Reliability-Optimal Offloading for Mobile Edge-Computing in Low-Latency Industrial IoT Networks
Jie Wang,Yulin Hu,Yao Zhu,Tong Wang,Anke Schmeink
DOI: https://doi.org/10.1109/twc.2024.3396161
2024-01-01
Abstract:In this paper, we study a multi-access mobile edge computing (MEC) network in the industrial Internet-of-Things (IoT) scenario, which aims at providing a joint computation service for a group of sub-tasks offloaded from multiple user equipments (UEs). The whole MEC service, including a communication phase and a computation phase, is required to satisfy both a low latency and a high reliability requirement. We derive the end-to-end reliability (error probability) of the whole MEC service and provide corresponding reliability-optimal design frameworks, where both the perfect channel state information (CSI) and outdated CSI scenarios are considered. In particular, we characterize the low-latency communication behavior with the consideration of the finite blocklength (FBL) impact, and exploit the extreme value theory to study the delay violation probability in the computation phase. Following the characterizations, in the perfect CSI scenario, a design framework minimizing the instantaneous end-to-end error probability is provided, i.e., via optimally choosing the time length for each user’s offloading and the time length for the computation phase. We rigorously prove the convexity of the problem, investigate the relationships among the variables in the optimal solution, based on which a low-complexity method is proposed achieving the optimal solution. In addition, for the scenario with only the outdated CSI, after deriving the expected end-to-end error probability conditioned on the outdated CSI value, a corresponding optimal time allocation design is provided as well, where the convexity of the formulated problem is characterized and the optimal solution is obtained. Via simulations, we validate our analytical model and evaluate the network performance under the design.