Harvesting-aware Adaptive Energy Management in Solar-Powered Embedded Systems

Nga Dang1, Zana Ghaderi2, Eli Bozorgzadeh3, Moonju Park4
1Google Inc., 2University of California, Irvine, 3Univ. of California, Irvine, 4Incheon National University, South Korea


Solar-powered embedded systems are challenged by variations and uncertainties in energy availability due to seasonal changes and weather conditions. Such systems need to deploy various schemes to adapt their energy consumption in order to achieve energy sustainability. We target solar-powered real-time systems with adaptive QoS model. A holistic middleware framework for energy management that orchestrates DVFS and application QoS in solar-powered real-time systems is proposed. Extensive experiments are carried out to show the effectiveness and the robustness of our technique on two case studies on a solar-powered smart camera system. Simulation results show an improvement of 19%-50% in terms of total QoS compared to a DVFS framework for real-time systems which does not provide adaptive QoS model.