A Modeling Framework of Cyber-Physical-Social Systems with Human Behavior Classification Based on Machine Learning

摘要

Cyber-Physical-Social Systems (CPSS) is an emerging complicated topic in recent years which focuses on the researches of a combination of cyberspace, physical space and social space. Different from traditional Cyber-Physical-Systems, CPSS contain human who interacts with the cyber and physical part more frequently. So how to capture and analyse human behaviors play a vital role in CPSS performance evaluation. To improve the analysis accuracy of CPSS, the paper proposes a new modelling framework - stohMCharts (stochastic hybrid MARTE statecharts) which is an extension of MARTE statecharts for stochastic hybrid system modelling and analysis. Compared to MARTE statechart, in stohMCharts, we can model the CPSS in a unified way. Also, we associate stohMCharts to NSHA (Networks Stochastic Hybrid Automata) and use statistical model checker UPPAAL-SMC to verify the stohMCharts. We apply an autonomous car as an example to explain the efficiency of our proposed approaches.