Resource Allocation of Low Earth Orbit Satellites Based on Deep Reinforcement Learning
Aiming at the problem that the optimization of energy efficiency(EE)and spectrum efficien-cy(SE)in low earth orbit(LEO)satellites cannot maintain a consistent growth trend,a method for opti-mizing the trade-off between EE and SE in LEO satellites is proposed.This method models the LEO satel-lite resource allocation scenarios,simplifies the dynamic model by dividing time slots,and optimizes the throughput by adjusting sub-carrier power,thereby optimizing EE and SE.In addition,EE and SE are weighted by introducing weight factors and unifying the units of EE and SE,so as to achieve the maximum balance between them.In order to deal with the large state action space problem,the Dueling DQN algo-rithm is used to achieve a better control strategy.Simulation results show that compared with other deep reinforcement learning(DRL)algorithms,the proposed algorithm converges faster and the convergence value is increased by 10.1%and 18.2%higher,respectively.When the noise power changes,the SE ob-tained by Dueling DQN is increased by 15.6%,compared with other DRL algorithms.
NETL
NSTL
万方数据
