QoE-Aware Dual Control System to Guarantee Battery Lifetime for Mobile Video Applications

• Published in: IEEE transactions on consumer electronics Vol. 65; no. 4; pp. 454 - 463
• Main Authors: Groba, Angel M., Lobo, Pedro J., Chavarrias, Miguel
• Format: Journal Article
• Language: English
• Published: New York IEEE 01-11-2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Batteries; Battery-lifetime guarantee; Control systems; Decoding; dynamic voltage and frequency scaling (DVFS); dynamic-power regulation; Energy conservation; mobile video; Power consumption; Power demand; QoE awareness; Quality of experience; Real-time systems; Streaming media; Switches; User satisfaction
• ISSN: 0098-3063; 1558-4127
• DOI: 10.1109/TCE.2019.2932778

The use of consumer-electronics mobile terminals to decode and view digital video is constantly growing. Consumers expect from them a good quality of experience (QoE) during the complete viewing of the video content, which is not always possible due to the battery-operated nature of this type of devices. In this paper, a dual closed-loop control system is proposed to act on the CPU frequency such that, by default, it controls a QoE-related variable to be at its lowest valid level in order to save as much energy as possible. This energy savings are computed as an energy bonus (EB) that can be further used in moments of higher power consumption. Anyway, when the EB decreases below a dynamic threshold, the system switches to an exception mode in which it controls the EB to be enough to guarantee the expected playback lifetime at the expense of lowering QoE. Besides, the system can be configured by users to be more or less aggressive in risking lifetime vs QoE by adjusting the mode-switching threshold. Upon implementation in the operating system of a commercial development board, results show how the system is able to meet the desired lifetime with a reduction of up to 70% of the time with QoE restrictions with respect to a previous lifetime-guaranteeing approach. Furthermore, the worst-case computation-time overhead of the system is only a 4.2%.