Short-Packet Edge Computing Networks With Execution Uncertainty

Abstract

Low-latency computational tasks in Internet-of-Things (IoT) networks require short-packet communications. In this paper, we consider a mobile edge computing (MEC) network under time division multiple access (TDMA)-based short-packet communications. Within the considering network, a mobile user partitions an urgent task into multiple sub-tasks and delegates portions of these sub-tasks to edge computing nodes (ECNs). However, the required computing resource varies randomly along with execution failure. Thus, we explore the execution uncertainty of the proposed MEC network, which holds broader implications across the MEC network. In order to minimize the probability of execution failure in computational tasks, we present an optimal solution that determines the sub-task lengths and the blocklengths for offloading. However, the complexity of the optimal solution increases due to the involvement of the Q function and incomplete Gamma function. Consequently, we develop a low-complexity algorithm that leverages alternating optimization and majorization-maximization (MM) methods, enabling efficient computation of semi-closed-form solutions. Furthermore, to reduce the computational complexity associated with sorting the offloading order of sub-tasks, we propose two sorting criteria based on the computing speeds of the ECNs and the channel gains of the transmission links, respectively. Numerical results have validated the effectiveness of the proposed algorithm and criteria. The results also suggest that the proposed network achieves significant performance gains over the non-orthogonal multiple access (NOMA) and full offloading networks.