NOMA based Efficient Spectrum Sharing for Underwater UAV System with Multi-agent Reinforcement Learning

Abstract

The underwater communications have been challenged by the long propagation delay as well as the limited utilizable bandwidth. As a result, although the underwater UAV system can benefit many applications, the communication performance still works as a barrier for its wide adoption. This paper proposes the utilization of non-orthogonal frequency division multiple access to increase the efficiency of underwater UAV networks, which will ideally avoid the negative effect from long propagation time, but rise multiuser interference problem. To solve this problem, this paper model the multi-user interference as a non-cooperative game, and resort the multiagent reinforcement learning to approach the Nash Equilibrium. This method can avoid the a priori constraint of scenario information, which usually cannot be obtained in the underwater UAV communications. A simple but descriptive simulation has validated the feasibility of proposed method.