Tracking Mm-Wave Channel Dynamics: Fast Beam Training .
Tracking mm-Wave Channel Dynamics:Fast Beam Training Strategies under MobilityJoan Palacios1, 2 , Danilo De Donno1 , and Joerg Widmer112IMDEA Networks Institute, Madrid, SpainUniversidad Carlos III de Madrid, Madrid, SpainE-mail: {firstname.lastname}@imdea.orgsuch as link blockage, device rotation, etc., can cause considerable signal drop. To sum up, fast and efficient beamtraining/tracking strategies are of paramount importance tomaintain seamless connectivity in a mm-wave network withnode mobility.The design space of beam search proposals in the literaturecan be divided into three main categories: (1) sequentialscanning strategies [3], [4]; (2) adaptive algorithms employingantenna patterns with configurable beamwidth [5]–[8]; (3)parallel beam search with simultaneous, multi-direction scanning [9], [10]. The vast majority of these works concentrates,however, on static networks without investigating the impactof the training latency on the overall Quality of Service (QoS)of realistic networks with mobility. Within the state-of-theart solutions on this subject, a further subdivision can bemade on the basis of the employed mm-wave transceivers.Since traditional multiple-input multiple-output (MIMO) digital beamforming (DBF) is, at present, impractical at mmwave frequencies because of cost and power consumptionconstraints, analog beamforming (ABF) and hybrid analogdigital beamforming (HBF) represent the only feasible solutions. Using ABF [3], [4] provides poor performance fortwo main reasons. First, the constant amplitude and thelow phase resolution constraints of the mm-wave RF phaseshifters [11] give rise to antenna sectors with high sidelobesand reduced flatness, leading to imprecise beam training.Second, the use of a single RF chain allows for only onecommunication beam, thus resulting in reduced throughput andhigh-overhead beam search. In HBF architectures [5]–[10],the precoding/combining operations are divided between theanalog and digital domains, while using much fewer RF chainsthan antenna elements. The availability of multiple RF chainsenables parallel, multi-stream processing and simultaneousmulti-direction scanning.In this paper, we consider a scenario consisting of a fixedAP and a mobile UE, both equipped with a low-complexitymm-wave HBF transceiver and communicating with directional antenna patterns. Our overall objective is to maximizethe communication rate over time. To this end, we proposetwo strategies, a deterministic one for beam training anda probabilistic one for beam tracking, to rapidly estimatemultiple, suitable directions of communication between APand UE. Here, beam training is a beam search mechanismwithout any prior knowledge that explores the entire azimuthalAbstract—In order to cope with the severe path loss,millimeter-wave (mm-wave) systems exploit highly directionalcommunication. As a consequence, even a slight beam misalignment between two communicating devices (for example,due to mobility) can generate a significant signal drop. Thisleads to frequent invocations of time-consuming mechanismsfor beam re-alignment, which deteriorate system performance.In this paper, we propose smart beam training and trackingstrategies for fast mm-wave link establishment and maintenanceunder node mobility. We leverage the ability of hybrid analogdigital transceivers to collect channel information from multiplespatial directions simultaneously and formulate a probabilisticoptimization problem to model the temporal evolution of the mmwave channel under mobility. In addition, we present for the firsttime a beam tracking algorithm that extracts information neededto update the steering directions directly from data packets,without the need for spatial scanning during the ongoing datatransmission. Simulation results, obtained by a custom simulatorbased on ray tracing, demonstrate the ability of our beamtraining/tracking strategies to keep the communication rate only10% below the optimal bound. Compared to the state of the art,our approach provides a 40% to 150% rate increase, yet requireslower complexity hardware.I. I NTRODUCTIONThe fifth generation of mobile communications (5G) isenvisaged to deliver multi-Gbps wireless connectivity and toenable a plethora of new applications. It is well establishedthat achieving extremely high data rates is impractical withcurrently available 4G systems due to the heavily congestedand fragmented spectrum below 6 GHz. In view of this, thelarge amount of unoccupied spectrum in the millimeter wave(mm-wave) bands