Variability modeling in automotive embedded systems
Daniel Hilton · 2014
To satisfy the needs of the highly competitive and customer oriented automotive market most manufacturers target the specific needs of their prospective customers by creating a highly flexible product line. Over the past two decades the complexity of automotive vehicles has grown rapidly [1]. As vehicle functionality has increased, so has the amount of possible configurations that a vehicle can have. This is handled by introducing variation points in the embedded software in order to adapt its use to the different configurations. However this introduction is often given incidental treatment by developers which leads the system architecture becoming increasingly inconsistent and highly complex over time. As a consequence of this the impact of a vehicle configuration becomes extremely time-consuming and hard to evaluate. To be able to meet the increasing demands on functional safety in the automotive embedded systems the impact of this variability must be known and evaluated. This thesis aims to explore the possibility of recovering a model which describes the software variability within Scania’s embedded systems. The method that is used involves the definition of a model through using the information and data that is retrieved from the process of architecture recovery. This results in a model that provides an architectural overview of where the vehicle configuration choices affect the software of the embedded systems within Scania. The conclusion is that although the recovered model provides information about the variability there are large limitations when recovering variability from legacy systems through architecture recovery.