Vincenzo Punzo

Analysis and Comparison of Microscopic Traffic Flow Models with Real Traffic Microscopic Data

The evermore widespread use of microscopic traffic simulation in the analysis of road systems has refocused attention on submodels, including car-following models. The difficulties of microscopic-level simulation models in the accurate reproduction of real traffic phenomena stem not only from the complexity of calibration and validation operations but also from the structural inadequacies of the submodels themselves. Both of these drawbacks originate from the scant information available on real phenomena because of the difficulty with the gathering of accurate field data. In this study, the use of kinematic differential Global Positioning System instruments allowed the trajectories of four vehicles in a platoon to be accurately monitored under real traffic conditions on both urban and extraurban roads. Some of these data were used to analyze the behaviors of four microscopic traffic flow models that differed greatly in both approach and complexity. The effect of the choice of performance measures on the m...

Integration of Driving and Traffic Simulation: Issues and First Solutions

Driving simulators are very suitable test beds for the evaluation and development of intelligent transportation systems (ITSs). However, the impact of such systems on the behavior of individual drivers can properly be analyzed through driving simulators only if autonomous vehicles in the driving scenario move according to the system under evaluation. This condition means that the simulation of the traffic surrounding the interactive vehicle should already take into account the drivers behavior as affected by the system under analysis. Currently, this “loop” is not properly tackled, because the effects on individuals and traffic are, in general, separately and, often, independently evaluated. The integration of traffic and driving simulations, instead, may provide a more consistent solution to this challenging evaluation problem. It also opens up new scenarios for enhancing the credibility of both traffic modeling and driving simulation and for their combined development. For instance, because drivers directly interact with driver/traffic models in a driving simulation environment, such models may also be tested against nonnormative behavior, and this case seems the only way to test driver/traffic models for safety applications. Based on this idea, this paper describes the integration of a driving simulation engine known as SCANeR and a traffic-flow microsimulation model known as AIMSUN. Methodological and technical issues of such integration are first presented, and future enhancements for higher consistency of the simulation environments are finally envisaged.

Making NGSIM Data Usable for Studies on Traffic Flow Theory: Multistep Method for Vehicle Trajectory Reconstruction

Despite the importance of NGSIM data for research on traffic flow theory, these data proved to be massively affected by measurement errors in the vehicles spatial coordinates, errors that were further amplified in the differentiation process when speed and acceleration values were calculated. If not properly accounted for, these errors would make NGSIM data unusable for any study on traffic flow theory. However, the techniques applied in the literature to correct vehicle trajectory data are not suitable for the scope; these techniques do not treat the cause of the bias appropriately and are limited to smoothing out the effects, which are the high-and medium-frequency disturbances in the data. Therefore, in this study the mechanism that was the root of the NGSIM data errors was illustrated, and the limits of available techniques were shown. Then, clarification that extremely high errors, the outliers, need special treatment to be fixed was provided. A multistep filtering procedure aimed at (a) eliminating outliers giving rise to unphysical values for acceleration by local reconstruction of the vehicle trajectory and (b) cutting off the residual random disturbances from the signal while still preserving the driving dynamics was proposed. Both operations were performed, with the requirement for internal consistency of the trajectory being taken into account. Results related to a single vehicles trajectory from the NGSIM I-80 data set and results from the application to the complete set of trajectories from the same data set are presented. The results necessitated correction of NGSIM data before further processing.

Can Results of Car-Following Model Calibration Based on Trajectory Data Be Trusted?

