Genetic algorithm-based calibration of microscopic traffic simulation model for single-lane roundabouts

Abstract

A calibration procedure for microscopic simulation models based on a genetic algorithm is proposed. Focus is made on single-lane roundabouts for which many random factors such as gap- acceptance affect operations. A comparison is performed between the capacity functions based on a meta- analytic estimation of critical and follow up headways and simulation outputs of a roundabout built in Aimsun microscopic simulator. Aimsun parameters were optimized using the genetic algorithm tool in MATLAB® which automatically interacted with Aimsun through a Python interface. Results showed that applying the genetic algorithm in the calibration process of the microscopic simulation model, a good match to the capacity functions was reached with the optimization parameters set. By this way, automa- tion of the calibration process results effective for analysts which use traffic microsimulation for real world case studies in the professional sphere.