Spectrum Sharing in Cache-Empowered Wireless Networks: Reservation-Based Versus Auction-Based Approaches

Spectrum sharing has attracted much recent attention from both industry and academia due to its significant potential of achieving high radio resource efficiency and scaling the service capability in wireless networks. However, how to fairly and efficiently share the radio spectrum or bandwidth in cache-empowered wireless networks remains open. In this paper, we present a reservation-based and an auction-based approach for dynamic spectrum access in cache-empowered wireless networks that serves both on-demand and pushing traffics. More specifically, a virtual network operator (VNO) is introduced as a spectrum access scheduler, which pays to the spectrum owner for the bandwidth it uses while charging its users for their content requests. To ensure the quality of service, a VNO is punished if it fails to satisfy a user’s request in peak time. By balancing the bandwidth cost and penalty, multiple VNOs may maximize the spectrum efficiency in a distributed manner. To further improve the spectrum efficiency in cache-empowered wireless networks, we formulate a linear program and a nonlinear program for both the reservation-based and the auction-based spectrum access, respectively. A modified deep Q-network algorithm is also presented to substantially reduce the computational complexity and to attain a distributed online spectrum sharing protocol.