Study Of Intelligent Wireless Communications Networks

The integration of intelligent technology with wireless communication network enables the network to perform better.The application of intelligent technology in wireless communication is everywhere,which can be summarized as four categories: hard-to-model problem,hard-to-solve problem,detection and estimation problem as well as implementation with unified modeling problem.Currently,the intelligent technology includes different kinds of improved heuristic algorithms,approaches based on game theory,reinforcement learning,deep learning,etc.For various problems,it is meaningful to study the application of intelligent technology in specific wireless communication scenario.Orthogonal Frequency Division Multiplexing(OFDM)technique and Multiple-Input Multiple-Output(MIMO)technique are the key technology of wireless communication.This dissertation focuses on the study of the application of intelligent technology in OFDM and MIMO systems.The main contributions of this dissertation can be listed as follows.(1)The robust auction mechanism for access permissions(ACP)trading in orthogonal frequency division multiple access(OFDMA)heterogeneous cellular networks is studied.Firstly,it is demonstrated that a critical challenge of designing such an auction mechanism is to ensure the economic properties(i.e.,truthfulness,individual rationality and budget balance).Secondly,instead of the single-sided scenario in most of prior works,a general market model where multiple femtocells can trade ACP with multiple MUEs is studied.Therefore,a truthful double auction for access permission(TDAP)is proposed.It is shown analytically that the auction mechanism is economically robust and computationally efficient.Moreover,through extensive simulation experiments,it is shown that TDAP can highly improve auction efficiency outperforming prior auction design.(2)The false-name-proof auction mechanism for ACP trading in OFDMA heterogeneous cellular networks is studied.Most of the related auction mechanisms can achieve strategy-proof(truthfulness)but can not resist the false-name bid(FNB)cheating,which means bidders can manipulate the market by submitting bids with multiple false names.The cheating behavior can improve the utility of cheater,but reduce the revenue of the auctioneer.In view of this situation,the challenge is designing a strategy-proof auction mechanism to resist FNB cheating.Therefore,a false-name-proof mechanism for ACP transaction(FMAT)is proposed.This mechanism is proved to be strategy-proof and false-name-proof.Besides,FMAT incurs a polynomial time complexity,which is better than most of existing strategy-proof auction mechanisms for access management in OFDMA heterogeneous cellular network.(3)A low-complexity optimization scheme for resource allocation is studied to balance the tradeoff between system capacity and proportional fairness in OFDMA based multicast systems.The major challenge is to solve the non-convex optimization problem with strict proportional fairness constraint.In this case,constrained team progress algorithm(CTPA)is proposed to solve this non-convex problem by allocating sub-channels to each multicast group based on sub-channel gains and proportional fairness constraint.Then a mapping power algorithm(MPA)is designed to guarantee strict proportional fairness with efficient power allocation.It utilizes the mapping relation between power and throughput.The advantage of CTPA-MPA is analyzed in three aspects: complexity,fairness and efficiency.(4)A power control approach based on the idea of multi-agent reinforcement learning in OFDMA heterogeneous cellular network is proposed.Subject to the quality of service(QoS)constraint of macrocell users(MUEs),the approach is to maximize the aggregate femtocells capacity based on Q-learning improved ant colony system for power control(QACSP),enabling the femtocell base stations(FBSs)to accomplish the power control by distributed multi-agent reinforcement learning with acceptable computational complexity.(5)For MIMO interference channels,two best-effort intelligent interference alignment(BEIA)schemes are proposed.Considering each transmitter-receiver pair has a constant number of data streams,the network selects the maximum number of interfering transmitters to align their signals given the feasibility conditions.In the event that not all interfering signals are aligned at each receiver,an upper bound of the average throughput is derived.It is shown that the proposed schemes have superiority over the traditional methods,such as time division multiple access(TDMA)and cluster interference alignment(CIA),in low and moderate signal-to-noise ratio(SNR)region in terms of average user throughput.Considering the performance of both average user throughput and minimum user throughput,the proposed max-min relative interference distance alignment scheme is better than the proposed scheme of equal interfering transmitters number alignment.