AI-Enabled Distributed Spectrum Sharing

This chapter investigates the AI-enabled distributed spectrum sharing in the wireless edge network, where multiple BSs share the same bandwidth and may suffer severe interference. This inevitably degrades the network spectrum efficiency (SE). To tackle this issue, conventional methods involve centralized power control strategies that rely on global instantaneous CSI. In this chapter, we propose a multi-agent independent actor-critic (MAIAC) power control algorithm for each BS to optimize local transmit power by enabling lightweight collaborations between the core network (cloud) and different BSs (edge). In particular, each BS learns its the power control policy locally with a cloud-based global reward mechanism. Simulation results show that the proposed MAIAC algorithms can respectively achieve over \(99\%\) SE performance of conventional algorithms in quasi-static scenarios, and approximately \(89\sim 100\%\) in dynamic scenarios with significantly reduced time complexity.