Calibration of car-following models against trajectory data has been widely applied as the basis for studies ranging from model investigation and benchmarking to parameter correlation analysis. Other theoretical issues, such as inter- and intradriver heterogeneity or multianticipative driving behavior, are also addressed in such studies. However, very few of these studies attempted to analyze and quantify the uncertainty entailed in the calibration process and its impacts on the accuracy and reliability of results. A thorough understanding of the whole calibration problem (against trajectory data), as well as of the mutual effect of the specific problems raised in the field literature, does not yet exist. In this view, a general methodology to assess a calibration procedure was proposed and applied to the calibration of the Gipps’ car-following model. Compact indicators were proposed to evaluate the capability of a calibration setting to find the known global solution regarding the accuracy and the robustness against the variation of the starting conditions of the optimization algorithm. Then a graphical inspection method, based on cobweb plots, was proposed to explore the existence and nature of the local minima found by the algorithms, as well as to give insights into the measures of performance and the goodness-of-fit functions used in the calibration experiments. The methodology was applied to all calibration settings (i.e., combinations of algorithms, measures of performance, and goodness-of-fit functions) used in the field literature so far. The study allowed the highlighting and motivation, for the model under investigation, of the limits of some of these calibration settings. Research directions for the definition of robust settings for the problem of car-following model calibration based on real trajectory data are outlined.

Comparison of Simulation-Based and Model-Based Calibrations of Traffic-Flow Microsimulation Models

Parameter calibration of traffic microsimulation models usually takes the form of a simulation-based optimization problem, that is, an optimization in which every objective function evaluation calls for a simulation. It is recognized that such a problem is computationally intractable. Running time grows exponentially both in the number of parameters and in the digits accuracy. In addition, considerable computing time is required by each objective function evaluation. This means that only heuristic techniques can be applied. Accordingly, results of the application of the OptQuest/Multistart algorithm to the calibration of AIMSUN microsimulation model parameters on a freeway network are presented. Furthermore, it is claimed that the search for an effective solution to the calibration problem cannot be exhausted by the choice of the most efficient optimization algorithm. The use of available information concerning the phenomenon could allow calibration performance to be enhanced, for example, by reducing dimensions of the domain of feasible solutions. It is argued that this goal could be achieved by using information from the stationary counterpart of microscopic traffic-flow models that depict the aggregate variables of traffic flows as a function of drivers’ microscopic parameters. Because they have a closed analytical formulation, they are well suited for faster calibrations. Results show that values of parameters from stationary model-based calibrations are not far from the optimal ones. Thus the integration of the two approaches cannot be excluded but is worth investigating.

Steady-State Solutions and Multiclass Calibration of Gipps Microscopic Traffic Flow Model

This study addresses the problem of calibration of the Gipps microscopic traffic flow model. The approach consists of first deriving traffic stream models, in the form of steady-state solutions of car-following models, and then fitting such models to stationary traffic data. To this end, traffic stream models for the Gipps model were first attained, and an explicit formula for the flow at capacity as a function of microscopic parameters is provided. Analysis of the models for different combinations of microscopic parameters explains the widely held belief that the Gipps model is unable to reproduce unstable traffic phenomena. To be suitable for model parameter calibration in simulation practice, single-class models were generalized to a multiclass traffic scenario for which a calibration procedure was developed. Once applied to real motorway traffic data, the multiclass scenario proved its effectiveness in terms of error statistics. Values of calibrated parameters were all significant and consistent with expectations. Moreover, they were consistent with the observed aggregate measures (e.g., flow at capacity). Finally, unlike nonstationary, model-based approaches, the computing time required by the multiclass calibration presented is negligible, allowing calibration of a large number of parameters, that is, calibration of different classes of vehicles.

