Beam Management for Millimeter Wave …

1 [ ] 30 Sep 2018 beam Management for Millimeter Wave BeamspaceMU-MIMO SystemsQing Xue and Xuming FangKey Lab of Information Coding & TransmissionSouthwest Jiaotong University, Chengdu 610031, ChinaEmail: XiaoCommunication Theory DepartmentRoyal Institute of Technology, Stockholm 100 44, SwedenEmail: Millimeter wave (mmWave) communication has at-tracted increasing attention as a promising technology for5 Gnetworks. One of the key architectural features of mmWave isthe use of massive antenna arrays at both the transmitter andthereceiver sides. Therefore, by employing directional beamforming(BF), both mmWave base stations (MBSs) and mmWave users(MUEs) are capable of supporting multi- beam simultaneoustransmissions. However, most researches have only considered asingle beam , which means that they do not make full potentialofmmWave. In this context, in order to improve the performanceofshort-range indoor mmWave networks with multiple reflections,we investigate the challenges and potential solutions of downlinkmulti-user multi- beam transmission, which can be described asa high-dimensional ( , beamspace ) multi-user multiple-inputmultiple-output (MU-MIMO) technique, including multi-user BFtraining, simultaneous users grouping, and multi-user multi- beam power allocation.

2 Furthermore, we present the theoreticaland numerical results to demonstrate that beamspace MU-MIMO compared with single beam transmission can largely improvetherate performance of mmWave Terms Millimeter wave (mmWave), beamspace MIMO,multi-user beamforming (BF) training, inter- beam INTRODUCTIONA ccording to Cisco forecast, global mobile data trafficwill increase sevenfold between 2016 and 2021 [1]. Recentresearches showed that mmWave communications, operatingin 30-300 GHz bands, are promising technologies for meetingthe explosive growth of mobile data demand. Compared withexisting microwave systems, mmWave systems are faced withtwo major challenges: severe propagation loss and sensitivityto blockage. To compensate for high propagation loss, direc-tional BF has been widely used as an essential technique toform a highly directional beam pattern with large antennagain.

3 Thanks to the short wavelengths of mmWave radiosranging from 10 mm to 1 mm, massive antenna arrays can bepacked into the limited dimensions of mmWave , with directional BF, it is possible to form multiplebeams at both mmWave transmitter and receiver sides inmobile networks. That is, mmWave systems are in fact able toprovide high-dimensional MIMO operations [2] [4] and canrealize spatial spectrum reuse at close distance [5]. However,most current work does not make full potential of instance, the work in [6] [9] was focused on singlebeam transmission scenarios, and the work in [5], [10] [12]considered the scenarios where only the transmitter side wasoperating with multiple beams. Moreover, since mmWaveradios have limited ability to diffract around obstacles ( ,human body), the connection between each pair of transmitterand receiver is vulnerable to blockage this context, aiming at increasing the achievable rate andmaintaining connectivity of mmWave mobile networks, weinvestigated the challenges and potential solutions (includingmulti- beam selection, cooperative beam tracking, multi-beampower allocation, and synchronization) associated with single-user multi- beam simultaneous transmissions ( , beamspaceSU-MIMO) in [13].

4 It is worth mentioning that the proposedscheme is only applicable to the short-range scenarios withmultiple NLOS paths ( , first or second order reflectionsfrom floor and/or ceiling in indoor scenarios). In order to fur-ther enhance the performance of mmWave systems, we extendour previous work to multi-user scenarios, namely beamspaceMU-MIMO, on the basis of existing research results. To thebest of our knowledge, there has been no work on this the communication environment with multi-user is morecomplex than that with single user, not only need we tofurther expand the strategies proposed in [13], but also willwe face some new challenges for implementing beamspaceMU-MIMO. For instance, due to the transmit beams selectedby different MUEs may be (partially) overlapped, the inter-user interference should be seriously considered in beamspaceMU-MIMO. This study mainly focuses on the issues of multi-user BF training, simultaneous users grouping, and multi-usermulti- beam power allocation in beamspace rest of the paper is organized as follows.

5 In SectionII, the network model and the basic idea of beamspace MU-MIMO are introduced. Section III first describes muti-user BFtraining and then proposes a multi-user grouping Section IV, the potential solutions of power allocation forbeamspace MU-MIMO are presented and analyzed. SectionV shows some numerical results to evaluate the proposedscheme. Finally, Section VI concludes the SYSTEMOVERVIEWIn this study, we consider a short-range indoor mmWave net-work with one reference MBS andUtotalsparsely distributedMUEs. LetRdenote the set of these MUEs. Meanwhile, boththe MBS and MUEs are equipped with massive antenna , with directional BF and space division technique, theyare capable of supporting multiple orthogonal beams simulta-neously and can realize spectrum reuse, as illustrated in the maximum number of beamsthat the MBS and MUEu(u R) can form, respectively,we generally havebMBSmax bumax.

