Decentralized Continuous Game for Task Offloading in UAV Cloud

Abstract

UAV cloud which integrates the flexibility and re-silience of mobile cloud computing (MCC) with multiple UAV system provides drones the ability of processing compute-intensive application by offloading task to cloud. However, such task with heterogeneous quality of experience (QoE) requirement generated by massive drones becomes a troublesome burden for cloud resource allocation. Especially the endurance issue related to the energy efficiency makes the problem more complicated. This paper proposes a game theory based decentralized continuous offloading algorithm. Each drone in the UAV cloud optimizes the percentage of offloading task executed at cloud, while minimizes its overhead composed by QoE requirement and energy consumption. This algorithm can be proved to a potential game that can reach a bilateral satisfaction Nash Equilibrium (NE) by finite iteration. Numerical results under various scenario corroborate not only the effectiveness and stability of the proposed continuous offloading game, but also the superiority of computation complexity and communication overhead.