Partial Task Offloading and Resource Allocation in WiFi-Cellular Multi-Access Edge Computing Networks

Enabling task offloading simultaneously in WiFi-cellular co-existing networks can effectively enhance the performance of an multi-access edge computing (MEC) system. In this paper, to minimize the task processing delays and energy consumption of all the user devices (UDs), we propose a joint partial task offloading, computing resource allocation and transmit power control scheme for a WiFi-cellular MEC network, which consists of a cloud server (CS) and two edge servers (ESs) equipped on the WiFi access point (AP) and the cellular base station (BS), respectively. We consider multiple partial offloading patterns, namely, the task of a UD can be divided and the sub-tasks can be partially offloaded to the AP, the BS, and the CS through the transmissions of the two networks, simultaneously. A twin delayed deep deterministic policy gradient (TD3)-based deep reinforcement learning (DRL) algorithm is designed to solve the optimization problem efficiently. We analyze the algorithm complexity and conduct extensive simulations in 4 scenarios. The convergence of our algorithm is validated and its superiority is demonstrated in comparison with 3 reference schemes.