WLAN Parameter Tuning Based On Deep Reinforcement Learning

With the rapid development of wireless communication technology,5G technology has stepped into real life from the national development strategy.The concept of interconnection of all things brought by 5G has greatly enhanced the value of wireless networks in the future society.WLAN can be used as a supplement to5 G,and it is widely used in many scenarios.To improve the service quality of the wireless local area network,the application of the 5G frequency band is indispensable.Due to the high fading of the 5G frequency band,the distribution of wireless AP is relatively dense.Due to the mutual interference between APs with highdensity coverage,this seriously affects the use value of wireless local area networks.Therefore,this article refines the scenario of a high-density AP distribution network,and proposes the following solutions to the problem of severe mutual interference between APs.The main work of the full text is as follows:1.Aiming at the problem of difficulty in estimating the saturated throughput of wireless local area network,this paper studies the modeling of single AP saturated throughput estimation based on Bianchi,and builds a model that can estimate the saturated throughput of wireless local area network on this basis.The model proposed in this paper extends the single AP behavior model proposed by Bianchi to the scenario of high-density AP distribution,and then further derives and calculates the Bianchi model to obtain the saturation of the wireless local area network in the scenario of high-density AP distribution.The estimation method of throughput.However,it should be noted that this model estimates the saturated throughput of downlink transmission in a high-density AP network scenario.2.The dynamic scene of wireless local area network with high-density AP distribution is greatly affected by user behavior and has strong Markov properties.In view of this situation,the control problem of carrier sensing threshold and AP transmit power is modeled as a MDP,and the DDPG is used to find the whole The best solution for the network.Traditional methods are often based on rules or models to adjust the carrier sensing threshold and AP transmit power.However,in the scenario of a high-density AP-distributed wireless local area network,the traditional method often consumes a lot of manpower and material resources,but it cannot be obtained.Good decision-making plan.The advantage of reinforcement learning is to solve decision-making problems,especially continuous decision-making problems.Therefore,reinforcement learning is very suitable for solving the problem of wireless local area network parameter design.The simulation experiment results in this paper also further confirm this conclusion.3.In a wireless local area network with high-density AP distribution,the computational complexity of the optimization problem increases exponentially with the increase in the number of APs.It is proposed that according to the interference relationship between APs,the distribution of APs can be established as a graphical model G(?,?).The traditional neural network of the strategy network and evaluation network in DDPG is replaced with GCN.Combined with this method,the decision cost can be further reduced,so that the DDPG algorithm can converge to a stable status.In this way,not only the training cost is reduced,but also the real-time performance of the algorithm is improved.In this way,the algorithm can reduce the co-frequency interference of the network in a more complex environment and increase the saturation throughput value of the network.