Millimeter Wave Communications for Future Mobile …

1 [ ] 17 Jul 20171 Improved Handover Through Dual Connectivityin 5G mmwave Mobile NetworksMichele Polese,Student Member, IEEE, Marco Giordani,Student Member, IEEE,Marco Mezzavilla,Member, IEEE, Sundeep Rangan,Fellow, IEEE, Michele Zorzi,Fellow, IEEEA bstract The Millimeter wave ( mmwave ) bands offer thepossibility of orders of magnitude greater throughput for fifthgeneration (5G) cellular systems. However, since mmwave signalsare highly susceptible to blockage, channel quality on any onemmWave link can be extremely intermittent. This paper imple-ments a novel dual connectivity protocol that enables Mobile userequipment (UE) devices to maintain physical layer connections to4G and 5G cells simultaneously. A novel uplink control signalingsystem combined with a local coordinator enables rapid pathswitching in the event of failures on any one link. This paperprovides the first comprehensive end-to-end evaluation of han-dover mechanisms in mmwave cellular systems.

2 The simulationframework includes detailed measurement-based channel modelsto realistically capture spatial dynamics of blocking events, aswell as the full details of MAC, RLC and transport to conventional handover mechanisms, the studyreveals significant benefits of the proposed method under paper was accepted for publication on the IEEE JSAC Special Issue onMillimeter Wave Communications for Future Mobile Terms 5G, Millimeter wave, multi-connectivity, han-dover, blockage, INTRODUCTIONThe Millimeter wave ( mmwave ) bands roughly corre-sponding to frequencies above 10 GHz have attracted consid-erable attention for next-generation cellular wireless systems[1] [5]. These bands offer orders of magnitude more spectrumthan conventional cellular frequencies below 3 GHz up to200 times by some estimates [1]. However, a key challengein delivering robust service in the mmwave bands is channelintermittency: mmwave signals are completely blocked bymany common building materials such as brick and mortar,[1], [6] [8], and even the human body can cause up to 35 dBof attenuation [9].

3 As a result, UE mobility, combined withsmall movements of obstacles and reflectors, or even changesin the orientation of a handset relative to the body or a hand,can cause the channel to rapidly appear or of the main tools to improve the robustness of mmWavesystems ismulti-connectivity[5]: Each Mobile device (UE oruser equipment in 3 GPP terminology) maintains connectionsto multiple cells, possibly including both 5G mmwave cellsand/or conventional 4G cells. In the event that one link isblocked, the UE can find alternate routes to preserve the con-nection. In cellular systems, this robustness is calledmacro-diversityand is particularly vital for mmwave systems [5].Michele Polese, Marco Giordani and Michele Zorzi are with the Departmentof Information Engineering, University of Padova, Padova,Italy Mezzavilla and Sundeep Rangan are with NYU WIRELESS, TandonSchool of Engineering, New York University, Brooklyn, NY (email: to implement multi-connectivity in the network layerfor mmwave systems remains largely an open problem.)

4 Cur-rent 3 GPP cellular systems offer multiple mechanisms for fastswitching of paths between different cells including conven-tional handover, multi-connectivity and carrier aggregation these methods are summarized below. However, mmWavesystems present unique challenges: Most importantly, the dynamics of mmwave channelsimply that the links to any one cell can deterioraterapidly, necessitating much faster link detection and re-routing [3]. Due to the high isotropic pathloss, mmwave signals aretransmitted in narrow beams, typically formed with high-dimensional phased arrays. In any link, channel qualitymust be continuously scanned across multiple possibledirections which can dramatically increase the time ittakes to detect that the link has failed and a path switchis necessary [10], [11]. One of the main goals of 5G is to achieve ultra-low la-tency [12] (possibly<1ms). Thus, service unavailabilityduring path switches must be kept to a ContributionsTo address these challenges, in this paper we expand themain results and findings of our previous works [10], [11],[13], [14], to provide the first global end-to-end comprehen-sive evaluation of handover and path switching of mmWavesystems under realistic dynamic scenarios, and assess howa dual-connectivity1(DC)

5 Approach can enable faster, morerobust and better performing mobility management particular, in [10] and [11] we proposed a novel multi-connectivity uplink measurement framework that, with thejoint effort of the legacy LTE frequencies, enables fair androbust cell selection, in addition to efficient and periodictracking of the user, suitable for several control-plane cellularapplications ( , we showed that periodic measurement re-ports can be used to trigger handovers or adapt the beams ofthe user and its serving cell, to grant good average throughputand deal with the channel dynamics experienced at mmWavefrequencies). In [13], we evaluated the tracking performance ofa user s signal quality considering real experiments in commonblockage scenarios, combined with outdoor statistical , in [14], we discussed two possible ways to integrate5G and LTE networks to improve the reliability of next-1 Although many of the ideas and techniques discussed in this paper applyto more general multi-connectivity scenarios, for concreteness in the followingwe will specifically refer todualconnectivity, in which a UE is simultaneouslyconnected to one 5G mmwave base station and one legacy LTE Mobile users, and described a preliminary ns 3simulation framework to evaluate the performance of extending our previous contributions, in this paper wealso propose: The use of a DC scheme to enable the base stations toefficiently track the UE channel quality along multiplelinks and spatial directions within those links.

