Research Of Intelligent Handover Algorithm Based On SDWN In Co-Channel Deployment

In recent years,wireless local area network(WLAN)has developed rapidly due to its convenience and easy expansion.With the large-scale application of WLAN,a large number of access points(APs)work in same frequency band,and station(STA)can detect multiple APs.The high mobility of STA leads to frequent handovers.Traditional WLAN handover mechanisms cannot meet STA's quality of service(Qo S)requirements,and it can easily lead to load unbalanced of network.Therefore,the problem of STA handover and load unbalancing of network under co-frequency deployment has become an urgent problem to be solved.Due to the poor scalability and lack of flexibility of traditional network architecture,it is difficult to manage movement of STAs,while software defined network(SDN)has the advantages of logical centralization,separation of control and data forwarding,and programming,so combine SDN with WLAN can effectively solve the problem of mobility management in WLAN.This thesis uses software defined wireless network(SDWN)architecture and introduces Reinforcement Learning(RL)to solve the problem of STA handover and load unbalancing of network.The specific work as follows:First of all,in view of the problem that STA cannot dynamically handover during movement,STA only selects the optimal AP according to received signal strength indication(RSSI)and handover delay is too long,a network architecture based on SDWN is designed,architecture uses logic access point(LAP)technology to achieve seamless handover.And a handover algorithm based on Deep Q-Network(DQN)is proposed.The algorithm uses RSSI to characterize states of STA,uses STA throughput,system throughput and fairness as reward,and algorithm uses a fully connected neural network.As the decision-making network outputs relevant decisions.In addition,in view of load unbalancing of network,a load balancing scheme based on Double DQN(DQN)is proposed.The scheme is mainly divided into two parts:selecting migration STA in overloaded AP and executing handover algorithm to select low-load AP for STA.The algorithm uses convolutional neural network(CNN)and fully connected neural network.CNN is based on AP throughput,AP of packet loss rate,the number of AP-associated stations,and signal to interference plus noise ratio(SINR)extracts load characteristics of AP,and fully connected neural network uses these characteristics to make association decisions.Finally,Mininet-Wifi simulation software is used to conduct experiments on the above schemes respectively.The simulation results show that the two schemes can effectively improve performance of WLAN network.