Multi-objective calibration of traffic microsimulation models

Abstract

Use of traffic microsimulation for emission estimation and traffic analysis has increased over the last decade, increasing the need for a more detailed calibration. This paper calibrates the Toronto Waterfront Area microsimulation network three times based on different objective functions. A genetic algorithm with a multi-criteria objective function is used to minimize the Root-Mean-Square of Errors between model estimates and field measurements for counts, speeds, and standard deviation of acceleration. The terms for speed and standard deviation of acceleration are then removed in turn from the objective function, and the model is recalibrated. The three calibrated models are compared according to their calibrated parameters, their simulated driving cycles and driving cycle parameters, their Vehicle-Specific Power distribution, and estimated emission factors. The comparison shows that aggressive driving decreases and emission factors are closer for the network that is calibrated to all three goodness-of-fit measures.

Keywords:

• Microsimulation
• multi-objective optimization
• driving cycles
• vehicle-specific power
• MOVES
• genetic algorithm
• emissions
• Toronto