Abstract
We can reproduce traffic accidents on a traffic simulator by implementing drivers' error in a driving phase such as recognition, determination and operation. Since traffic accidents are the phenomena that cause serious damage but have quite low probability, it is difficult to obtain statistical prediction. Therefore, the approaches to identify the types and regions of the parameters contributing significantly to the occurrence of traffic accidents from a vast parameter space may become more effective. It is important to execute simulations efficiently and to manage the simulation results for the sensitivity analyses. In this paper we describe a framework that has been developed to execute and manage a huge number of parametric studies.