6 LetQ(Q R) be theset of MUEs served simultaneously by the MBS andUbe thenumber of MUEs inQ, we have1 U bMBSmax. Furthermore,supposing thatbandbuare the number of operating beamsof the MBS and MUEuin actual transmissions, respectively,and considering that the transmit and receive beams are usedin pairs in mmWave networks, we haveb= u Qbu bMBSmax,where1 bu multi-user multi- beam simultaneous transmissionscheme investigated in this study can be described asbeamspace MU-MIMO defined as Definition 1. Fig. 1 showsan example of beamspace MU-MIMO in two-dimensional(2D) perspective. Note that the analysis is also applicabletothree-dimensional (3D) mode. For ease of analysis, similarto[10], [13], [14], we replace the MBS withUvirtual MBSs(vMBSs) located at the same position. Each vMBS servesdifferent MUEs with different transmit beam sets. Moreover,when1< bu bumax, the transmission mode between MUEuand its corresponding vMBS is beamspace SU-MIMO ( ,for MUE1 and MUE3 in Fig.)

7 1) and it is beamspace SU-SISO whenbu= 1( , for MUE2). In this context, beamspaceMU-MIMO can be defined as a set of beamspace SU-MIMOand/or SU-SISO technologies with space division 1( beamspace MU-MIMO): The beamspace MU-MIMO is defined as an mmWave communication mode that anMBS with multiple orthogonal beams can transmit simultane-ously to a set of MUEs, where each MUE is with one or moreoperating beams. That is, denotingQas the set of MUEs,bandbuas the number of transmitting and receiving beams of theMBS and MUEu(u Q), respectively, the multi-user multi- beam simultaneous transmissions can be termed asN NUbeamspace MU-MIMO, whereUis the number of MUEs inQ,Nis the total number of transmitting and receiving (T-R) beampairs between the MBS and the simultaneous transmittingMUEs,1 N min{b, u Qbu}.In order to implement beamspace MU-MIMO and, mean-while, to achieve optimal system performance, we face manychallenges as ) Multi-user BF training: Since only one transmit/receivedirection s link quality can be detected at a time in tra-ditional BF training ( , in ), the efficiencyof the optimal beam selection is generally very low.

8 Itmeans that the existing beam selection solutions are notentirely applicable to beamspace MU-MIMO. This studyutilizes the capability of supporting multiple beams bothat the MBS and MUEs to detect the quality of multiplelinks simultaneously, and thus to increase the efficiency ofmulti- beam selection for beamspace ) Inter-user interference coordination: The best transmit beamMBS (vMBSs)MUE1 MUE2 MUE3 NLOS linkLOS linkFig. 1. An example of 2D view of beamspace MU-MIMO (U= 3). Notethat, in order to make the figure clear, we do not show the side lobes selected by different MUEs may be (partially) over-lapped, , in Fig. 1, one NLOS link for MUE1 is inconflict with the LOS link for MUE2 over the transmitbeam. And the inter-user interference will be severe in thiscase. To address this issue, we can switch some MUEs initial selected T-R beam pairs with conflicting beams toa suitable candidate (if available), or assign the MUEswith the same conflicting transmit beam to different in different groups will be served in time ) Power allocation: Considering that the transmission per-formance of different links for different MUEs may varywidely, the appropriate power allocation strategies shouldbe seriously considered to maximize the achievable rate ofbeamspace beam /USERMANAGEMENT FORBEAMSPACEMU-MIMOIn this section, we first give an efficient multi-user BFtraining mechanism for beamspace MU-MIMO.

9 It is worthmentioning that the BF training (or beam steering) operationsare designed to determine the best T-R beam pair setNupair(u R) that best matches the LOS path and/or NLOS pathsbetween a vMBS and its corresponding MUE, hereafter calledvMBS-MUE. In this study, after the successful completionof BF training, directional BF is established. Moreover, toguarantee the network performance, we further adjust theinitial selected T-R beam pair sets by analyzing inter-userinterference. Table I summarizes the main notations usedthroughout the Multi-user BF TrainingThe multi-user BF training mechanism in this study aimsat improving the downlink performance of beamspace MU-MIMO. The corresponding strategies for uplink transmissionare left as our future work. For ease of illustration, we dividethe region of the MBS/MUEs into a number of transmit/receivesectors ( , orthogonal beam directions).

10 Meanwhile, weassume that MUEs can distinguish signals received fromdifferent beams. The proposed mechanism mainly consists ofthree phases and, moreover, the conceptual flow of the firsttwo phases is illustrated in Fig. 2. The details are as follows.(i) Transmit Training:In this phase, all MUEs are in thequasi-omni mode and the MBS scansntxtransmit sectors TABLE ISUMMARY set of MUEs within the coverage of the MBSQThe set of MUEs served simultaneously (Q R)UNumber of MUEs inQbMBSmaxMaximum number of simultaneous transmit beams at MBSbumaxMaximum number of simultaneous receive beams at MUEubNumber of the operating beams at the MBSbuNumber of the operating beams at MUEu(u Q)NuTXBest transmit beam set of MUEu(u R)NuRXBest receive beam set of MUEu(u R)NupairBest T-R beam pair set of vMBS-MUEu(u R)NucdCandidate T-R beam pair set of vMBS-MUEu(u R)NuOperating T-R beam pair set of vMBS-MUEu(u Q)CmThe set of MUEs with conflicting beamm(Cm R) Threshold of SNR (or SINR) tTransmitting beamwidth rReceiving beamwidthRuiTransmission distance of linkifor MUEu(i Nu)quality simultaneously withntxdirectional beams.