Cellular Network Capacity Optimization For Unlicensed Spectrum And Ultra-Dense Deployment

In recent years,with the popularity of smart mobile devices and the increasing of diverse applications,the demand for mobile data traffic has grown exponentially.In order to cope with the challenge of high-speed traffic demand,Long Term Evolution-Unlicensed(LTE-U)and ultra-dense network,the key technologies of the fifth generation cellular mobile communication system,have attracted the attention in academia.The LTE-U technology improves the performance of the network by allowing the Long Term Evolution(LTE)system to share the unlicensed band with the Wi-Fi network to increase the available frequency band,while the ultra-dense cellular network technology reuses the spectrum to improve the spectrum efficiency and achieve the network capac-ity increasing.This paper studies how to boost the capacity of cellular network under unlicensed spectrum deployment and ultra-dense deployment.Firstly,we propose an LTE-U multi-cell framework utilizing licensed and unlicensed spec-trum at the same time.Based on this framework,we study joint resource allocation for LTE over the legacy licensed spectrum and the unlicensed spectrum.Specifically,we formulate a mixed-integer power-channel-time allocation problem aiming at maximizing the network throughput,with the constraints of protecting the coexisting Wi-Fi networks and hardware limitation on bandwidth oc-cupancy of user equipments(UEs),while guaranteeing per UE's throughput.To solve the resource allocation problem,we exploit the delay column generation(DCG)approach to decompose the original optimization problem and then propose a novel algorithm based on Karush-Kuhn-Tucker(KKT)conditions.Furthermore,we analyze the gap between the proposed algorithm and the opti-mal solution.Simulation results show the effectiveness of the proposed algorithm in terms of the network throughput and convergence speed.Moreover,the influence from network parameters,such as UE density and Wi-Fi network density,is also investigated in the simulation.Secondly,we study the mean packet throughput(MPT)per UE in a densely deployed dynamic time-division duplex(D-TDD)network considering not only random locations of UEs,small cell BSs(SBSs)and gateways,packet arrival process,interference,but also different kinds of back-haul.We use an approximate method to derive the successful transmission probability considering interference via stochastic geometry.The uplink(UL)and downlink(DL)MPT considering re-transmission,scheduling,and backhaul is analyzed through queueing theory tools and Markov chain.Moreover,we analyze the optimization between MPT and backhaul cost based on the de-ployment cost model.Numerical results demonstrates that the analysis results match well with simulation results under different kinds of backhaul,confirming the accuracy of the analysis.It also revealed that backhaul network,retransmission and scheduling have great impact on network performance.More results have shown the appropriate choice of gateway density will improve the D-TDD network performance and reduce the deployment cost.Finally,we study the learning-based UL/DL configuration under densely deployed D-TDD networks.A general model of interference analysis and UL/DL configuration is proposed and the influence of configuration on cell is expressed by the dissatisfaction of cells.A learning-based DL/DL configuration algorithm is proposed considering the UL/DL traffic,inter-cell interference,and dissatisfaction of other cells.Simulation results show the performance of proposed algorithm in improving network throughput and convergence.