Dynamic Computation Scheduling for Hybrid Energy Mobile Edge Computing Networks

Mobile edge computing (MEC) has grown rapidly due to the demand for resource-intensive applications on the Internet-of-Things. Computation offloading enables these applications to run at the network edge, but meanwhile it leads to significant energy expenses for network operators. One promising solution is to power the base stations (BSs) with a hybrid energy supply that combines unstable harvested energy and steady energy from the smart grid. This paper investigates the problem of computation scheduling for mobile devices (MDs) in a multi-MEC network with hybrid energy sources. Our objective is to maximize the long-term time-averaged service utility by jointly optimizing the battery supply of BSs, harvestable energy, CPU-cycle frequency, transmission power, task-partition factor, and association vector of MDs. We leverage the Lyapunov optimization framework and propose an online algorithm. Experimental work shows that our algorithm outperforms other benchmark schemes in terms of service utility and queue stability.