Deep Multi-User Reinforcement Learning for Centralized Dynamic Multichannel Access

We consider the problem of dynamic multichannel access for transmission maximization in multi-user wireless communication networks. At the beginning of each time slot, the centralized node evaluates all K channels and allocate a channel to transmit a packet for each user. After each time slot, the centralized node receives feedback signals for each user indicating whether the packet was successfully delivered or not. The objective is to find a multi-user strategy that maximizes global channel utilization with a low collision in a centralized manner without any prior knowledge. Obtaining an optimal solution for centralized dynamic multichannel access problem is a difficult problem due to the large-state and large-action space. To tackle this problem, we develop a centralized dynamic multichannel access framework based on double deep recurrent reinforcement learning. Specifically, at each time slot, the centralized node map current state to channel access actions and select multiple channels where the sum of corresponding value function obtains its maximum value based on neural network. The proposed method decreases collision through centralized allocation policy to achieve an optimal multi-user channel allocation.