The dense heterogeneous network provides standardized connectivity and access guarantees for 5G communication services.However,the complex network environment and high level of dynamism pose challenges to network sele...The dense heterogeneous network provides standardized connectivity and access guarantees for 5G communication services.However,the complex network environment and high level of dynamism pose challenges to network selection decisions.Existing vertical handover algorithms often overlook the dynamic nature of user mobility and network condition,resulting in problems such as handover failure and frequent handover,ultimately impacting the quality of the user communication service.To address these problems,we propose an intelligent switching method,iMALSTM-DQN,which integrates an improved Multi-level Associative Long Short-Term Memory model(iMALSTM)with Deep Reinforcement Learning(DRL).The algorithm leverages iMALSTM to predict the global network state in the next moment based on the global user movement trajectory and historical network status information within a region,thereby enhancing the prediction accuracy of network states.Subsequently,based on the predicted network state,we employ the Deep Q Network(DON)model to make handover decisions,adaptively determining the optimal switching and network selection strategy through interaction with the environment.Experimental results demonstrate that the proposed algorithm enhances decision timeliness,significantly reduces the number of switch failures,and alleviates the problem of frequent handovers resulting from network dynamics.展开更多
基金National Key Research and Development Program of China(No.2022YFB3903404,2024YFC3015403)National Natural Science Foundation of China(NSFC No.42271431,42271425)Hubei Province Major Science and Technology Innovation Program(2024BAA011)。
文摘The dense heterogeneous network provides standardized connectivity and access guarantees for 5G communication services.However,the complex network environment and high level of dynamism pose challenges to network selection decisions.Existing vertical handover algorithms often overlook the dynamic nature of user mobility and network condition,resulting in problems such as handover failure and frequent handover,ultimately impacting the quality of the user communication service.To address these problems,we propose an intelligent switching method,iMALSTM-DQN,which integrates an improved Multi-level Associative Long Short-Term Memory model(iMALSTM)with Deep Reinforcement Learning(DRL).The algorithm leverages iMALSTM to predict the global network state in the next moment based on the global user movement trajectory and historical network status information within a region,thereby enhancing the prediction accuracy of network states.Subsequently,based on the predicted network state,we employ the Deep Q Network(DON)model to make handover decisions,adaptively determining the optimal switching and network selection strategy through interaction with the environment.Experimental results demonstrate that the proposed algorithm enhances decision timeliness,significantly reduces the number of switch failures,and alleviates the problem of frequent handovers resulting from network dynamics.
