Joint Offloading and Resource Allocation with Partial Information for Multi-user Edge Computing.

Task offloading has received enough attention from academia and industry, which relieves the limitation of computing and storage capacity of devices and improves the quality of service (QoS). Currently, most studies on task offloading require that tasks information is completely known for all users in the system. However, in many cases, the information is partially observable, so our work mainly solves task offloading and resource allocation for multi-user edge computing, where the task information collected by the center controller is partial, i.e. certain users’ tasks information is missing. In view of this situation, we propose a prediction-based information completion scheme (PBIC) to alleviate the impact of partial information collected. To minimize the total latency of the system, each user can select one base station (BS) for partial task offloading, and each edge server and BS can allocate computing and bandwidth resources to users respectively. We also propose a scheme combining adaptive information retransmission and predictive completion, namely AIRPC, to further improve the performance of the system from a long-term perspective. Numerical results show the superiority of the proposed schemes over existing methods. Compared with the conventional scheme, the delay cost can be reduced by 27.5% to 34.6% with AIRPC.