above 6 GHz becomes very appealing [1].Communications at mm-wave frequencies are challengingsince the channel suffers from severe path loss, atmosphericabsorption, human blockage, and other environmental obstructions [2]. The short wavelength of the mm-waves allows beamforming arrays with many antennas to be implemented in asmall form factor, thus providing sufficient link margin. On theother hand, highly directional communications complicate thelink establishment and maintenance between an Access Point(AP) and a User Equipment (UE). In fact, AP and UE mustperform a time-consuming beam training procedure in order todetermine the best directions of transmission/reception, whichincurs significant overhead (and waste of network resources).The problem is exacerbated in scenarios with mobility, sinceeven a slight beam misalignment or environmental changes,1
domain and that is carried out both in the initial access phaseand, periodically, during the AP-UE communication. Beamtracking, instead, is an ongoing estimation that, starting fromthe current steering directions, probabilistically infers howthey evolve due to node mobility. The main contributions ofthe paper are as follows: II. R ELATED WORKMost of the literature on mm-wave beam search focuses onstatic scenarios without user mobility [4]–[8], [10]. Such anassumption may lead to wrong conclusions about the actualperformance of the algorithms in real networks. A comparativeanalysis of initial access techniques in mm-wave networks ispresented in [4], where performance metrics such as detectionprobability and delay are analyzed. The problem of trackingthe AP-UE beams to handle the channel dynamics is leftas future work. The design of HBF codebooks relying onbeamforming vectors with different beamwidths is presentedin [5]–[8], where it is assumed that phase shifters with alarge number of quantization bits are available at mm-wavefrequencies. However, the design of high-resolution mm-wavephase shifters is extremely challenging [11]. Finally, the simultaneous reception from multiple beams to accelerate thebeam search is exploited in [10].To the authors’ knowledge, only very few works in theliterature address the problem of fast beam search in realistic,dynamic scenarios with node mobility. A smart beam steering algorithm for 60 GHz link re-establishment under usermobility is presented in [3]. The algorithm uses knowledgeof previous feasible sector pairs to narrow the sector searchspace, thereby reducing the associated overhead. Numericalresults show that the proposed strategy is very effective, butstill incurs non-negligible latency in complex scenarios withsignificant blockage. A temporal channel evolution model fornon line-of-sight (NLOS) mm-wave scenarios is presentedin [9]. HBF at both the AP and UE is considered and abeam tracking technique based on sequentially updating theprecoder and combiner is developed. However, in [9], theangle of arrival (AoA) and angle of departure (AoD) deviationsdue to mobility are modeled as very small uniform randomvariables, which are not appropriate to characterize actualmobility or significant, sudden changes in the channel due toobstacles. In [12], a linear dynamic system model to analyzethe occurring errors due to link blockage and device movementis proposed. Based on the model, the authors propose twoprobing protocols that are effective in identifying the beamerrors. However, no beam training/tracking strategy is implemented in order to find alternative antenna sector pairs oncethe beam errors are identified. Finally, it is worth highlightingthat none of the above-mentioned works [3], [9], [12] analyzesthe impact of the beam search accuracy and overhead on theevolution over time of the achievable rate under mobility.We design a two-stage beam training protocol that approaches the performance of an exhaustive search overall possible beam directions, but has very low latency anduses implicit feedback (i.e., it does not require a dedicatedfeedback channel). The key aspect of our beam searchstrategy is a particular HBF combiner matrix whichtakes a reduced number of sequential, multi-stream signalmeasurements to cover all the possible combinations ofantenna weights.We propose, to the authors’ knowledge for the firsttime in the mm-wave HBF context, a beam trackingalgorithm that is able to track the mm-wave channeldynamics without any training slots, but simply usingknown portions of the data packet (e.g., the preamble).To this end, we formulate a probabilistic optimizationproblem, solved by gradient descent, whose objectivefunction