User Scheduling Based on Multi-Agent Deep Q-Learning for Robust Beamforming in Multicell MISO Systems

Maximizing the rate in multiple input single output (MISO) systems using distributed algorithms is an important task that typically incurs high computational cost. In this work, we propose two deep Q-learning-based user scheduling schemes to solve the beamforming problem of sum-rate maximization with per base station power constraints in multicell MISO scenarios. The two key features of the proposed algorithms are that they are executed in a distributed fashion and are robust with respect to channel state information (CSI) errors. Simulation results show that in the presence of CSI errors the proposed schemes outperform state-of-the-art algorithms both in terms of average spectral efficiency and execution time.