Model-based system of systems verification considering selfishness

Abstract

A System of Systems (SoS) is composed of many independent constituent systems (CSs) to achieve complex goals (SoS-level goal), which is not accomplished by a single system. In SoS engineering, analyzing the SoS-level goal achievement is one of the most important problems. There is no common definition of SoS, but it is common that each CS has managerial independence and operational independence. This indicates that CSs in SoS can act in their own way. Furthermore, SoS uses a bottom-up method that includes existing systems as a configuration method. Those characteristics make SoS-level manager hard to know the internal information of each CS. Also, even if CSs act in a predictable way, when a large number of CSs simultaneously behaves and responds to the environment, we cannot avoid uncertainty about the behavior of the SoS. In this thesis, we propose Action-Benefit-Cost modeling approach, which is based on economics and game theory, to model an SoS (ABC-SoS) without detailed internal information of each CS. SoS is modeled using observable behaviors (actions) of each CS. Those actions are composed their interest and consumed resources by acting the actions. Because actions are modeled without internal information, we analyze the behaviors of an SoS using only observable behaviors and assumed decision mechanism, which is beneficial and cost. Also, we propose the verification supporting tool, SIMSoS. The tool contains discrete-time simulator and simulation-based model checking technique. Simulation quantifies uncertainty of SoS by executing ABC-SoS model, so we can get the quantitative SoS-level goal achievement verification result. We evaluate our tool correctness using the mathematically calculable scenario. We use mass casualty incident response system scenario and model the scenario as ABC-SoS model. Using the case scenario, we show the ABC model and SMC can quantitatively analyze SoS-level goal achievement.