Supervised Learning Grant Prediction for Cellular Mobile Device Power Savings
Buch in deiner Nähe kaufen
Beschreibung
Optimizing power consumption in mobile devices is crucial to reduce their considerable environmental footprint and improve the user experience. Such mobile devices are used on a daily basis by billions of users for tasks like web browsing, communication or video streaming. These connectivity features are supported by a critical component, called the modem, which is responsible for maintaining a constant and reliable connection with the network at a minimal battery energy drain cost. In this work, a novel power saving mechanism is proposed relying on a cognitive system that proactively predicts time intervals with no data transfer and manages the energy drain accordingly. If no data is expected to be received for future time intervals, low power mode is activated, and parts of the modem’s receive chain are deactivated. This mechanism, called cognitive power control, has the advantage of being compliant with existing standard power-saving procedures and can therefore be built on top of current architectures and protocols. In essence, the proposed cognitive system is designed using supervised learning-based algorithms trained with real-life modem traffic patterns. The problem of cognitive power control is formalized as a multi-objective optimization problem in order to study the trade-off between power consumption, data loss and cognitive system complexity. Using theoretical foundations of multi-objective optimization, problem scalarizations and sub-problems are explored as guided investigations in several subsets of efficient solutions corresponding to different starting assumptions, e.g., partially random traffic pattern assumptions, a generic class of supervised learning algorithms, or specific user objective preferences.
Autor*in
Jonathan Ah Sue
Themen in »Supervised Learning Grant Prediction for Cellular Mobile Device Power Savings«
Machine learning Power savings Cellular mobile device