6 In addition,to allow fast detection of link failures, we demonstratethat the uplink control signaling enables the network totrack the angular directions of communication to the UEon all possible links simultaneously, so that, when a pathswitch is necessitated, no directional search needs to beperformed (this approach greatly saves switch time, sincedirectional scanning dominates the delay in establishinga new link [15], [16]). The use of a local coordinator that manages the trafficbetween the cells. The coordinator performs both controlplane tasks of path switching and data plane tasks as atraffic anchor, at the Packet Data Convergence Protocol(PDCP) layer. In conventional cellular systems, thesecontrol and data plane functions are performed in the Mo-bility Management Entity (MME) and Serving Gateway(S-GW), which are often far from the cells. In contrast,the local coordinator is placed in close proximity to thecells, significantly reducing the path switch time.

7 The design of faster network handover procedures(namelyfast switchingandsecondary cell handover) that,by exploiting our DC framework, improve the mobilitymanagement in mmwave networks, with respect to thestandard standalone hard handover (HH) scheme. Theseprocedures are controlled by the LTE Radio ResourceControl (RRC) layer and, since the UE is connected tothe LTE and the mmwave eNBs, it is possible to performquick fallback to LTE with the fast switching command. A dynamic time-to-trigger (TTT) adaptation to enhancethe switch decision timing in highly uncertain link , we evaluate the proposed switching and handoverprotocols by extending the evaluation methodology we havedeveloped in [14], [17] [19]. The ns3-based framework weimplemented for this work makes it possible to use detailedmeasurement-based channel models that can account for boththe spatial characteristics of the channel and the channeldynamics arising from blocking and other large-scale events,which is important for a detailed and realistic addition, the simulator features a complete MAC layerwith HARQ, all the network-layer signaling, and an end-to-end transport protocol.

8 We believe that this is the firstexhaustive contribution which provides a global evaluationof the performance of a dual-connectivity architecture withrespect to a traditional standalone HH scheme in terms ofhandover and mobility management specifically tailored toa dynamic mmwave scenario. In particular, we simulatedthe user s motion in a typical urban environment. Separately,actual local blockage dynamics were measured and superim-posed on the statistical channel model, to obtain a realisticspatial dynamic channel model. We believe that this is thefirst work in which such detailed mmwave dynamic modelshave been used in studying study reveals several important findings on the in-teraction of transport layer mechanisms, buffering, and itsinteraction with physical-layer link tracking and handoverdelays. We also demonstrate that the proposed dual connec-tivity framework offers significant performance improvementsin the handover management of an end-to-end network withmmWave access links, including (i) reduced packet loss, (ii)reduced control signaling, (iii) reduced latency, and (iv)higherthroughput stability.

9 Moreover, we show that a dynamic TTTapproach should be preferred for handover management, sinceit can deliver non-negligible improvements in specific mobilityscenarios in which state-of-the-art methods Related WorkDual connectivity to different types of cells ( , macroand pico cells) has been proposed in Release12of LongTerm Evolution-Advanced (LTE-A) [20] and in [21]. How-ever, these systems were designed for conventional sub-6 GHz frequencies, and the directionality and variability ofthe channels typical of mmwave frequencies were not ad-dressed. Some other previous works, such as [22], consideronly the bands under6 GHz for the control channel of 5 Gnetworks, to provide robustness against blockage and a widercoverage range, but this solution could not provide the highcapacities that can be obtained when exploiting mmWavefrequencies. The potential of combining legacy and mmWavetechnologies in outdoor scenarios has also been investigatedin [23], highlighting the significant benefits that a mmWavenetwork achieves with flexible, dynamic support from LTEtechnologies.

10 Articles [24], [25] propose a multi-connectivityframework as a solution for mobility-related link failuresandthroughput degradation of cell-edge users, enabling increasedreliability with different levels of the literature on handover in more traditional sub-6 GHz heterogeneous networks is quite mature, papers on han-dover management for mmwave 5G cellular are very recent,and research in this field has just started. The survey in [26]presents multiple vertical handover decision algorithms thatare essential for heterogeneous wireless networks, while article[27] investigates the management of the handover processbetween macro, femto and pico cells, proposing a theoreticalmodel to characterize the performance of a Mobile user inheterogeneous scenarios as a function of various handoverparameters. However, these works are focused on low fre-quency legacy cellular systems. When dealing with mmWaves,frequent handover, even for fixed UEs, is a potential drawbackthat needs to be addressed.