is designed so as to model the temporal evolutionof channel paths due to device movements. Note that thisproblem is quite different from the problem of MIMOchannel estimation using known pilot symbols.We develop a simulation framework to assess the performance of the proposed beam training/tracking strategiesand compare them against existing approaches in theliterature. Specifically, we propose and implement a fastprotocol for link establishment and maintenance underuser mobility which dynamically switches between beamtraining and beam tracking based on the real-time QoS.Our simulator integrates a ray-tracing tool to accuratelymodel the time-varying mm-wave channel, taking intoaccount blockage, ray clustering, and mobility effects,and guaranteeing spatial consistency over time.Numerical experiments show that the performance providedby our solution is very close to the optimal “oracle-based”algorithm. Furthermore, the high accuracy and reduced latencyoverhead characterizing our beam training/tracking strategiesresult in a significant rate increase over state-of-the-art solutions which in addition require higher complexity hardware.Compared to ABF solutions which share the disadvantageof converging towards only one communication beam, ourapproach based on HBF is capable of achieving multiplexinggains by simultaneously transmitting multiple parallel datastreams over different paths.We use the following notation in the paper. A is a matrix, ais a vector, and A denotes a set. kak2 is the Euclidean norm ofa, while kAkF , A , AT , AH , and A 1 denote the Frobeniusnorm, determinant, transpose, Hermitian, and inverse of A,respectively. [A]B,: ([A]:,B ) are the rows (columns) of thematrix A indexed by the set B, and I is the identity matrix. E[·]denotes the expectation operator and d·e the ceiling function.III. M OTIVATION AND SYSTEM MODELThe use of highly directional antennas with very narrowbeams at both the AP and UE complicates the mm-wave linkestablishment and maintenance. As for the link establishment,the 60-GHz IEEE 802.11ad standard [13] implements a timeconsuming beam training procedure based on an exhaustivesearch to find the most suitable directions of transmissionand reception. Once a connection is established, the linkquality degradation due to user mobility is handled throughbeam refinement procedures that search around the previous2
Fig. 2. Frame structure encompassing beam training/tracking and datatransmission. Data slots can be indifferently either downlink (DL) or uplink(UL) slots.Fig. 1. Block diagram of the AP-UE mm-wave transceiver architectureimplementing HBF.initial directions of transmission. Once the initial access isaccomplished, pure data frames with directional antenna patterns at both the AP and UE are used. As explained laterin §V, the beam tracking can be performed with pure dataframes, i.e., using known portions of just two data slots (onefor UE beam tracking and one for AP beam tracking) withoutrequiring any dedicated training slots. The allocation of atraining frame to perform a new full beam search from scratchcan be triggered periodically or when the link quality fallsbelow a certain threshold. Based on the work in [14], [15],we assume frames of duration T 10 ms, each divided into100 slots of duration Tslot 100 µs, a sufficiently small valueto ensure channel coherence at mm-wave frequencies.As experimentally demonstrated in [16], the mm-wavechannel between AP and UE is composed of “ray clusters”,each cluster carrying a fraction of the total power. DefiningTslot 100 µs as the time granularity of our system, we canexpress the MUE MAP channel matrix at each time slot as:rK LMAP MUE X Xαk aUE (θk )aH(1)H AP (φk )Lsector pair in order to determine a new combination of beamswith improved link quality. However, in large and crowdedscenarios with mobility, such procedures may fail to copewith high channel dynamics, which would necessitate fastmechanisms to scan a large angular domain (instead of justadjacent directions) to find alternative communication links.In case simply probing adjacent beams is unsuccessful, anew exhaustive beam search procedure has to be performed.This leads to a high latency which deteriorates the overallsystem performance. Motivated by this challenging problem,we propose two smart and efficient strategies, one for beamtraining and one for beam tracking, to accelerate the linkestablishment and maintenance between mm-wave devices inmobility scenarios.We consider a mm-wave system with one fixed AP andone moving UE, both featuring the same HBF architectureconsidered