Optimisation of networked control systems using model-based safety analysis techniques

We propose a novel approach to the optimization of networked embedded safety critical systems in which Genetic Algorithms are used to find optimal tradeoffs among safety, reliability and cost in the design of such systems. The aim is to automatically evolve initial designs that do not necessarily meet dependability requirements to designs that fulfil such requirements with minimal costs. The approach departs from earlier work in that the safety and reliability model (i.e. a set of system fault trees) is automatically synthesised from an engineering model of the system. It also moves beyond the classical '' success-failure '' model by introducing a failure scheme in which components can exhibit more that one failure modes which include the loss but also the commission of functions as well as value and timing failures. We discuss the approach, and compare the performance of two implementations, based on two different Genetic Algorithms, which have been applied on a set of well known benchmark examples.