Henk J. van Zuylen

The most likely trip matrix estimated from traffic counts

For a large number of applications conventional methods for estimating an origin destination matrix become too expensive to use. Two models, based on information minimisation and entropy maximisation principles, have been developed by the authors to estimate an O-D matrix from traffic counts. The models assume knowledge of the paths followed by the vehicles over the network. The models then use the traffic counts to estimate the most likely O-D matrix consistent with the link volumes available and any prior information about the trip matrix. Both models can be used to update and improve a previous O-D matrix. An algorithm to find a solution to the model is then described. The models have been tested with artificial data and performed reasonably well. Further research is being carried out to validate the models with real data.

Construct support vector machine ensemble to detect traffic incident

This study presents the applicability of support vector machine (SVM) ensemble for traffic incident detection. The SVM has been proposed to solve the problem of traffic incident detection, because it is adapted to produce a nonlinear classifier with maximum generality, and it has exhibited good performance as neural networks. However, the classification result of the practically implemented SVM depends on the choosing of kernel function and parameters. To avoid the burden of choosing kernel functions and tuning the parameters, furthermore, to improve the limited classification performance of the real SVM, and enhance the detection performance, we propose to use the SVM ensembles to detect incident. In addition, we also propose a new aggregation method to combine SVM classifiers based on certainty. Moreover, we proposed a reasonable hybrid performance index (PI) to evaluate the performance of SVM ensemble for detecting incident by combining the common criteria, detection rate (DR), false alarm rate (FAR), mean time to detection (MTTD), and classification rate (CR). Several SVM ensembles have been developed based on bagging, boosting and cross-validation committees with different combining approaches, and the SVM ensemble has been tested on one real data collected at the I-880 Freeway in California. The experimental results show that the SVM ensembles outperform a single SVM based AID in terms of DR, FAR, MTTD, CR and PI. We used one non-parametric test, the Wilcoxon signed ranks test, to make a comparison among six combining schemes. Our proposed combining method performs as well as majority vote and weighted vote. Finally, we also investigated the influence of the size of ensemble on detection performance.

Predicting Urban Arterial Travel Time with State-Space Neural Networks and Kalman Filters

A hybrid model for predicting urban arterial travel time on the basis of so-called state-space neural networks (SSNNs) and the extended Kalman filter (EKF) is presented. Previous research demonstrated that SSNNs can address complex nonlinear spatiotemporal problems. However, SSNN models require off-line training with large sets of input-output data, presenting three main drawbacks: (a) great amounts of time and effort are involved in collecting, preparing, and executing these training sessions; (b) as the input-output mapping changes over time, the model requires complete retraining; and (c) if a different input set becomes available (e.g., from inductive loops) and the input-output mapping has to be changed, then retraining the model is impossible until enough time has passed to compose a representative training data set. To improve SSNN effectiveness, the EKF is proposed to train the SSNN instead of conventional approaches. Moreover, this network topology is derived from the urban travel time prediction...

Capacity Reduction at Incidents: Empirical Data Collected from a Helicopter

Incidents on freeways cause large delays for road users. These delays depend largely on the capacity at the incident location, which is determined by the drivers’ behavior at the accident location. Few empirical facts are available on traffic operations during an incident. This paper presents high-quality videos of the traffic flow around two accidents recorded from a helicopter. From the collected images, traffic counts have been performed at the exact location of the incident. This has two advantages. First, the capacity at the bottleneck per lane could be estimated. Second, truck counts could be converted to passenger car units at the location of the bottleneck. Counts show that the (outflow) capacity of the remaining lanes is about 50% lower than the (free-flow) capacity of the same number of lanes. This means that the road capacity in the opposite direction is reduced by half by the rubbernecking effect. The capacity of the road in the direction of the accident is reduced by more than half because not all lanes are in use. The images provide information on the causes for the capacity reduction. A leader accelerates and the follower accelerates a short time later. The average time between these two accelerations is estimated at about 3 s, but the video also shows a large spread of these times. The results can be used to assess consequences of incidents, in an analytical way and in macroscopic or microscopic traffic simulators.

A comparison of outlier detection algorithms for ITS data

In order to improve the veracity and reliability of a traffic model built, or to extract important and valuable information from collected traffic data, the technique of outlier mining has been introduced into the traffic engineering domain for detecting and analyzing the outliers in traffic data sets. Three typical outlier algorithms, respectively the statistics-based approach, the distance-based approach, and the density-based local outlier approach, are described with respect to the principle, the characteristics and the time complexity of the algorithms. A comparison among the three algorithms is made through application to intelligent transportation systems (ITS). Two traffic data sets with different dimensions have been used in our experiments carried out, one is travel time data, and the other is traffic flow data. We conducted a number of experiments to recognize outliers hidden in the data sets before building the travel time prediction model and the traffic flow foundation diagram. In addition, some artificial generated outliers are introduced into the traffic flow data to see how well the different algorithms detect them. Three strategies-based on ensemble learning, partition and average LOF have been proposed to develop a better outlier recognizer. The experimental results reveal that these methods of outlier mining are feasible and valid to detect outliers in traffic data sets, and have a good potential for use in the domain of traffic engineering. The comparison and analysis presented in this paper are expected to provide some insights to practitioners who plan to use outlier mining for ITS data.

