부산대학교 도서관

Different from the conventional transmit-then-compute scheme, which has been widely considered in mobile edge computing (MEC) networks, we propose a transmit-while-compute scheme to reduce the average latency. Specifically, by means of the partial offloading technique, the computational task is divided into multiple subtasks which are sequentially offloaded to the edge server node (ESN) with short-packet communications. Therefore, the transmission and computation of the task can be executed almost simultaneously. Moreover, to minimize the average latency, we first prove the optimization of corresponding tasks is quasi-convex, then provide an optimal solution to determine the blocklength and the size of subtasks. A low-complexity solution based on the alternating optimization (AO) method is proposed as well. Numerical results show that the average latency of the proposed scheme is very close to the computing time of the task.