Cooperative Communciation Resource Optimization Strategy Of Wireless Heterogeneous Network Based On Reinforcement Learning

With the development of the Internet and communication technology,the number of mobile users has surged,and user demands are increasing.Deploying Heterogeneous Network(HetNet)is considered to be one of the most common methods to solve the mentioned problems.However,problems such as unbalanced load and huge energy consumption will occur in HetNets.To solve the mentioned problems,it is necessary to design an effective cooperative optimization strategy of wireless HetNet communication resource.In addition,existing optimization strategies are usually limited to fixed scenarios,and cannot achieve ideal performance at low cost when applied to different HetNets.To solve the mentioned problems,it is necessary to improve the generalization ability of the collaborative optimization algorithm of wireless HetNet communication resource and reduce the training cost.To solve the problems of unbalanced load and huge energy consumption in HetNets,this paper proposes a user-base station association and power control optimization algorithm based on Deep Deterministic Policy Gradient(DDPG).First,we improve the DDPG algorithm so that it can deal with the discrete user-base station association problem and the continuous power control problem at the same time with a faster convergence.Secondly,compared with the traditional resource optimization algorithm based on simulated annealing algorithm,the proposed algorithm has fewer parameter settings and shorter training time.The simulation results show that the proposed algorithm can achieve the improvement of the sum of all user rates and energy efficiency.To solve the problems that the generalization ability of the collaborative communication resource optimization algorithm needs to be improved and the training cost needs to be reduced,this paper proposes a transfer learning communication resource optimization algorithm based on DDPG.Specifically,the collaborative resource optimization algorithm designed for a single communication background has weak generalization ability and poor performance in other HetNets.In addition,training an optimization algorithm based on deep reinforcement learning(DRL)in new scenarios is computationally expensive.To solve the mentioned problems,this paper designs a transfer learning communication resource optimization algorithm based on DDPG.The simulation results show that compared with the collaborative communication resource optimization algorithm designed for single scenario,the generalization ability of the proposed algorithm has been improved;compared with the collaborative communication resource optimization algorithm based on DRL,the algorithm accelerates the training speed and reduces the training cost in the new scenarios.