Task Offloading in UAV-Assisted Mobile Cloud-Edge Computing Networks: An AoP-Aware HAPPO Approach

In recent years, unmanned aerial vehicles (UAVs) equipped with edge servers have emerged to expand service coverage. However, their inherent limitations in bandwidth, computational capacity, and energy resources restrict their ability to meet the demands of computation-intensive applications. To overcome these challenges, an air-ground collaborative mobile edge computing (MEC) network with enhanced resources is employed to improve Quality of Service (QoS). This network can provide support for a wide range of real-time applications where the computed results are critical. We adopt Age of Processing (AoP) as a metric to measure the freshness of processed results. This study investigates AoP-aware task offloading and resource allocation in a multi-UAV-assisted MEC system, optimizing the trade-off between user-experienced AoP and UAV energy consumption through joint design of user-UAV association, resource allocation, offloading ratios, and transmit power. To solve the formulated non-convex problem, we employ a cooperative Heterogeneous-Agent Proximal Policy Optimization (HAPPO) approach, utilizing a sequential policy update strategy within a centralized training, decentralized execution framework to mitigate inefficiencies from overlapping UAV coverage. Simulation results demonstrate that the proposed approach can achieve better performance compared with the other approaches.