No abstract

The purpose of this paper is to explore various traffic modeling aspects and theories that may overcome some of the limitations in existing microscopic simulation models. A multiregime microscopic traffic simulation approach has been formulated featuring realistic and comprehensive carfollowing and lane-changing logic. A prototype implementation of the multiregime approach was developed in C++ and extensively tested. The multiregime simulation results demonstrate the efficiency and validity of the proposed models for a broad range of traffic scenarios. The test and validation results indicate that the model and program outperformed traditional methods and other existing traffic simulation programs. The validity and efficiency of the model is attributed to the fact that the regimes were added to the model incrementally to reflect increasing agreement with real-world traffic flow. The techniques and corresponding models will be used to improve existing microscopic traffic simulation models and programs.

To achieve high fidelity and high credibility of a microscopic traffic simulation, model validation is of the utmost importance. A systematic validation effort was performed on an advanced microscopic traffic simulation that uses a multiregime simulation (MRS) approach for its car-following and lane-changing logic. The vigorous validation procedure includes animation comparison and quantitative/statistical analysis at both macroscopic and microscopic levels, and is illustrated through case studies. For validation at the macroscopic level, the averages and other statistics of traffic variables are compared, and fundamental relationships of traffic flow parameters are studied. For validation at the microscopic level, the speed change pattern, vehicle trajectory plots, and headway distribution from simulations are compared with those of the real-world system. The importance of the real-world data sets to model validation is emphasized, and animation comparison that provides intuitive appreciation of the model validity is frequently used. The validation and testing not only show the validity of the MRS model but also illustrate a systematic approach that can be used to validate other simulation models.

This paper discusses the concept of validation and proposes a multistage validation framework for traffic simulation models. The framework consists of conceptual validation and operational validation. The operational validation involves two levels of statistical tests: a twosample t test and a two-dimensional two-sample Kolmogorov-Smirnov test. The validation experience employing the proposed framework demonstrates the fact that while a model can be valid for one level of detail, it can be invalid for another. The validation results also illustrate that the proposed multistage validation procedure can account for the complexity of the validation task and its conclusions.

A multistage validation framework that accounts for the realistic nature of traffic simulation output data is proposed. The framework consists of conceptual validation and operational validation. The operational validation comprises a qualitative approach, which involves static and animated Turing tests, and a quantitative approach, which involves three levels of statistical tests. Particularly in the third-level statistical test, the autocorrelation and nonstationary nature of traffic simulation output data is emphasized, its implications on validation methods are explored, and a univariate nonseasonal autoregressive-integrated-moving-average (ARIMA) modeling approach is proposed. Finally, numerical results for an actual freeway network are presented. The validation results illustrate that the proposed multistage validation procedure can account for the complexity of the validation task and its conclusions.