Deep Learning-Based Spectrum Sharing in Next Generation Multi-Operator Cellular Networks

Owing to the exponential increase in wireless network services and bandwidth requirements, sharing the radio spectrum among multiple network operators seems inevitable. In wireless networks, enabling efficient spectrum sharing for resource allocation is quite challenging due to several random factors, especially in multioperator spectrum sharing. While spectrum sensing can be useful in spectrumsharing networks, the chance of collision exists due to the inherent unreliability of wireless networks, making operators reluctant to use sensing-based mechanisms for spectrum sharing. To circumvent these issues, we utilize an alternative approach, whereby we propose an efficient spectrum-sharing mechanism leveraging a spectrum coordinator (SC) in a multi-operator spectrum-sharing scenario assisted by deep learning (DL). In our proposed scheme, before the beginning of each timeslot, the base station of each operator transmits the number of required resources based on the number of packets in the base station’s queue to SC. In addition, base stations also transmit the list of available channels to SC. After gathering information from all base stations, SC distributes this collected information to all the base stations. Each base station then utilizes the DL-based spectrum-sharing algorithm and computes the number of resources it can use based on the number of packets in its queue and the number of packets in the queues of other operators. Furthermore, by leveraging DL, each operator also computes the cost it must pay to other operators for using their resources. We evaluate the performance of the proposed network through extensive simulations. It is shown that the proposed DL-based spectrum-sharing mechanism outperforms the conventional spectrum allocation scheme, thus paving the way for more dynamic and efficient multi-operator spectrum sharing.