On Model-Based Performance Analysis of Collective Adaptive Systems

This paper fosters the analysis of performance properties of collective adaptive systems (CAS) since such properties are of paramount relevance practically in any application. We compare two recently proposed approaches: the first is based on generalised stochastic petri nets derived from the system specification; the second is based on queueing networks derived from suitable behavioural abstractions. We use a case study based on a scenario involving autonomous robots to discuss the relative merit of the approaches. Our experimental results assess a mean absolute percentage error lower than 4% when comparing model-based performance analysis results derived from two different quantitative abstractions for CAS.