Uncertainty and Predictability of Urban Link Travel Time: Delay Distribution–Based Analysis

Travel times that vehicles experience in urban road networks are intrinsically uncertain because of the stochastic character of delays at signalized intersections. The ability to capture delay characteristics at signalized intersections is critical for estimating and predicting travel times on urban links. Much research has been done to predict travel time on urban links on the basis of traffic state information (e.g., volumes or speeds). However, the results have not been promising. One important reason is that delays experienced by vehicles on urban links are uncertain because of both traffic conditions and traffic control at intersections. This paper addresses the causes of travel time uncertainty. A probabilistic delay distribution model with stochastic arrivals and departures is proposed to investigate delay uncertainty in both undersaturated and oversaturated conditions. The delay distributions with the Poisson arrival process and with the binomial arrival process are compared. Results show that different arrival patterns have little influence on the delay distribution in undersaturated conditions, although they have significant influence on the delay distribution in oversaturated conditions. The delay distributions under different degrees of saturation are overlapping, which indicates that it is difficult to determine the traffic state for a given single-valued delay and vice versa. The “width” of the delay distribution based on percentiles is used to quantify the delay uncertainty at signalized intersections. The dynamics of delay uncertainty and of delay uncertainty under different degrees of saturation are investigated.

State Space Neural Networks for Freeway Travel Time Prediction

The highly non-linear characteristics of the freeway travel time prediction problem require a modeling approach that is capable of dealing with complex non-linear spatio-temporal relationships between the observable traffic quantities. Based on a state-space formulation of the travel time prediction problem, we derived a recurrent state-space neural network (SSNN) topology. The SSNN model is capable of accurately predicting experienced travel times - outperforming current practice by far - producing approximately zero mean normally distributed residuals, generally not outside a range of 10% of the real expected travel times. Furthermore, analyses of the internal states and the weight configurations revealed that the SSNN developed an internal models closely related to the underlying traffic processes. This allowed us to rationally eliminate the insignificant parameters, resulting in a Reduced SSNN topology, with just 63 adjustable weights, yielding a 72% reduction in model-size, without loss of predictive performance.

A hybrid model of partial least squares and neural network for traffic incident detection

Development of a universal freeway incident detection algorithm is a task that remains unfulfilled and many promising approaches have been recently explored. The partial least squares (PLS) method and artificial neural network (NN) were found in previous studies to yield superior incident detection performance. In this article, a hybrid model which combines PLS and NN is developed to detect automatically traffic incident. A real traffic data set collected from motorways A12 in the Netherlands is presented to illustrate such an approach. Data cleansing has been introduced to preprocess traffic data sets to improve the data quality in order to increase the veracity and reliability of incident model. The detection performance is evaluated by the common criteria including detection rate, false alarm rate, mean time to detection, classification rate and the area under the curve (AUC) of the receiver operating characteristic. Computational results indicate that the hybrid approach is capable of increasing detection performance comparing to PLS, and simplifying the NN structure for incident detection. The hybrid model is a promising alternative to the usual PLS or NN for incident detection.

Using Probe Vehicle Data for Traffic State Estimation in Signalized Urban Networks

Probe Vehicle Data (PVD) is becoming more and more common for the collection of information about the traffic state. In most cases, the information that can be obtained from a probe vehicle refers to the position, the speed and the direction of movement at certain time intervals. Especially in urban networks, the raw GPS data needs a cleaning process to map the measured position to the road network. The cleaned information about positions in the network at fixed moments where GPS signals are collected can be used to derive travel time along certain routes.

Valuation of Different Types of Travel Time Reliability in Route Choice: Large-Scale Laboratory Experiment

Travelers sometimes experience extremely long travel times on a route. The Travel Simulator Laboratory (TSL) of Delft University of Technology is used to study the effect of these extreme experiences on route choice. The TSL allows for a completely controlled experiment and good research methodology. It is hypothesized that together with the average travel time and the variance, the most extreme travel times experienced influence a persons perception of the attractiveness of the route. Travel information is assumed to be able to alter this perception. Data from 2,500 respondents were gathered with the TSL. Route switching behavior after a regular experience and after an extreme experience was analyzed. People switched routes after an extreme experience significantly more often when the information that they received before the route choice was wrong than they did when it was correct. Moreover, travelers clearly preferred a route that is sometimes bad and most of the time good over a route that is symmetr...