Self-adaptation in Automotive Embedded Systems using a Multi-layered Control Approach

Abstract

In this work, we present an approach for self-adaptation in automotive embedded systems using a hierarchical, multi-layered control approach. We model automotive systems as a set of constraints and define a hierarchy of control loops based on different criteria. Adaptations are performed at first locally on a lower layer of the architecture. If this fails due to the restricted scope of the control cycle, the next higher layer is in charge of finding a suitable adaptation. We compare different options regarding responsibility split in multi-layered control and a version with centralized control option, in a self-healing scenario with a setup adopted from automotive in-vehicle networks. We show that a multi-layer control architecture has clear performance benefits over a central control, even though all layers work on the same set of constraints. Furthermore, we show that a responsibility split w.r.t. network topology is preferable over a functional split.