Seamless and Intelligent Resource Allocation in 6G Maritime Networks Framework via Deep Reinforcement Learning

Sixth-generation (6G) communication networks will fulfill users' requests for high data speeds and low latency without causing network outages throughout the world. However, marine communication in deep-sea waters is expanding as maritime traffic grows. To serve mission-critical applications, a growing number of maritime end-users require high throughput and low latency. Although deep-sea satellite connections will enable 6G networks, however due to limited service capacity owing to the long propagation delay between end-user and satellites. We propose unmanned aerial vehicles (UAVs) as aerial backhauling and a relay medium in the marine communication network assisted by satellites and coastal base stations (BSs). The power allocation strategy of multi-satellites for a 6G network is investigated in this work. The power allocation problem is non-convex, hence deep reinforcement learning (DRL) is used to solve it instead of the conventional optimization technique. Our research aims to optimize the network's total capacity in the case when the satellites are randomly and densely dispersed. We provide a deep neural network and a method for mapping wireless power allocation for multi-satellites. The proposed technique can achieve a better overall capacity in comparison to the water-filling and the Q-learning method. The simulation results further demonstrate that the suggested technique offers a notable increase in stability and convergence speed.