Resource Allocation And Optimization In Energy-harvesting-based D2D Communication

With the significant growth of wireless cellular devices and its traffic demand,cellular network will face the challenges of spectrum scarcity and high energy consumption.The 3GPP network working group continues to promote effective wireless communication technologies to free up cellular networks from the spectrum constraints and to control the growing energy consumption.At the same time,the high expansion of mobile devices will accelerate energy consumption and bring about some hidden problems,such as environmental pollution(greenhouse gas emissions).Therefore,the rapid growth of wireless cellular devices and its traffic demand,as well as the environmental pollution and energy supply problems,drive the industry to explore better wireless communication solutions which have higher energy efficiency and spectral efficiency.Energy harvesting(EH)-based D2D Communication underlaying Cellular Network(EH-DCCN)combines strengths of D2D communication and energy harvesting so as to offload cellular traffic from base station with green communication.Thus,EH-DCCN can realize the demands of future network about high energy efficiency and spectral efficiency.So,EH-DCCN is an important part of future green wireless cellular communication.However,due to the fluctuation of environment and the immaturity of energy conversion technology,this new energy technology makes the available energy of devices have a strong randomness,intermittence and heterogeneity.Based on this characteristices,high efficient energy management schemes,which are the cornerstone to realize EH-DCCN,become one of most important research points in EH-DCCN.In order to satisfy the demands of different network transmission,this dissertation designs efficient multi-domain resource scheduling and allocation mechanisms by joint considering characteristics of D2D communication mode and energy collection.In this way we can promote the development of EH-DCCN.So,the main works and innovation points of this dissertation can be illustrated as follows:First of all,this dissertation investigates the resource allocation schemes in terms of the spectrum resource matching and the power allocation under the assumption that one cellular user(CU)and multiple EH-based D2D pairs(EH-DPs)share the same spectrum resource.Considering the joint influence of the available energy and channel condition on the resource allocation schemes,a channel-and energy-sensing resource allocation strategy is proposed.Due to the fact that the available energy becomes one of crucial impact factors to design the resource allocation policies in EH-DCCN,a sum-rate maximization problem of the whole cellular network with consideration of Quality of Service(QoS)and available energy constraints is formulated.The maximization problem is a non-concave mixed-integer nonlinear programming(MINLP)problem,which has been proved to be NP-hard.To solve the problem,we first relax it with a concave lower bound on the original problem and then obtain the theoretical performance of the lower bound by outer approximation algorithm(OAA).Moreover,a heuristic algorithm,an Energy-aware space matching approach(ESM),is proposed to acquire a suboptimal solution with low computational complexity.Finally,numerical simulation results indicate our considered resource allocation strategy is more effective than the strategy only based on channel state information under the EH-DCCN.Moreover,the performance in aspects of the sum rate and the matching probability shows that the ESM can approximately obtain the theoretical performance of the lower bound on the original problem under the scenarios with higher ratio of CU and EH-powered D2D numbers.Secondly,after the spectrum resource allocation,the energy availability will make the cellular spectrum occupation unbalanced in time domain.This dissertation designs the resource scheduling scheme in time domain and the corresponding power allocation algorithm to balance time domain spectrum load.And finally,the load balance scheme can reduce the mutual interference during users' transmission and improve the overall transmission energy efficiency.In the short term,the time-domain spectrum scheduling and its corresponding interference control model are established to optimize EH-DP total energy efficiency.To efficiently solve the short-term fractional non-convex programming problem,this dissertation proposes a two-layer convex approximation iteration algorithm(CAIA),which jointly combines Dinkelbach fractional programming and convex relax scheme.CAIA can obtain an approximate optimal solution under the offline situation which needs all status information including channel status information(CSI)and energy status information(ESI).It's hard to obtain the overall CSI and ESI during a period of time in practice.Thus,we propose a heuristic algorithm,which is named as time-division scheduling scheme(TDSS).With the CSI and ESI of the latest time slot,TDSS can simply balance the transmission requirements between two adjacent time slots based on the basal transmission demand consumption of user side.Simulation results verify that the energy efficiency of EH-DP can be further promoted by the short-term load balance scheme based on energy characteristics,which is more suitable for EH-DCCN application.Last but not least,considering the non-full buffer data traffic mode,this dissertation designs transmission scheduling algorithm based on Q-Learning(QL)under a more practical data and energy model.Facing multiple utility requirements,this dissertation establishes a long-term weighted average form of energy efficiency and delay under the circumstance that cellular user and EH-DP one-to-one sharing and one-to-more sharing scenarios,respectively.However,the above established long-term weighted average model is a non-convex multi-objectives optimization.Thus,this dissertation uses the centralized and distributed infinite-horizon discrete-time and finite-state Markov decision processes(DFMDP)for one-to-one and one-to-more scenarios,respectively.Because the convergence rate of QL is sensitive to the size of state-action space,the centralized and distributed infinite-horizon DFMDPs for one-to-one and one-to-more scenarios are improved to accelerate the convergence rate.At first,the size of state-action space of QL is lowered by eliminating superfluous information according to the characteristics of system model.And then,the users in the similar environment share their Q table to increase experience.Simulation results show that the improved QL can obtain the same performance as QL does with rapid convergence rate.Simultaneously,the maximal sustainable data arrival rate which can get proper energy efficiency and delay tradeoff performance is analyzed under different simulation scenarios.This conclusions are helpful for the practical energy efficiency and delay resource scheduling deployment in EH-DCCN scenario.