| The performance of wireless mobile devices is lacking in terms of energy consumption and throughput estimation of the cellular and WiFi radios. In this thesis, we first propose a novel bluetooth based mobile-to-mobile collaborative system that smartly mitigates the cellular tail energy wastage among multiple devices while being fair to participating nodes. Our approach is unique as we focus on low-data-rate background traffic instead of application data. Second, based on the notion that background data is delay insensitive, we propose a dynamic algorithm using reinforcement learning that buffers background transfers of multiple applications on single mobile device, to reduce the WiFi tail energy wastage while keeping the buffering delay low. Finally, we propose an accurate throughput estimation technique for TCP Cubic. Compared to existing throughput estimation techniques on mobile devices, our approach has significantly lower data and energy overhead while providing more accurate estimation results. |
