SYSTEM AND METHOD FOR AUTOMATED MODEL CALIBRATION, SENSITIVITY ANALYSIS, AND OPTIMIZATION

摘要

The problem of model calibration is solved with an architecture for automatically managing interchangeable input parameters, interchangeable output objective functions, and interchangeable optimization methods. Prior to calibration, this architecture allows a user to quickly and easily eliminate the vast majority of input values and combinations, thus drastically simplifying the process of calibration (via simulation-based optimization). Embodiments of the subject invention can be used to provide an efficient and practical self-calibration method for computer models, having any number of input and output parameters. Users can adjust the selection of parameters to affect, and in some cases fully control, required computer run times for automated calibration. Beyond calibration, the same architecture can be effectively applied towards problems of sensitivity analysis and optimization. The subject invention has been successfully implemented in the context of traffic simulation