Energy-Efficient Task Offloading and Resource Allocation in Mobile Edge Computing with Sequential Task Dependency

In this paper, we investigate the computation task with its sub-tasks subjected to sequential dependency for a mobile edge computing (MEC) system, under both slow fading channel and fast fading channel. To minimize mobile device's energy consumption while limiting task processing delay, offloading strategy (which decides to offload since which sub-task), communication resource (in terms of transmit power in every fading block), and computation resource (in terms of CPU frequency for local computing) are optimized jointly. In slow fading channel, channel state always keeps stable. With offloading decision given, the optimization of the rest variables is non-convex but is transformed to be convex. Golden search method is only required to find the optimal solution by decomposing the investigated problem into two levels. Then the optimal offloading decision can be easily selected. In fast fading channel, channel state may fluctuate even when offloading data. Online policy depending on instant channel state is desired and the optimal solution is derived. In addition, it is proved the derived online policy will converge to an offline policy when channel coherence time is short enough, which can help to save computation complexity. Numerical results verify the effectiveness and correctness of our proposed strategies and analysis.