in [5], [6], [8]–[10] and depicted in Fig. 1. TheAP is equipped with a uniform linear array (ULA) of MAPisotropic radiators connected to NAP RF transceiver chainsthrough a network of analog/RF phase shifters. The numberof antennas and RF transceiver chains at the UE side isMUE and NUE respectively. The HBF transceiver configurationallows AP and UE to communicate via NS data streams,with NS min(NAP , NUE ). To this end, the AP applies anNAP NS digital baseband (BB) precoder PBB followed byan MAP NAP RF precoder, PRF , to the symbol sequence tobe transmitted. The transmit power constraint is ensured byimposing k[PRF pBB ]:,i k22 1, for i 1, 2, ., NS . The final APprecoder is then given by the MAP NS matrix P PRF PBB .The transmitted signal passes through the MUE MAP channelmatrix H and impinges on the UE antennas together withwhite noise. Since the UE also implements HBF, it is able toconcurrently receive NS streams of data. To do that, it appliesa MUE NUE RF combiner CRF followed by a NUE NS digitalbaseband combiner CBB . The final UE combiner is given bythe MUE NS matrix C CRF CBB .We assume that AP and UE communicate using the framestructure in Fig. 2. Two different types of frames can beallocated: (i) beam training frames, which contain both trainingand data transmission phases, and (ii) pure data frames. Inthe initial access procedure, the allocation of a training frameis mandatory, since AP and UE need to determine suitablek 1 1where K is the number of clusters, L is the number of subpaths per cluster, aUE(AP) (·) is the ULA response vector at theUE (AP) whose expression can be found in [8, Eq. (3)], andαk is the complex gain on the -th sub-path of the k th cluster,which includes path loss, Doppler shift, and delay spreadeffects. The variables θk [0, 2π] and φk [0, 2π] are the thAoD/AoA of the k th cluster at the UE and AP respectively.In this work, we assume channel reciprocity [5], [6], that is,the AP AoDs in the downlink correspond to the AP AoAs inthe uplink. The same applies to the UE as well. Note that,in order to simplify the notation, we consider the AP andUE implementing horizontal (2-D) beamforming only, whichimplies that all scattering happens in the azimuthal domain.Extension to planar antenna arrays and, therefore, to 3-Dbeamforming is straightforward.IV. P SEUDO -E XHAUSTIVE B EAM T RAINING (PE-T RAIN )The use of directional antennas for mm-wave communication requires that AP and UE find suitable directions oftransmission, both in the initial access phase and, periodically,during the communication. As illustrated is Fig. 2, if nofeedback channel is available, two separate stages are requiredin the beam training phase, namely UE beam training (using3
downlink training sequences from the AP) and AP beamtraining (using uplink training sequences from the UE). Inthe following, we propose a pseudo-exhaustive beam training(PE-Train) protocol which is able to search over all possiblebeam directions with very low latency overhead. It usesomnidirectional transmission at the AP for UE beam training,and simultaneous, multi-stream transmission over the bestestimated directions at the UE for AP beam training.to channel-only information. The UE can now post-process themeasurement Ŷ to obtain:H 1Y AHŶUE (W )where AUE [aUE (θ1 ), aUE (θ2 ), ., aUE (θN )] is a spatial filter matrix and θi 2πiN , i 1, 2, ., N , is a set ofN equally spaced discrete angles covering the 360 azimuthal domain. As evident from Eq. 4 and from the ex2pression E[Y] Pt AHUE Hpo , the N 1 vector [Y]i , fori 1, 2, .N , contains the expected signal power impingingon the UE from each angular direction. Such information canbe directly used by the UE to estimate its Lest most suitable(i.e., the most powerful) directions of transmission/receptionH 1θ [θ1 , ., θLest ]. Since the product AHin Eq. 4 canUE (W )be precomputed and stored in the UE memory, the computational cost to estimate Y is just a matrix-vector multiplication.A. Stage I: UE beam trainingWe consider a mm-wave AP with the HBF architecture inFig. 