Do We Really Need to Calibrate All the Parameters? Variance-Based Sensitivity Analysis to Simplify Microscopic Traffic Flow Models

Automated calibration of microscopic traffic flow models is all but simple for a number of reasons, including the computational complexity of black-box optimization and the asymmetric importance of parameters in influencing model performances. The main objective of this paper is therefore to provide a robust methodology to simplify car-following models, that is, to reduce the number of parameters (to calibrate) without sensibly affecting the capability of reproducing reality. To this aim, variance-based sensitivity analysis is proposed and formulated in a “factor fixing” setting. Among the novel contributions are a robust design of the Monte Carlo framework that also includes, as an analysis factor, the main nonparametric input of car-following models, i.e., the leaders trajectory, and a set of criteria for “data assimilation” in car-following models. The methodology was applied to the intelligent driver model (IDM) and to all the trajectories in the “reconstructed” Next Generation SIMulation (NGSIM) I80-1 data set. The analysis unveiled that the leaders trajectory is considerably more important than the parameters in affecting the variability of model performances. Sensitivity analysis also returned the importance ranking of the IDM parameters. Basing on this, a simplified model version with three (out of six) parameters is proposed. After calibrations, the full model and the simplified model show comparable performances, in face of a sensibly faster convergence of the simplified version.

“No Free Lunch” Theorems Applied to the Calibration of Traffic Simulation Models

In 1997, Wolpert and Macready derived “No free lunch theorems for optimization.” They basically state that “the expected performance of any pair of optimization algorithms across all possible problems is identical.” This is to say that there is no algorithm that outperforms the others over the entire domain of problems. In other words, the choice of the most appropriate algorithm depends upon the specific problem under investigation, and a certain algorithm, while providing good performance (both in terms of solution quality and convergence speed) on certain problems, may reveal weak on certain others. This apparently straightforward concept is not always acknowledged by optimization practitioners. A typical example, in the field of traffic simulation, concerns the calibration of traffic models. In this paper, a general method for verifying the robustness of a calibration procedure (suitable, in general, for any simulation optimization) is proposed based on a test with synthetic data. The main obstacle to this methodology is the significant computation time required by all the necessary simulations. For this reason, a Kriging approximation of the simulation model is proposed instead. The methodology is tested on a specific case study, where the effect on the optimization problem of different combinations of parameters, optimization algorithms, measures of goodness of fit, and levels of noise in the data is also investigated. Results show the clear dependence between the performance of a calibration procedure and the case study under analysis and ascertain the need for global solutions in simulation optimization with traffic models.

Nonstationary Kalman filter for estimation of accurate and consistent car-following data

Difficulty in obtaining accurate car-following data has traditionally been regarded as a considerable drawback in understanding real phenomena and has affected the development and validation of traffic microsimulation models. Recent advancements in digital technology have opened up new horizons in the conduct of research in this field. Despite the high degrees of precision of these techniques, estimation of time series data of speeds and accelerations from positions with the required accuracy is still a demanding task. The core of the problem is filtering the noisy trajectory data for each vehicle without altering platoon data consistency; i.e., the speeds and accelerations of following vehicles must be estimated so that the resulting intervehicle spacings are equal to the real one. Otherwise, negative spacings can also easily occur. The task was achieved in this study by considering vehicles of a platoon as a sole dynamic system and reducing several estimation problems to a single consistent one. This process was accomplished by means of a nonstationary Kalman filter that used measurements and time-varying error information from differential Global Positioning System devices. The Kalman filter was fruitfully applied here to estimation of the speed of the whole platoon by including intervehicle spacings as additional measurements (assumed to be reference measurements). The closed solution of an optimization problem that ensures strict observation of the true intervehicle spacings concludes the estimation process. The stationary counterpart of the devised filter is suitable for application to position data, regardless of the data collection technique used, e.g., video cameras.

Human-Like Adaptive Cruise Control Systems through a Learning Machine Approach

In this work an Adaptive Cruise Control (ACC) model, with human-like driving capabilities,based on a learning machine approach, is proposed. The system is based on a neural network approach and is intended to assist the drivers in safe car-following conditions. The proposed approach allows for an extreme flexibility of the ACC that can be continuously trained by drivers in order to accommodate their actual driving preferences as these changes among drivers and over time. The model has been calibrated against accurate experimental data consisting in trajectories of vehicle platoons gathered on urban roads. Its performances have been compared with those of a conventional car-following model.