Joint Task Offloading and Resource Allocation for LEO Satellite-Based Mobile Edge Computing Systems With Heterogeneous Task Demands

To satisfy the growing demand for supporting intelligent Internet of Things (IoT) applications in remote areas, satellite mobile edge computing (SMEC) systems are expected to be widely deployed. Meanwhile, as IoT applications are increasingly diversified, the tasks received by SMEC systems present heterogeneous demands for various resources and quality of service (QoS) metrics, necessitating customized design to accommodate such heterogeneity. In this paper, we investigate the problem of joint task offloading and resource allocation in SMEC systems with heterogeneous task demands. We first propose a customized task utility model that captures the diversified demands, which is an extension of traditional unified task utility models. Based on the model, we then formulate a mixed integer non-linear programming (MINLP) problem for joint optimization of task offloading, computing resource allocation, transmission power control, and user association, aiming to maximize the sum utility of all tasks. To solve the MINLP, a multi-layer iterative framework is proposed that decomposes the original problem into two subproblems, which are solved by successive convex approximation (SCA) and deep reinforcement learning (DRL) algorithms, respectively. Simulation results show that, by applying the proposed task utility model, the average utility performance of UEs can be improved by 50% and 100% compared to applying a single metric-based utility model and a typical classification-based utility model, respectively; the proposed task offloading and resource allocation scheme achieves a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$30\%\hbox{--}80\%$</tex-math></inline-formula> performance gain compared to benchmark schemes.