Abstract
Advances in parallel simulation capabilities are now enabling the possibility of simulating multiple scenarios of large problem configurations. In emergency management applications, for example, it is now conceivable to consider simulating phenomena in large (city- or state-scale) vehicular networks. However, an informed understanding of simulation results is needed for real-time decision support tools that make use a number of simulation runs. Of special interest are insights into trade-offs between accuracy and confidence bounds of simulation results, such as in the quality of predicted evacuation time in emergencies. In some of our emergency management projects, we are exploring approaches that not only aid in making statistically significant interpretations of simulation results but also provide a basis for presenting the inherent qualitative properties of the results to the decision makers. We provide experimental results that demonstrate the possibility of applying our approach to large-scale vehicular network simulations for emergency planning and management.