Task-Driven Robust Integration of Communication and Computation for Edge-Intelligent Networks

In this paper, we investigate the issue of integrated communication and computation for multiple time-sensitive computation-intensive user equipments (UEs) with different types of tasks in edge-intelligent networks. Especially, we consider a practical edge-intelligent network with both communication and computation uncertainties, where channel state information (CSI) is partially obtained by the base station (BS) and the task complexity is inaccurately estimated by the mobile edge computing (MEC) server. To effectively mitigate the influences of these unfavorable uncertainties and guarantee user fairness, a task-driven robust design algorithm for integrated communication and computation with the objective of minimizing the maximum system delay among all UEs is put forward by jointly optimizing transmit power at the UEs, receive beamforming at the BS and computing resources at the MEC server based on task types. Both theoretical analysis and simulation results confirm the robustness and the effectiveness of the proposed algorithm for edge-intelligent networks.