Monte Carlo Tree Search for Network Planning for Next Generation Mobile Communication Networks

In this paper, we investigate the network planning problem in mmWave mobile communication systems, where the narrow-beam antennas can adjust azimuths and downtilts of antennas so as to maximize the power coverage of the network, as well as the system throughput. Searching for the optimal configurations of antennas generally yields a combinatorial opti-mization problem, which cannot be addressed even for a medium scale antenna set case. We formulate this optimization task as a finite Markov decision process, and develop a multi-layer Monte Carlo tree search method to produce a promising solution with reasonable complexity, which evaluates the outcome of given azimuth and downtilt settings without acquiring all configurations of antennas. Experiments in a real urban environment show that our proposed scheme outperforms the state-of-the-art algorithms over 10% in terms of system throughput while guaranteeing high power coverage.