A Novel Lyapunov based Dynamic Resource Allocation for UAVs-assisted Edge Computing

Mobile edge computing (MEC), as a key component in the development of IoT and 5G technologies, can provide extra computation resources in edge servers for mobile devices to complete their computation tasks with low latency and high reliability. Considerable efforts on computation offloading and resource allocation have been developed to reduce the energy consumption and computation latency in edge computing. Nonetheless, the system utility of heterogeneous edge computing system (e.g., UAVs-assisted edge computing), in which multiple unmanned aerial vehicles (UAVs) are involved in an edge computing system to serve as edge servers still needs to be further investigated. To this end, in this paper, we propose a novel Lyapunov based Dynamic Resource Allocation (LDRA) for UAVs-assisted Mobile Edge Computing, which can effectively choose suitable edge servers for mobile devices to offload and complete their computation tasks with low system cost and great system utility of UAVs-assisted edge computing system, as well as acceptable computation latency and great reliability for computation tasks of mobile devices. Particularly, a random queue model for edge servers is conducted in our LDRA scheme to support the dynamic of offloaded computation tasks of mobile devices. Additionally, a system cost model of UAVs-assisted edge computing is developed considering the combination of multiple constraints, such as both the mobility of UAVs and mobile devices, energy consumption, communication cost, etc. With the objective of minimizing the system cost and maximizing the system utility in providing edge resources to complete the offloaded computation tasks of mobile devices, by introducing Lyapunov optimization, a dynamic resource allocation scheme is proposed to effectively determine edge servers to offload tasks of mobile devices with considering both the real-time execution state of offloaded tasks in edge servers and states of the communication link. Through analysis and performance evaluations, our results show that our proposed LDRA scheme can achieve a great balance between system cost and system stability. Additionally, our results also demonstrate that our LDRA scheme also can achieve better system utility in comparison with existing schemes.