Throughput Region And Resource Allocation In Wireless Networks With Statistical QoS Provisioning

Future wireless communication networks are expected to support high traffic loads and to simultaneously provide reliable services with diverse quality of service(QoS)requirements.However,due to the time-varying nature of the underlying wireless channels,it is consid-erably difficult or even impossible to provide deterministic(or strict/instantaneous)QoS guarantees in realistic wireless networks.For example,in a point-to-point wireless commu-nication system experiencing Rayleigh fading,in which the deterministic QoS requirement is characterized by the strict minimum data rate,the maximum achievable data rate is also random by nature,even if the system always transmits data at its maximum power,resulting in that the strict data rate requirement may not be satisfied at some points.Consequently,in recent years,the statistical(or non-strict/time-averaged)QoS guarantee(e.g.,time-averaged rate,time-averaged queueing delay,etc.)has become an important alternative metric to provide the required levels of performance and quality to end-users over wireless networks.From the wireless communications perspective,it is extremely important to dynamically allocate resources(e.g.,power and bandwidth)to provide statistical QoS guarantees due to the scarce and time-variant channel capacity,and it is considerably interesting to know what a QoS guaranteed system with limited resource is capable of,or its "throughput region".As a result,considering the fact that the network resources,such as the available bandwidth and the maximum transmit power,are limited,in this thesis,we mainly focus on how to analyze the throughput region(e.g.,the maximum throughput)of statistical QoS guaranteed wireless networks and how to provide statistical QoS guarantees over wireless networks in an efficient way.The main contributions of this thesis are outlined as follows.Firstly,we investigate the throughput regions of wireless systems in the presence of random data arrivals and statistical QoS requirements,which are statistically characterized by the queueing-bound violation probability.By combining the concepts of effective capacity and effective bandwidth,we propose a unified analytical framework to investigate the through-put regions.Employing the proposed unified framework,we further acquire the explicit expressions of the throughput regions of QoS guaranteed wireless systems with the chan-nel state information(CSI)known/unknown at the transmitter.Specifically,the acquired throughput region is characterized by the first-and second-order statistics of the random data arrivals,and it is shown that the QoS requirements affect the throughput regions by the second-order statistics of the random data arrivals and the random data transmissions.Our theoretical analysis further demonstrates that the throughput region under QoS constraints is tighter than the conventional stable throughput region,and that the first-order statistic of the random data arrivals is sufficient to characterize the throughput region when the system can tolerate an arbitrarily long queueing delay.In particular,we prove that the queueing-bound violation probability decays exponentially with the queueing-bound,which is verified by our simulation results.Secondly,we investigate the resource allocation for the newly emerging wireless powered communication networks(WPCN)with statistical QoS requirements characterized by time-averaged queueing delay.In WPCNs,wireless powered devices(WPD)harvest energy from a power station(PS)via wireless energy transfer in the downlink and then communicate with an information receiving station(IRS)in the uplink.Each WPD is equipped with an energy buffer and a data buffer to store the random harvested energy and the bursty data ar-rivals,respectively.In order to minimize the time-averaged power consumption and also to provide time-averaged queueing delay guarantees,we propose a cross-layer Online Power and TIMe Allocation(OPTIMA)algorithm,which does not require any a prior distribution knowledge of channel states and data arrivals.Most importantly,we prove that the proposed OPTIMA algorithm achieves the power-delay tradeoff as[O(1/V),O(V)]with V being a system control parameter,and provides a significant method to control the power-delay per-formance and to guarantee the QoS in terms of the time-averaged queueing delay as required in system design.In other words,if the system is predefined to guarantee the statistical QoS in terms of the time-averaged queueing delay,it is only to select an appropriate system con-trol parameter V.Our simulation results verify the theoretical analysis and demonstrate the advantages of the proposed OPTIMA algorithm.Thirdly,we investigate the power efficient resource allocation in Orthogonal Frequency Division Multiple Access(OFDMA)cellular networks with statistical QoS requirements,which are characterized by the minimum time-averaged data rate.Specifically,the base sta-tion is supplied by both conventional grid and renewable energy.By jointly considering the harvested energy control,power allocation and sub-carrier assignment,we propose a sub-optimal online resource allocation(SORA)algorithm to minimize the time-averaged grid power consumption.Specifically,the performance of the SORA algorithm is theoretically deterministic:(a)the computational complexity of the SORA algorithm increases linearly with the number of users and sub-carriers;(b)the SORA algorithm is asymptotically opti-mal as the battery capacity is to be infinite.Most importantly,we develop an implementation architecture to take the SORA algorithm into practice,and also analyze the low implemen-tation costs(e.g.,low computational complexity,trivial signaling overhead,etc.).Our simu-lation results verify the theoretical analysis and demonstrate the advantages of the proposed SORA algorithm.