Spectrum Sharing Mechanism And QoS Guarantee Technology Of LTE-unlicensd Systems

Applying Long Term Evolution(LTE)technology in the unlicensed band,i.e.,LTE-unlicensed(LTE-U),combines cellular mobile communication technology based on centralized control and flat structure with license-free band resources to break through the bottleneck of the spectrum scarcity,relieve the pressure of data transmission over the licensed band,and effectively improve the capacity of the cellular network.However,the application of LTE-U technology faces two major challenges: the problem of fair coexistence with legacy networks in the unlicensed band,especially Wi-Fi,and the difficulty of guaranteeing the quality-of-service(Qo S)over the unlicensed band with stochastic and dynamic channel availability.This dissertation studies the fair coexistence mechanism of LTE-U with Wi-Fi and the Qo S guarantee technologies in LTE-U networks.For enabling the fair coexistence between LTE-U and Wi-Fi networks,i.e.,the fair access opportunities of all LTE-U and Wi-Fi users,the channel access mechanism and resource allocation of LTE-U networks are studied.The fair coexistence problem is formulated as a joint time,power and frequency allocation optimization problem for the throughput proportional fairness(PF)of all LTE-U users and Wi-Fi users.The resource allocation problem is further decoupled into two sub-problems,namely,the channel occupation time ratio PF-oriented optimization problem for the allocation of two networks' time ratio,and the LTE-U users' instantaneous transmission rate PForiented optimization problem for the joint subcarrier and power allocation of the LTE-U network.Then,an analytic model is proposed for calculating the channel occupancy time ratios of two networks and the optimal channel access configuration.The simulation results validate the accuracy of the proposed coexistence analysis model and the effectiveness of the resource allocation algorithm.For the coexistence of co-located and coordinated LTE-U and Wi-Fi,a downlink non-orthogonal multiple access(NOMA)-LTE-U channel access mechanism and the corresponding resource allocation are studied for fair coexistence and higher spectrum efficiency.A hybrid channel access mechanism based on grouped NOMA is proposed: the in-service Wi-Fi user and all the LTE-U users with their channel qualities better than the Wi-Fi user are assigned in the first group;all the remaining LTE-U users are assigned in the second group;The two groups access the channel at different time,which can avoid the inter-group interference.The users in the same group decode their received signals in a descending order of their channel qualities,which can avoid the intra-group interference.Therefore,the hybrid channel access mechanism can improve the spectrum efficiency of the unlicensed channel due to the NOMA gain and interference avoidance.Based on the new mechanism,the inter-group time ratio allocation and intra-group power allocation are formulated for achieving the max-min fairness for the throughput of all LTE-U users and the in-service Wi-Fi user.A two-level bisection algorithm is proposed to iteratively search the optimal power allocation scheme in each group and the optimal channel occupancy time ratio between two groups,which can achieve the fair coexistence between LTE-U and Wi-Fi networks.For a heterogeneous network with one LTE macro base station(MBS)and multiple LTE-U small base stations(SBSs),the user admission control is studied for exploiting the high reliability of the LTE MBS over the licensed channel and the offloading capability of the LTE-U SBSs over the unlicensed channel.An analytic evaluation model for LTE-U SBS's serving capability is proposed to capture the uncertainty of the available capacity and access delay of the unlicensed channel.A utility function for the users' throughput in different areas and channels is designed and maximizing the utility function leads to an optimal user admission control policy which can achieve a tradeoff between high throughput and load balancing.A suboptimal algorithm is proposed to distributedly perform the user admission control in each region and can solve the exponentially increased space complexity with the number of considered regions.For edge computing applications in the LTE-U-enabled Internet of things(Io T),the transmission scheduling and computing scheduling for Io T tasks are studied for high efficiency and low delay of task computing.The process of tasks' local queuing,local computing,task uploading,edge server queuing,edge server computing,and task scheduling decisions are modeled with a constrained MDP(CMDP).Under the condition of the limited budget of the edge computing cost,a task scheduling algorithm based on the constrained deep Q-learning is proposed.Leveraging a virtual reward function,the constrained task scheduling problem can be transformed into a unconstrained task scheduling problem,which can be solved by deep Q-learning based on the environmental feedback when the dynamics of channel availability and task generation pattern is unknowable.The framework of the deep Q-learning is customized for our problem,which can promote the learning efficiency and performance.