Minimizing Task Delay for Mobile Edge Generation in D2D Underlaying Cellular Network

A mobile edge generation (MEG) framework is proposed for latency-sensitive tasks in social-aware D2D underlaying cellular networks. User devices (UDs) are organized into socially cohesive communities for cooperative generation via cellular and D2D links. A joint seed-and-content-based (JSCB) protocol is proposed to enable adaptive switching UD association between seed acquisition with local generation and direct content sharing, while a spillover mechanism is incorporated to handle overdue transmissions. An average task delay minimization problem is formulated to jointly optimize UD association, transmission mode, D2D pairing, and beamforming. To address the hybrid, temporally coupled problem, a joint matching and proximal policy optimization (JMPPO) algorithm is developed, in which discrete and continuous actions are decoupled through hierarchical deep reinforcement learning and matching. Numerical results demonstrate that JSCB achieves reduced delay via adaptive scheduling, while JMPPO outperforms learning-based and traditional baselines under diverse conditions.