Performance Study Of Energy Harvesting Communication Systems Based On Network Calculus

With the evolution of communication techonologies,green commnucations and the Internet of Everything has become the main development trend of the future communication networks.Energy harvesting communication systems(EHCSs)are able to harvest various kinds of energy to sustain their data transmissions.As a result,the advantages of using EHCSs include abundant energy sources,reducing carbon emissions and increasing the flexibility of equipment deployments.Due to these advantages,EHCSs will play an important role in realizing the green commnucations and the Internet of Everything in the future.Typically,an EHCS harvests energy stochastically and finitely,which leads to its service characteristics more complex than those of a conventional communication system.Therefore,how well quality of service or how much traffic throughput can be guaranteed is an indispensable question while using EHCSs to transmit data.Stochastic network calculus,which is recognized as an excellent tool in performance analysis by academia,has great potential in answering this question.Motivated by this,this thesis applies stochastic network calculus to study the performance of EHCSs,with object to providing theoretical supports to such systems in parameter configurations and traffic access control.The main contents and contributions are as follows.(1)Study of Multi-Server Systems Based on Stochastic Network CalculusAS one kind of EHCSs,renewable energy harvesting communication systems(REHCSs)are some large communication notes including solar base stations and wind base stations which can transmit data simultaneously through several parallel channels.The service processes of such systems can be characterized by an analytical multi-server model.In order to study the performance of REHCSs,this part first extends the application scope of stochastic network calculus to the multi-server scenarios.Three different system service curves which characterize the system service process are derived under different conditions.After that,the service curve and the traffic characteristics are used together to study the traffic queueing delay performance,including the delay distribution and the mean delay.Finally,the accuracy of the analysis is verified with comparison of the exact results based on queueing theory under a M/M/N scenario.It is highlighted that the analytical results of the multi-server systems are absent in the state-of-art stochastic network calculus.Hence,this part supplements the theory of foundations for stochastic network calculus.More specifically,this part constructs a general framework to study the performance of multi-server systems and reveals how system performance is affected by various parameters,such as the number of servers,the traffic characteristics and etc.(2)Performance Study of REHCSsThis part first investigates the relationship between the energy depletion probability and the battery capacity,based on which the per-flow queueing delay performance is then studied under some typical traffic scheduling policies.Furthermore,in terms of the analytical results of the multi-server systems,the system service characteristics and the delay and battery capacity constrained per-flow throughput are studied respectively under a Rayleigh multi-channel scenario.Finally,the accuracy of the analysis is validated by simulations.The numerical results point out that the data packet size,the energy block size,the interdependence among the service process of each channel all have a great difference on the per-flow throughput of a REHCS.Compared to the existing studies which mainly focuses on the single-flow-single-channel REHCS,this part proposes a generally analytical framework to study the multi-flow-multi-channel REHCSs with consideration of the traffic scheduling policy and the stochastic features of multi-flow arrivals,multi-channel services,energy harvesting and energy consumptions.It is highlighted that this is the first time to study the relationship between the energy harvesting rate and the system performance in such a system in the literature.(3)Performance Study of Point-to-Point Wireless Powered Communication Systems(WPCSs)As another kind of EHCSs,WPCSs can be used in sensor networks and the Internet of Things.This part foucses on a point-to-point WPCS scenario.The service characteristics are analyzed for both uplink and downlink based on the fundamental feature of a WPCS that both uplink and downlink have to suffer a non-transmission phase respectively during each wireless charging cycle.After that,two upper bounds on the traffic delay distribution are obtained:One is tight with high calculation complexity while the other is simple but may be conservative.Moreover,the traffic throughput is derived under the traffic delay constraints.Finally,the accuracy of the analysis is validated by simulations.The numerical results confirm that the system performance depends on the length of the system transmission block,the system time allocation policy and the battery capacity.Compared to the existing studies,it is highlighted that this is the first time to study the delay and delay-constrained throughput performance of a WPCS in the literature.Specifically,a theoretical approach is proposed to analyze the uplink and downlink performance which can be sustained by a WPCS at the same time.Moreover,this part sheds new insights on the performance guarantee of the delay-sensitive WPCSs.(4)Performance Study of Multi-User WPCSsIn this part,the WPCSs is extended to a multi-user scenario.The buffer overflow probability is first studied on each node with the given buffer capacity,based on which the buffer-constrained throughput is then obtained.Additionally,when the buffer-constrained throughput is enforced as a performance requirement,the power allocation policy of the access point and the battery capacity configuration of each user are both investigated.Finally,the optimal system time allocation policy is proposed to achieve the maximum buffer-constrained throughput which can be sustained by each user at the same time.It is highlighted that this is the first time to study the data backlog performance under a WPCS scenario in the literature.Compared to other related resource allocation policies which are proposed in the literature,the one studied in this part takes buffer constraints into account.Hence,it is more applied to the buffer-limited WPCSs.