CONTEXTO: Leveraging Energy Awareness in the Development of Context-Aware Applications

We introduce a new context classification and recognition framework for the development and deployment of mobile, context-aware applications. The framework is complemented with an energy calculator that specifically assists mobile developers in estimating the energy footprint of context-aware applications during the development process with the framework. The framework abstracts from the raw context information gathering, allows for sensor fusion, enables the prediction of custom and higher-level contexts, and provides for context sharing. 1. Introduction and Motivation The evolution of mobile devices and the general availability of information sources that describe the situation and environment (i.e., the context) of mobile users offer new opportunities for innovative applications (1). By constantly monitoring the contexts in which mobile users are situated, applications obtain a potential to adapt their behaviour to current contexts more intelligently and without user intervention. However, such mobile context awareness comes at a price: Novel challenges of the mobile environment and specific constraints of mobile devices and their use (e.g., limited battery life, a comparably small screen size, dependence on network infrastructure) can severely impact the acceptance of mobile context-based approaches. In addition, adequate developer support for the realisation of context-aware applications is currently lacking. Consequently, most application developers are on their own when realising the sensing and interpreting of context information, or the sharing of context. With the increasing interest in, and a growing market for, context-aware applications, developers are more and more in charge of carefully designing context-aware applications and they need to be able to competently address issues such as privacy (2), availability, precision of context recognition, or energy requirements. In this contribution, we address the energy-related implications of developers' choices of sensing components, processing algorithms, and granularity or temporal frequency of sensing. We specifically aim at developer energy awareness and present CONTEXTO, an energy-aware framework for offline context classification and recognition on mobile devices. The framework provides a layered, component-based architecture that can easily be extended, modified, or customised. It follows established software engineering patterns to provide high learnability and a low threshold for beginners. Within the framework, the energy requirements for all used components on a specific device are always made transparent, and information about energy requirements can be used early in the design process with the help of the framework's energy calculator, and at runtime.