Modelling and Verification of Dependable Component-Based Vehicular Control-System Architectures

Domains such as vehicular control system design illustrate the need for component-based development methods based on architectural component models, and formal, design-time verification of extra-functional requirements such as schedulability. However, methods for verification and prediction of component-based systems are still impractical, requiring improved performance and accuracy.We show how using {em dependent finite state machines} (DFSMs) enables practical property prediction, through a case study illustration of an automotive cruise control system, with varying behaviour depending on configuration choices, represented in the SaveComp component model (SaveCCM). Variability is a hallmark of component-based design, in particular of product line architectures in automotive control. The parameterised component protocol types in DFSMs permit analysis of behaviour dependencies and allow refined predictions for improved accuracy in worst-case execution time (WCET) bounds for particular configurations. Since task schedulability critically depends on WCET, schedulability can be predicted more accurately. Many other approaches take a whole-of-system analysis approach, requiring computation of a detailed behavioural model of the entire system. In contrast, hierarchical DFSMs permit propagation of behaviour constraints through networks of mutually dependent state machine behaviour models. The propagation operates on hybrids of protocol state machines and simplified property models -- such as formulae or table representations for WCET properties of state machines. Combining SaveCCM and schedulability analysis with DFSM semantics and dependency analysis allows scalable and accurate analysis of SaveCCM systems and extends the range of compositional extra-functional properties studied and analysed in the context of DFSMs.