Probabilistic Analysis and Predictions of Component-Based Real-Time Systems

Software components are suitable vehicles to introduce advanced analysis techniques in a software-engineering context for embedded control software; a feat that has yet to be fully accomplished. We are adopting a component-based approach to con-trol software development. We are extending and combining methods from disparate disciplines, such as probabilistic re-liability predictions, stochastic scheduling analysis and soft-ware component technologies. We study theories and meth-ods for probabilistic modelling, analysis, and prediction of control software executing in resource-constrained embed-ded computers. Combining the behaviour models and the architectural model of a component assembly we are deriving stochastic properties, such as reliability, expected delays, and resource consumption. Using components as the fundamental unit of reuse, we employ run-time monitoring techniques to extract probabilistic models of the component behaviour.