VISSIM Microscopic Traffic Simulator: A systematic Sensitivity Analysis of driving behavior models parameters from macroscopic and mesoscopic perspectives

Abstract

This paper illustrates a systematic sensitivity analysis experiment of the driving behavior models’ parameters in VISSIM. The aim is to explore their impact on selected macroscopic (average travel time and number of lane changes) and mesoscopic (distribution of lane change locations along the road) outputs. The experiment starts by a one-at-a-time screening technique to determine the initial key parameters set and their reliable ranges and effective variation steps. Next, the elementary effects screening technique uses the first round outputs for verification and to reveal the interaction between these parameters. Out of 22 parameters related to car-following and lane changing models in VISSIM, 15 parameters showed various degrees of impact on the proposed macroscopic sensitivity measures. At the mesoscopic level, Maximum Look Ahead Distance, Safety Distance Reduction Factor, CC0, CC1, CC3, and CC4, and Lane Changing Distance were significant for free and discretionary lane changing maneuvers respectively. Although the same set of 15 parameters affects all sensitivity measures, their relative importance differs among those measures and there was no specific trend or correlation between their values and the magnitude of their influence.