Short- and long-term cost and performance optimization for mobile user equipments

Task offloading strategy optimization in mobile edge computing (MEC) has always been a hot issue. However, the mobility of a user equipment (UE) seriously affects the UE's cost and performance. This paper proposes three mobility types depending on whether the mobility characteristic of a UE is known, and formulates an energy minimization problem and a latency minimization problem to optimize the cost and performance, respectively. We first develop greedy strategy based task offloading algorithms for UEs according to their mobility characteristics. However, accurately obtaining the mobility characteristics of the UEs over a long time in practice is a huge challenge, especially in a highly random environment like the MEC. To address the issue, we use a Lyapunov optimization method to develop the algorithms that do not require any prior knowledge of the mobility characteristics to minimize the long-term energy and latency of UEs. Experimental results show that the greedy strategy based algorithms can optimize the cost and performance of UEs by using their mobility characteristics, and perform better than the Lyapunov optimization based algorithms in a short-term. However, the Lyapunov optimization based algorithms perform better than the greedy strategy based algorithms over a long-term. © 2020 Elsevier Inc.