Incorporating adaptivity using learning in avionics self adaptive software: A case study

Self Adaptive software is forging its way into avionics. Such software, while being adaptive, needs to meet safety, determinism, and real time responsiveness, like all avionics systems. We model avionics self adaptive software as a multiagent system. Each agent uses a BDI (Belief Desire Intention) model for adaptiveness and also incorporates learning to address several constraints. We illustrate our approach using a case study of adaptive Flight Management System (FMS). Our BDI model of adaptive FMS in Netlogo is a model that is adaptive while being deterministic and also responds in real-time. We propose a learning algorithm that agents use to adapt themselves better and also a way of measuring their adaptivity that provides quantitative gains illustrating the system's adaptability.