Abstracts 18th annual scientific meeting of the American Uro-Gynecologic society September 25–28 1997 Tucson, Arizona.pdfVIP

Verification and Validation of Agent-Based Simulations Using Approximate Model Checking ( ) Benjamin Herd B , Simon Miles, Peter McBurney, and Michael Luck Department of Informatics, King’s College London, London, UK benjamin.c.herd@kcl.ac.uk Abstract. This paper focusses on the usefulness of approximate prob- abilistic model checking for the internal and external validation of large- scale agent-based simulations. We describe the translation of typical val- idation criteria into a variant of linear time logic. We further present a prototypical version of a highly customisable approximate model checker which we used in a range of experiments to verify properties of large scale models whose complexity prevents them from being amenable to conven- tional explicit or symbolic model checking. Keywords: Agent-based simulation · Verification · Validation · Model checking 1 Introduction Agent-based simulation (ABS) is rapidly emerging as a popular paradigm for the simulation of complex systems that exhibit a significant amount of non-linear and emergent behaviour. It is applied successfully to an ever-increasing number of real-world problems and could in many areas show advantages over tradi- tional numerical and analytical approaches. Due to the high level of complexity, however, ABSs are difficult to understand, to verify and to validate. In order to deal with the large number of behaviours that a model can exhibit, random variance in the output and an often huge input parameter space, comprehensive experiments need to be conducted in all stages of the development process as well as during productive use. Insight into the dynamics is typically obtained by analysing the output (often by executing complex database queries), by statisti- cal analysis and inductive reasoning