategy in reducing total travel times, particularly at situations where drastic changes in traffic demand and road capacity occur.

Macroscopic Modeling of Freeway Traffic Using an Artificial Neural Network

Traffic flow on freeways is a complex process that often is described by a set of highly nonlinear, dynamic equations in the form of a macroscopic traffic flow model. However, some of the existing macroscopic models have been found to exhibit instabilities in their behavior and often do not track real traffic data correctly. On the other hand, microscopic traffic flow models can yield more detailed and accurate representations of traffic flow but are computationally intensive and typically not suitable for real-time implementation. Nevertheless, such implementations are likely to be necessary for development and application of advanced traffic control concepts in intelligent vehicle-highway systems. The development of a multilayer feed-forward artificial neural network model to address the freeway traffic system identification problem is presented. The solution of this problem is viewed as an essential element of an effort to build an improved freeway traffic flow model for the purpose of developing real-time predictive control strategies for dynamic traffic systems. To study the initial feasibility of the proposed neural network approach for traffic system identification, a three-layer feed-forward neural network model has been developed to emulate an improved version of a well-known higher-order continuum traffic model. Simulation results show that the neural network model can capture the traffic dynamics of this model quite closely. Future research will attempt to attain similar levels of performance using real traffic data.