Optimization of Wireless Ad Hoc Network Node Layout Self-play Based on AlphaZero Algorithm

The success of AlphaZero algorithm in chess games provides ideas and methods for other fields. This research designs an intelligent deployment model of wireless ad hoc network node based on the deep reinforcement learning framework of AlphaZero algorithm. Aiming at the difficult problems of setting learning rate and preventing model overfitting in the process of model self-training optimization, this paper chooses gradually decreasing dynamics. Learning rate and regularization method are increased to ensure fast convergence in model training and improve the accuracy of model prediction.