Towards Power Consumption Optimization for Embedded Systems from a Model-driven Software Development Perspective

A power consumption optimization for battery-powered and resourceconstrained embedded systems is typically performed on the hardware layer while the application layer is often neglected. Because software applications affect the hardware behavior directly, power-related optimizations can result in major application design and workflow changes. Such in-depth modifications should be considered in early design phases, where they are most effective. For embedded software development, current trends in software engineering such as Model-Driven Development (MDD) can be used for an early power consumption analysis and optimization even if the hardware platform is not yet finalized. However, power consumption aspects on the application layer are currently not sufficiently considered in MDD. In this paper, we present an approach to abstract hardware components of an embedded system using the Unified Modeling Language (UML) and annotate UML-based models with power characteristics. Additionally, we define a novel UML profile to capture the dynamic behavior of hardware components while interacting with software applications. With our approach, energy profiles can be derived to make the impact of software on power consumption in early design stages visible. Energy profiles are also suitable for software optimization and energy bug detection, which is demonstrated using a sensor node use case example.