Optimal Schedule of Mobile Edge Computing Under Imperfect CSI.

Mobile edge computing (MEC), as a prospective computing paradigm, can augment the computation capabilities of mobile devices through offloading the complex computational tasks from simple devices to MEC-enabled base station (BS) covering them. However, how to achieve optimal schedule remains a problem due to various practical challenges including imperfect estimation of channel state information (CSI), stochastic tasks arrivals and time-varying channel situation. By using Lyapunov optimization theory and Lagrange dual decomposition technique, we propose an optimal dynamic offloading and resource scheduling (oDors) approach to maximize a system utility balancing throughput and fairness under imperfect estimation of CSI. We derive the analytical bounds for the time-averaged data queues length and system throughput achieved by the proposed approach which depends on the channel estimation error. We show that without prior knowledge of tasks arrivals and wireless channels, oDors achieves a system capacity which can arbitrarily approach the optimal system throughput. Simulation results confirm the theoretical analysis on the performance of oDors.