Performance Optimization Of Heterogeneous Networks Based On Coupled Queue

To meet surging traffic demands,heterogeneous networks(HetNets)enable a more flexible,targeted and economical deployment of new infrastructure versus tower-mounted macro-only systems,which are very expensive to deploy and maintain.For a high network spectrum efficiency,load balancing across different tiers can be achieved by optimizing the association between users and base stations(BSs).To achieve a high energy efficiency,proper controls of BSs' activation(on/off status)and deployment density can significantly avoid unnecessary BS power consumption.However,some practical conditions are not considered in existing studies.1)Previous studies usually assumed that BSs were always busy transmitting packets to their associated users,which characterized a worst case of the performance metrics.In practice, one BS can either be busy or idle,depending on its queuing condition,in which case theperformance metrics such as the packet delay should be further studied with queuing taken into account.2)With the assumption of continuous BS transmission in previous literatures,the network power consumption linearly increases with the number of BSs only.Practically,the power consumption of a BS in the idle state is much lower than that in the busy state,the tuning of the network design parameters,for example,the bandwidth allocation and the BS deployment density,thus have a significant impact on the BS busy/idle status,which in turn affects the network energy efficiency.3)Most of the previous studies focus on a uniform user distribution.In reality,users might not be evenly distributed and may form a cluster in certain hot area.In such cases,the user association optimization in a per-tier fashion would result in a poor user experience in the overloaded areas,and a per-station association scheme is thus preferable.To address the above considerations,the thesis focuses on the optimization of both the network spectrum efficiency and the network energy efficiency with practical assumptions of queuing and non-uniform user distribution,which is elaborated in the following.1)Delay-optimal biased user association in HetNets.A thinned Poisson point process modelto characterize the locations of BSs in the busy state,and an explicit expression of the average traffic intensity of each tier is obtained.On that basis,an optimization problem is formulated to minimize the lower bound of the network mean queuing delay by tuning the biasing factor of each tier,which is shown to be a convex problem.The simulation results demonstrate that the network queuing performance can be significantly improved by properly tuning the biasing factor.It is further shown that the network mean queuing delay might be improved at the cost of a deterioration of the network signal-to-interference ratio (SIR)coverage,which indicates a performance tradeoff between real-time and non-real-time traffic in HetNets.2)Queue-aware optimal bandwidth allocation in HetNets.Based on the queuing analysis,a minimization problem of the network average power consumption and a maximization problem of the network SIR coverage are formulated,which are shown to be convex and concave with respect to the bandwidth allocation to each tier,respectively.By using an approximation of the average traffic intensity,closed-form solutions are obtained for both problems.Simulation results of a 2-tier HetNetdemonstrate that the network average power consumption and the SIR coverage can be significantly improved by the optimal bandwidth allocation.3)Queue-aware energy efficient base station density optimization in HetNets.By further using the approximation that BSs of a tier have the same SIR coverage,the cumulative distribution function(CDF)of the traffic intensity of each tier is obtained.On that basis,a minimization problem of the network average power consumption is studied by optimally tuning the activation ratio of micro BSs under the quality of serive(QoS)constraints of the network mean queuing delay and the network SIR coverage.Numerical results demonstratethat if the idle power coefficient is below a certain threshold,the optimal activation ratio should equal the one to minimize the network average power consumption per area.Otherwise,the optimal activation ratio should be obtained according to the QoS constraints.It is further shown that universal frequency reuse(UFR)outperforms spectrum partitioning(SP)in terms of both energy efficiency and spectrum efficiency in the considered scenario.4)Optimal biased association scheme with non-uniform user distribution in HetNets.A practical scenario is studied where one cell is overloaded due to the cluster of users.By maximizing the mean user utility in the area of this overloaded cell and its neighboring cells,the optimal biasing factor can be obtained.It is found that in the scenario where the overloaded cell is fully surrounded by a macro cell,the optimal biasing factor arithmically decreases with the user's intensity of the overloaded cell.Numerical results demonstrate that the mean user rate of the overloaded cell and the whole network can be significantly improved by properly tuning the biasing factor of the overloaded cell.