Avionics Self-adaptive Software: Towards Formal Verification and Validation

One of the future trends in the aerospace industry for ground and air operations is to make aircrafts self-adaptive, enabling them to take decisions without relying on any control authority. We propose a Belief, Desire, Intention (BDI) based multi-agent system for modelling avionics Self-Adaptive Software (SAS). Our BDI models are formally specified using Z notation and include a library of learning algorithms to cater to adaptability. Apart from satisfying various self-* properties that define adaptability features, avionics SAS, being safety critical systems, also have to satisfy safety and provide deterministic response meeting real-time constraints. We propose a validation framework to check for self-* properties. We also present a formal verification framework based on abstractions and model checking for verifying safety properties. The framework is illustrated through an avionics case study involving an adaptive flight planning system.