1 performing omnidirectional transmission of a trainingsequence s[t], for t 1, 2, ., Ts , in a reciprocal channel.Arranging s[t] into the 1 Ts row vector s, the MUE Tsdiscrete-time signal R impinging on the UE antennas becomes:R pPt Hpo s N(4)(2)B. Stage II: AP beam trainingwhere Pt is the transmit power, H is the channel matrix,po [1, 0, 0, ., 0]T is the MAP 1 omnidirectional precodingvector used at the AP, and N is a MUE Ts matrix withindependent, Gaussian-distributed complex noise with meanzero and variance σ 2 . The lack of a dedicated RF chain foreach antenna makes it impossible for a HBF transceiver todirectly access R. In fact, R is inevitably processed by ahybrid combiner which compresses it into a reduced dimensionspace, with consequent loss of information. We tackle thisproblem with the following strategy. First, we design aneasily invertible, orthogonal MUE MUE matrix W (e.g.,a Hadamard matrix) representing a basis for the full spaceof antenna configurations. Then, since we cannot directlyapply W to R because of the HBF limitations, we performmultiple, consecutive measurements, each time using as hybridcombiner a different sub-matrix of W. Thanks to the propertiesof W, we can reconstruct, at the end of the procedure, anestimated version of R and process it through a spatial filtermatrix to derive the received signal power from each angulardirection.Specifically, we build W such that the elements of NUE W belong to the feasible set of phase-shifter weights.For example, in the case of 2-bit phase shifters, the elementsof NUE W can assume only four values, namely 1 and j. Then, we divide W into NW dMUE /NUE e sub-matriceswith dimensions MUE NUE , i.e., W [W1 , W2 , ., WNW ].For each Wi , with i 1, 2, ., NW , we build the RFNUE Wi and the baseband combinercombiner CRF,i CBB,i INUE / NUE , where INUE is the NUE NUE identitymatrix. The overall hybrid combiner Ci CRF,i CBB,i Wi isthen applied to R at NW successive instants in order to extractthe following signal measurements:pYi W H(3)i ( Pt Hpo s N), i 1, 2, ., NWAt the end of Stage I, the UE initiates
millimeter-wave (mm-wave) systems exploit highly directional communication. As a consequence, even a slight beam mis- . maintain seamless connectivity in a mm-wave network with node mobility. The design space of beam search proposals in the literature . A comparative analysis of initial access techniques in mm-wave networks is presented in .
Motive Wave. It is a five wave trend but unlike a five wave impulse trend, the Wave 4 overlaps with the Wave 1. Ending Diagonals are the last section ("ending") of a trend or counter trend. The most common is a Wave 5 Ending Diagonal. It is a higher time frame Wave 5 trend wave that reaches new extremes and the Wave 3:5 is beyond the .
Business Ready Enhancement Plan for Microsoft Dynamics Customer FAQ Updated January 2011 The Business Ready Enhancement Plan for Microsoft Dynamics is a maintenance plan available to customers of Microsoft Dynamics AX, Microsoft C5, Microsoft Dynamics CRM, Microsoft Dynamics GP, Microsoft Dynamics NAV, Microsoft Dynamics SL, Microsoft Dynamics POS, and Microsoft Dynamics RMS, and
Wave a and Wave c are constructed of five waves as Elliott originally proposed. As opposed to the five wave impulse move in Elliott’s original version that could form either a Wave 1, Wave 3, Wave 5, Wave A or Wave C the harmonic version can only f
So, the wave 1, wave 3 and wave 5 are parts of impulsive wave in upward direction. [6] Though Elliott waves follow many rules but three basic rules are followed by each wave to interpret Elliott wave. These guidelines are unbreakable. These rules are as follow: Rule 1: Wave 2 is not retracted more than 100% of wave 1.
So, the wave 1, wave 3 and wave 5 are parts of impulsive wave in upward direction. [2] Though Elliott waves follow many rules but three basic rules are followed by each wave to interpret Elliott wave. These guidelines are unbreakable. These rules are as follow: Rule 1: Wave 2 is not retracted more than 100% of wave 1.
Microsoft Dynamics 365 for Operations on-premises, Microsoft Dynamics NAV, Microsoft Dynamics GP, Microsoft Dynamics SL, Microsoft Dynamics AX 2012 or prior versions, or Microsoft Dynamics CRM 2016 or prior versions. This guide is not intended to influence the choice of Microsoft Dynamics products and services or provide technical specification.
This guide is designed to improve your understanding of how to license Microsoft Dynamics 365, Business edition. This document does not apply to Dynamics 365, Enterprise edition, Microsoft Dynamics NAV, Microsoft Dynamics GP, Microsoft Dynamics SL, Microsoft Dynamics AX 2012, or Microsoft Dynamics CRM 2016 or any other prior version.
ASP.NET v4.0 either authenticates or encrypts view states, but it does not apply both operations at the same time. Forms Authentication Tickets: Since ASP.NET aims to be-come a rapid web development framework, it provides built-in solutions for many common problems in web devel-opment. One of them is user account support. Providing user account support for any site involves the same set of .
