Calibration and Validation of Simulation Models for Investigation of Traffic Assistance Systems

Abstract

In recent years, initial proposals have been presented for advanced driver assistance systems (ADAS) which globally optimize traffic flow by means of the interaction of autonomous vehicles. This kind of ADAS will hereinafter be referred to as traffic assistance system (TAS). For the design, optimization and evaluation of these TAS, investigative simulations simultaneously considering both microscopic and macroscopic behavior are necessary. Therefore, in this paper a two-level approach for calibration and validation of traffic simulations is presented. This contribution presents a new measurement concept that is needed to gather the required data for the suggested two-level approach for calibration and validation. This concept advocates simultaneous data acquisition sourced from both a vehicle (microscopic) and an overall traffic (macroscopic) perspective which is furtheron compared to simulative data of both systemic levels. The paper describes the concept of calibration and validation of a car-following model with respect to intra- und inter-driver-variability, which makes it necessary to consider a distribution for each parameter in use. First empirical parameter distributions and their subsequent use in the Gipp’s car following model are described in this paper. Results of a macroscopic validation with regard to headway distribution are presented. Compared to the current state of the art, the application of the two-level approach for calibration and validation with the gathered microscopic and macroscopic measurement data will enhance the possibilities to investigate the efficiency of TAS and yield results which are characterized by a higher degree of confidence.