Calibrating Steady-State Traffic Stream and Car-Following Models Using Loop Detector Data

The research reported in this paper develops a heuristic automated tool (SPD_CAL) for calibrating steady-state traffic stream and car-following models using loop detector data. The performance of the automated procedure is then compared to off-the-shelf optimization software parameter estimates, including the MINOS and Branch and Reduce Optimization Navigator (BARON) solvers. The model structure and optimization procedure is shown to fit data from different roadway types and traffic regimes (uncongested and congested conditions) with a high quality of fit (within 1% of the optimum objective function). Furthermore, the selected functional form is consistent with multiregime models, without the need to deal with the complexities associated with the selection of regime breakpoints. The heuristic SPD_CAL solver, which is available for free, is demonstrated to perform better than the MINOS and BARON solvers in terms of execution time (at least 10 times faster), computational efficiency (better match to field data), and algorithm robustness (always produces a valid and reasonable solution).