WLAN Association Control Mechanism For High Client-density Environments

In recent years,IEEE 802.11 WLAN technology has developed rapidly and is widely deployed in many scenarios such as homes,campuses,corporate parks,airports,etc.,to provide high-speed wireless connections for user devices.With the explosive growth of the number of user devices and the increasingly abundant Internet applications,WLANs have shown a trend of dense users in most public places.In a high client-density environment,due to limited spectrum channel resources,incremental deployment of APs cannot effectively improve network capacity,and channel contention among users is more intense.The traditional association control mechanism makes the client select the AP with the highest signal strength to connect.It is easy to cause load imbalance between APs,which seriously affects user experience.Therefore,there is an urgent need for an effective association control mechanism to improve WLAN performance and user experience in a high client-density environment.However,due to the many and complex connections between APs and clients,the dynamics of the environment,and the time-varying nature of user requirements,it is difficult to directly design a corresponding association control strategy,and one possible idea is to use machine learning techniques to assist the design.At present,the use of machine learning technology for wireless network optimization has gradually become a trend,and the reinforcement learning method in the field of machine learning can combine data generation with model training and introduce dynamic feedback signals in the learning process,which is suitable for WLAN association control problems in complex environments.Therefore,deep reinforcement learning technology is introduced in this thesis to design a self-learning and selfadapting association control scheme for complex environments,which not only performs intelligent association control during user access,but also performs dynamic user transfer during network operation.Specifically,this thesis studies the intelligent association control mechanism under the condition of multi-AP coordination and the dynamic user transfer mechanism driven by time-varying demand.The work mainly includes the following three aspects.Firstly,the intelligent association control mechanism under the condition of multi-AP coordination is investigated and designed.In this mechanism,strategies such as state reconstruction,two-dimensionalization,and discretization are used to define the elements of reinforcement learning,and an intelligent association control algorithm based on deep reinforcement learning is proposed,which solves the problem of AP load imbalance and user experience deterioration in high client-density environments.This mechanism implements association control that is transparent to users and is determined by the infrastructure side,and can significantly improve WLAN performance and user experience quality in high clientdensity environments.Secondly,the dynamic user transfer mechanism driven by time-varying demand is investigated and designed.In this mechanism,key factors affecting load balancing and user experiences,such as signal strength,AP load,user demand,and transfer frequency,are considered comprehensively to design a combination of triggered and periodic transfer mode and a reasonable transfer object selection strategy.Based on this,this thesis proposes a user transfer algorithm to solve the challenges caused by frequent changes in network status and user demand.This mechanism ensures that users can quickly and accurately transfer to the target AP based on the decision results,and further improves the WLAN performance and user experience in high client-density environments based on association control.Finally,the above research results have been applied to the investigation and design of a WLAN association control prototype system for high user-density environments.The system implements many functions such as intelligent association control,dynamic user transfer,automated testing,and data stream binding.Then,build a high client-density environment emulation platform based on multiple radio frequency interfaces,so that the prototype system can automatically perform model training and performance verification.In summary,this thesis not only investigates the association control mechanism under the condition of multi-AP coordination and proposes an intelligent association control algorithm based on deep reinforcement learning,but also designs a dynamic user transfer mechanism and proposes a transfer algorithm based on the characteristics of time-varying user demand,and finally develops a prototype system based on a high client-density environment emulation platform.The research work in this thesis is essential for improving WLAN throughput and making full use of potential bandwidth,and provides a theoretical basis and technical support for realizing WLAN load balancing and improving user experience in high client-density environments.