Research On The Intelligent And Efficient Access Control And 3-D Deployment In Wireless LAN

Nowadays,802.11 wireless local area networks(WLANs) have been widely deployed around the world and have become an indispensable part of people's life.With the increase of wireless devices,the development of mobile services and the growth of wireless traffic,users have higher performance requirements for WLAN such as throughput and delay.Driven by the requirements of users,802.11 standards are also evolving.However,the evolution of 802.11 standards is mainly focused on the physical layer.Without further optimization of the MAC layer protocol which manages the access control,users will have little benefit from the improved physical layer performance.On the other hand,in order to improve area throughput,joint deployment of multiple WLAN access points(APs) has become an inevitable trend.However,unlike the two-dimensional deployment of terrestrial APs,unmanned aerial vehicle(UAV) communication network based on WLAN presents new challenges to the de-ployment of APs.From the perspective of access control and AP deployment,in this thesis,we optimize the MAC layer protocol of WLAN,and study the access control problems that need to be solved urgently in MAC layer,in order to improve WLAN performance such as throughput,delay and fairness.On the other hand,we optimize the 3-D deployment of UAV communication network with the purpose of high energy efficiency and persistence.The main works and contributions of this thesis can be summarized as follows.1.MAC backoff algorithm in heavy node loaded WLANWith the popularization of intelligent devices,the number of nodes in WLAN is ever-increasing.When a large number of nodes associate with WLAN,the network suffers from severe packet collision and poor short-term fairness due to the inappropriate binary exponential backoff(BEB) algorithm.In this thesis,we provide an enhanced backoff(EBO)algorithm to improve the performance of BEB.Firstly,in order to reduce the collision probability,the size of backoff interval increases by the initial value of contention window after an unsuccessful transmission,and backoff intervals in different backoff stage are disjoint.Secondly,to improve the short-term fairness,we slow down the decre-ment of contention window by resetting the contention window to initial value after T consecutive successful transmissions.In this thesis,we not only the-oretically analyze the performance of EBO algorithm using discrete time 3-D Markov chain,but also evaluate the performance of EBO algorithm comparing with other backoff algorithms by simulations.2.Thompson sampling based rate adaptation in 802.11ac WLANWith the evolution of 802.11 standard,the physical layer of the standard becomes more and more complex.The rate search space of 802.11ac is huge due to the introduction of Multi-input Multi-output(MIMO) and channel bond-ing,and traditional rate adaptation(RA) algorithms are unsuitable.To solve this problem,we design a novel RA algorithm termed rate adaptation with Thompson Sampling(RATS).In this algorithm,we first consider compacting the search space by removing some rates to accelerate the convergence of the algorithm.Moreover,inspired by multi-armed bandit problem,we design RA algorithm based on Thompson Sampling.Specifically,we design rate adaptation with general Thompson Sampling(RAGTS)and rate adaptation with sliding-window Thompson Sampling(RASWTS) for stationary and non-stationary channel environments respectively.Simulation results demonstrate that the performance of the proposed RATS outperforms the existing method and is closer to the performance upper limit.3.Multi-agent deep reinforcement learning based on-demand channel bondingWith the continuous growth of wireless traffic,the traffic load of WLAN AP is increasing day by day.In order to meet different and varying traffic demand of APs,we propose an on-demand channel bonding(O-DCB) algo-rithm based on multi-agent deep reinforcement learning(DRL) in heterogeneous WLANs where the APs have different channel bonding capability.In O-DCB,each AP corresponds to an agent.The load size and average load arrival rate of APs are chosen as the state,and the average transmission delay of the total network and the average load size of all APs are used as the reward.Real traffic traces collected from a campus WLAN are used to train and test O-DCB.Simulation results reveal that the proposed algorithm has good convergence and lower delay than other algorithms.4.Deep reinforcement learning based 3-D deployment of UAV aerial APsDue to the energy limitation of UAV,persistence is the key problem of UAV communication network.In order to realize energy efficient,fair and per-sistent communication service,we propose a UAV aerial APs control algorithm based on DRL,where limited UAV aerial APs move around to serve users and recharge timely to replenish energy.Firstly,detailed UAVs communication system models are built,including channel model,data rate model and energy consumption model.Then,we carefully design the state,action and reward.Specifically,the locations,the residual energy of UAVs and the accumulative received data volume of users are chosen as the state.The flight direction and flight distance of UAVs are used as the action.Energy efficiency and fairness are designed as the reward.For the proposed algorithm,we give a detailed description and perform a lot of training.Simulation results reveal that the proposed algorithm shows a good convergence and outperforms other scheduling algorithms in terms of data volume,energy efficiency and fairness.