Network Selection And Resource Allocation For Unlicensed LTE Systems

Due to the conflict between the continuously growing mobile traffic and the limited spectrum resource,mobile networks still face enormous challenges.To address these challenges,mobile networks can share the unlicensed frequency bands(such as the 5.8 GHz spectrum)which were traditionally occupied by Wi Fi systems.This emerging technology is called the Long Term Evolution-Unlicensed(LTE-U)technology.LTE-U can improve the system capacity and user experience of mobile cellular networks.Besides,compared with existing Wi Fi networks,the spectrum efficiency of unlicensed bands can also be improved.In order to deploy the LTE-U system,it is necessary to study the network selection and resource allocation of LTE-U and Wi Fi networks to balance the traffic load and ensure the fair and friendly coexistence of LTE-U and Wi Fi.In view of network selection and resource allocation,this paper first designs a centralized method to solve the multi-homing problem in the LTE-U and Wi Fi coexistence system.Multihoming is a technology that allows a user in the system to use both Wi Fi and LTE-U networks simultaneously.In this context,it is necessary to deal with whether a user should access a certain network and how much network resource should be occupied by this user.We aim to maximize the overall system throughput under the constraints of users' quality of service(Qo S)and power.By analyzing the model,we have derived the sufficient conditions that users only choose LTE-U or Wi Fi for transmission on unlicensed bands.Because this problem is non-deterministic polynomial(NP)hard,we propose an effective heuristic algorithm for network selection and resource allocation.This heuristic algorithm can achieve close performance compared with the optimal algorithm that is of exponential complexity.In addition to verifying the effectiveness of the algorithm,simulation results also show that compared with the traditional single-homing network,the multi-homing mechanism can improve overall system throughput.Secondly,this paper proposes a distributed method based on Minority Game(MG).This method is of low computational complexity and low signaling overhead,and it does not require any CSI information.In order to achieve load balance between Wi Fi and LTE-U networks,we formulate the network selection and user association problem as a minority game.We also derive the cut-off value in minority game and prove its uniqueness.In this work,we have studied the two coexistence mechanisms in LTE-U,namely listen-before-talk(LBT)and duty-cycle muting(DCM).We have achieved the Nash equilibria of the model.In addition to the single Wi Fi AP scenario,our method is also applicable to multiple Wi Fi APs scenarios.Finally,we propose a distributed method based on Q-learning for RAT selection and user association.This method is also on the basis of the minority game model.However,the minority game is played according to Q-learning rather than the classic pure strategy and mixed strategy.The users make network selection by means of reinforcement learning while seeking their own interests.Meanwhile,they self-organize to a state where the two networks reach load balance.Simulation results show that Q-learning can suppress the inherent "crowd effect" in minority games and improve the stability of the system.