A Multi-agent Framework for Dependable Adaptation of Evolving System Architectures

We present a multi-agent framework for the formal verification of component-based systems after changes such as addition, removal and modification of components. The core of our approach is an Agent Verification Engine (AVE) that constructs evolvable Belief-Desire-Intention (BDI) agents to coordinate and plan the re-verification of component models after system changes. The engine provides BDI-agents with existing techniques for the compositional verification of component-based systems. We illustrate this integration for Satisfiability Modulo Theories (SMT) constraint analysis and demonstrate our framework on requirements arising from industrial control systems.