IEEE International Conference on Sensing, Communication, and Networking
Abstract
Computation offloading is a promising application of mobile clouds that can save energy of mobile devices via optimal transmission scheduling of mobile-to-cloud task offloading. Existing approaches to computation offloading have addressed various aspects of the tradeoff between energy consumption and application latency, but none of them explicitly considered the dependency in optimization on the mobile user''s context, e.g., user tendency, the remaining battery level. This paper captures such a user-centric perspective in the energy-latency tradeoff via a quality-of-experience (QoE) based cost function, and formulates the problem of data offloading scheduling as dynamic programming (DP). To derive the optimal schedule, we first introduce a database-assisted optimal DP algorithm and then propose a suboptimal but computationally-efficient approximate DP (ADP) algorithm based on the limited lookahead technique. An extensive numerical analysis has revealed that the ADP algorithm achieves near-optimal performance incurring only 2.27% extra cost on average than the optimum, and enhances QoE by up to 4.46 times compared to the energy-only scheduling.