Calibrating microscopic traffic simulation models using speed-flow data

Calibrating a micro-simulation model is an important step that precedes implementing the model into engineering practice or academic research. Compared with other measures of effectiveness (MOE), such as traffic delay or volume, using speed-flow data as the MOE has the following advantages: (1) The speed-flow data can be easily requested from local Traffic Management Centre, and thus, financial expenditure associated with data collecting can be reduced; (2) The speed-flow plot contains traffic flow information related to different traffic states and a higher number of parameters can be exposed to the calibration process. However, there are still some issues related to calibrating micro-simulation model using speed-flow data: (1) How to determine a reasonable raster resolution of the speed-flow plot; (2) Bias resulted from using the area of the cells occupied by speed-flow data as the MOE; (3) How to set the free-flow speed (FFS) distribution in the simulation model. In order to address these issues, the relationship between the raster resolution and the number of the cells occupied by speed-flow points and the area of these cells was analysed. A trade-off was observed and was used to determine a reasonable raster resolution. Data frequency information was incorporated into the MOE to refine the original 2-D MOE. In addition, four scenarios were designed in this study to identify the impact of various FFS distribution setting criteria. To verify the effectiveness of the proposed methodology, a case study was conducted using the micro-simulation software package VISSIM and MATLAB. The result manifests the effectiveness of the proposed methods.