A Novel Transmission Scheduling Based on Deep Reinforcement Learning in Software-Defined Maritime Communication Networks

With increasingly diversified communication services of ship users, quality of service (QoS) for data transmission becomes a bottleneck restricting the development of maritime communication. Aiming to solve this problem, firstly, a software-defined framework for maritime communication is presented to tackle the communication mode barrier in heterogeneous networks. Furthermore, under this framework, we propose a novel transmission scheduling scheme based on the enhanced deep Q-learning algorithm, which combines deep Q-network with softmax multiple classifier, also known as S-DQN algorithm. This scheme also mentions what the purpose of the optimization is (i.e., delay, cost, energy). We first employ Markov decision processes (MDPs) to achieve optimal scheduling strategy. Moreover, system builds up mapping relation between obtained information and optimal strategy by utilizing deep Q-network, and when incoming data arrives, it will make optimal strategy as fast as possible and accurately after a plethora of data self-learning. Simulation results show that the proposed scheme outperforms the other traditional scheme in terms of different QoS, which validate the effectiveness of the proposed scheme.