SUMO-Cadyts calibration with limited data quality

Abstract

Traffic simulation has been often used to support decision making in traffic management. Such simulation requires a large amount of fine data which cannot be easily obtained in practice especially when a large network is considered. Under this circumstance, the respective calibration work is challenged. In this study, the route choice calibration of SUMO and Cadyts is applied with a real case study in Hefei, China, to examine (1) to which extent the calibration can achieve and (2) which issues need to be especially taken into account. With the resource limitation the network data is based on the OSM with insufficient quality. Major adjustments are made on the number of lanes, turning lane allocations, speed limitations and traffic signal plans. The last two parts still suffer from the limited data quality. The collected flow data for calibration is also limited for only 15 minutes at five major intersections during different moments in the evening peak hour. The relative absolute error is used as the performance indicator. The results show that the simulation quality is improved with the given limited data and quality, but only to a certain degree, namely the percentage of the investigated links with a flow deviation less than 30% has increased from 38% to 72%. The spectrum of the relative errors has been greatly narrowed after the calibration. Gained experiences related to the application of SUMO-Cadyts calibration are also pointed out as references for further applications.