A Predictive Protocol for Mobile Context Updates with Hard Energy Constraints

As mobile devices have become powerful sensor platforms, new applications have emerged which continuously stream mobile user context (location, activities, etc.). However, energy is a limited resource on battery-equipped mobile devices. Especially frequent transmissions of context updates over energy-expensive wireless channels drain the battery of mobile devices in an uncontrolled manner. It is a fundamental algorithmic challenge to design protocols such that users can control the energy consumption on mobile devices while, at the same time, optimizing the quality of mobile applications. To address this trade-off in the area of context update protocols, we propose a novel protocol that maximizes the context accuracy perceived by a remote consumer while guaranteeing that the consumed energy stays under a given limit. Our update protocol exploits predictions about a user´s future behaviour to give priority to the most effective context updates. In our evaluation, we apply our predictive update protocol to a real-world trace of user context and show that the context accuracy is significantly increased compared to an update protocol which operates without predictions under the same energy budget.