Multi-channel Deep Q-network Carrier Sense Multiple Access

With the continuous development of network communication and the application of many specific scenarios, the dynamics of network traffic continues to increase, making the optimization of routing problems an NP-hard problem. When using traditional routing algorithms, accuracy and efficiency cannot be balanced. Recently, Deep Q-Network (DQN) has shown great potential for solving dynamic network problems. However, existing DQN-based routing solutions often overlook network environment issues related to packet level, packet size, expected transmission time, and do not generalize well when the network changes. In this paper, we present a new carrier sense multiple access (CSMA) protocol called MC-DQN CSMA, which employs Deep Q-Network to improve the performance of the network. First, we propose a distance constraint under the signal-to-interference-to-noise ratio (SINR) model, which effectively avoids interference and improves the probability of success. Based on the dynamic and unpredictable needs of Ad Hoc networks, we try to use DQN strategy to train the network’s agents without expert knowledge. Furthermore, we demonstrate the performance of the proposed algorithm by comparing it with other methods and describing it graphically, which focus on transmitting packets in multi-channel Ad hoc networks.