Game Theory based D2D Collaborative Offloading for Workflow Applications in Mobile Edge Computing

Device-to-device (D2D) collaborative offloading is a promising complement to the Device-Edge-Cloud hierarchical offloading paradigm, in which the computational tasks of an edge device can be offloaded to nearby devices with idle resources by D2D communication. However, because the devices are owned by different individuals, the conflicting interests among offloading requesters and resource providers present a substantial challenge for D2D offloading, especially when the tasks have dependency relationships with strict time constraints. To encourage the edge devices of individuals to participate in the offloading and maintain a sustainable collaborative community, this study presents a novel game theory-based D2D offloading approach for workflow applications in a dynamic mobile edge computing (MEC) environment. We first introduce a satisfaction metric to assess the collective benefits of the stakeholders. Then our offloading approach employs game theory to maximize collective benefits. To enable reaching a real-time offloading decision, no-regret dynamics is leveraged to accelerate the convergence of the game process. Experiments demonstrate that our approach can achieve high collective benefits with a satisfactory quality of service.