Smart Resource Configuration and Task Offloading with Ultra-Dense Edge Computing

The ongoing increasing number of mobile devices (MDs) with innovative applications yields unprecedented demands for user experience and network capacity expansion. The combination of ultra-dense network (UDN) and mobile edge computing (MEC) has been envisioned as a promising futuristic technology. It can remarkably improve the capacity of system and extend cloud-computing capabilities to proximate edge servers, by which the growing computation requirements of MDs will be met. However, what is not addressed well is how to optimize the configuration of computing resources with varying computation demands of MDs being satisfied, aiming at maximizing the operating earnings of operators while decreasing the cost of MDs. Considering the diverse computation demands in different regions and variational computing resources of edge servers, it is hard to solve this problem by traditional methods. To address this issue, an effective solution is generating an optimal computing resource configuration strategy and task offloading profile in time-varying UDN scenarios. Toward this end, a deep Q-network based scheme is proposed to achieve maximum long-term weighted network utility in such ever-changing environments. Simulation results validate the significant performance improvement of our scheme in weighted network utility and task offloading compared to conventional game-theoretical solution.