Вы находитесь на странице: 1из 5

© 2014 IEEE

Presented at IEEE VTC Spring 2014, 18–21 May 2014, Seoul, Korea
Available in IEEE Xplore®
Handoff Rates for Millimeterwave 5G Systems
Anup Talukdar, Mark Cudak, Amitava Ghosh
Radio Systems Research, Technology and Innovation
Nokia Solutions and Networks
Arlington Heights, Illinois, USA
{Anup.Talukdar, Mark.Cudak, Amitava.Ghosh}@nsn.com

Abstract—Millimeterwave band is a promising candidate for 5th might be quite rapid as the user device moves through the
generation wireless access technology to deliver peak and cell- network. Moving obstacles, hand motion and changes in
edge data rates of the order of 10 Gbps and 100 Mbps, orientation may all contribute to multiple, successive handoffs.
respectively, and to meet the future capacity demands. The main This paper examines many of these scenarios and attempts to
advantages of the millimeterwave band are availability of large provide a first order estimate of typical handoff rates for 5G
blocks of contiguous bandwidth and the opportunity of using mmWave cellular. This handoff rate estimate will help guide
large antenna arrays composed of very small antenna elements to network protocol design for 5G cellular.
provide large antenna gains. The line-of-sight operation
requirement in this band, driven by its unique propagation The rest of this paper is organized as follows. In
characteristics, makes it necessary to build the network with Section II, we mention the related works. In Section III, we
enough redundancy of access points and the users may have to describe deployment scenarios for a mmWave system. In
frequently handoff from one access point to another whenever its Section IV, we investigate the handoff events in a mmWave
radio link is disrupted by obstacles. In this paper we investigate system and develop a set of fundamental types of handoffs
the handoff rate in such an access network. Based on analysis of categories and estimate the handoff rates for these types of
various deployment scenarios, we observe that, typical average handoffs. In Section V we present the conclusions.
handoff interval is several seconds, although for certain type of
user actions the average handoff interval can be as low as 0.75 II. RELATED WORK
sec.
Recently there has been a strong research interest on
Keywords—MmWave, 5G, handoff. beyond-4G or 5G wireless access networks in the mmWave
band. A framework for Beyond-4G local area network in the
I. INTRODUCTION millimeter wave band has been proposed in [1], which can
Millimeterwave (mmWave) frequency bands have been deliver peak and cell-edge data rates greater than 10Gbps and
identified as a promising candidate for 5th generation (5G) 100Mbps, respectively, with sub-millisecond latencies. Results
cellular technology [1][2]. Spectrum in traditional cellular from measurements studies of the propagation characteristics at
bands, below 6GHz, is finite and as cellular data traffic demand 60GHz and 81-86 GHz bands have been presented in [5] and
continues to grow new frequency bands must be considered. [7], respectively. To the best of our knowledge, there has not
Unlike traditional cellular bands, large blocks of contiguous been any prior research on the investigation of handoff
spectrum may be allocated at mmWave bands allowing for characteristics in a mmWave access network.
bandwidths on the order of GHz or more. Moreover, the III. DEPLOYMENT SCENARIO OF A MMWAVE NETWORK
mmWave bands allow for multi-element antenna arrays
composed of very small elements, on the order of IC chip The primary deployment scenarios of mmWave networks
scales, providing large antenna gain and sufficient power are expected to be urban and semi-urban hotspot environments
output through over-the-air power combining [3]. This with high data rate and capacity demands and it will be
combination of large bandwidths and novel device underlaid within a 4G LTE/LTE-A deployments. In the
architectures allows mmWave cellular to provide peak rates on following, we discuss typical deployment scenarios.
the order of 10 Gbps and ample capacity to meet future A. Street-side
demand.
MmWave access point deployments on the street side are
However the propagation characteristics in the mmWave targeted to provide services to pedestrians on the street, street-
band are more challenging than traditional cellular. Diffraction side open-air restaurants customers etc. In a typical
at mmWave bands is effectively non-existent and propagation deployment, APs will be installed on street poles or other
behaves similar to visible light. Transmission through most structures at an elevated position to avoid large scale blocking
objects is diminished where foliage and other common by obstacles, but not so high that the short range of the air-
obstacles can produce severe shadowing [4]. Reflective power, interface affects the links. To ensure enough coverage, APs
on the other hand, is improved offering new opportunities for need to be deployed at intervals of street blocks, possibly on
completing the link [5]. With this in mind, a cluster network both sides of the street.
concept is envisioned where a set of coordinated access points
(AP) work together to provide ubiquitous coverage through AP
diversity [1][6]. In the event of shadowing, one AP will
rapidly hand off to another AP in the cluster. These handoffs
B. Campus/courtyard the paper is to analyze a typical deployment scenario which is a
In a university or college campus, a possible deployment representative of the handover category.
area is the courtyards in front of the classroom buildings. Other A. Fixed obstacles
similar scenarios are open air tourist spots.
A typical deployment scenario where handoff events are
C. Stadium generated by fixed obstacles is a tree-lined street. In Figure 1-
In a stadium users may use mmWave high data rate 2, example handoff events are shown in such a deployment for
services for augmented reality or shared experience. Access various numbers of obstacles in a 100m long street block. In
point deployments should be able to provide sufficient Figure 1, the trivial case of one single obstacle located
coverage to users covering the various action moments of the midway between the two access points, AP1 and AP2 is
events. illustrated. Initially, the user is connected to AP1 at time
instant t=0. As the user walks down the street at the speed of
IV. HANDOFF EVENTS IN A MMWAVE AIR-INTERFACE 3km/h, its LOS to AP1 is obstructed by the obstacle at time
DEPLOYMENT t=63.5s. At this point, the user is in LOS of AP2 and a
Handoff of a user terminal from one AP to another is handover occurs from AP1 to AP2. The user then continues to
performed when its radio link to its current AP either fails or remain connected to AP2 for the rest of the block. In Figure 2,
too weak to provide minimum data rate. It is also possible that the handoff events are illustrated for a more complex scenario
handoffs are initiated by the network for the purpose of load of 3 obstacles. The user alternates between AP1 and AP2 as it
balancing. In this paper, the primary focus is on the handoff moves along the sidewalk. The obstacle O1 generates a
events due to link failure or very poor link condition in a handoff at t=16.5sec; the front edge of the obstacle O2
system comprising only mmWave base stations. We didn’t generates two handoffs: first, at t=42.4s when the terminal
consider handoff’s from mmWave to wide area systems (and loses LOS to AP2 and handoffs to AP1, secondly at t=63.5s
vice-versa) like LTE/LTE-A. when it loses the LOS to AP1 and handoffs back to AP2.
In a mmWave air-interface, radio link failures for a terminal Similarly the front edge of O3 also generates two handoffs. In
in the coverage area of an AP can happen for two reasons: a general scenario with a large number of obstacles in the
obstacles between the terminal and the AP, and user terminal street block, it can be concluded that two handoffs occur for
orientation is such that its antenna is not pointed to the AP to each obstacle.
make a link. Although, non-LOS reflective links may exist 100m
between the terminal and the AP, they can be significantly AP1 AP2

weaker than the LOS link (by 15 dB or larger[5]). For the


handoff rate estimation in this paper we simplify the analysis obstacle
20m
and assume that handoff occurs whenever the LOS link is 3m
3Km/h
disrupted. UE

The characteristics of the handoff events in various t=0s t=63.5s t=120s


deployment scenarios, thus, are determined by the nature of the Figure 1 Handoff events for one fixed obstacle
obstacles and the user behaviors impacting the terminal 100m

orientation. For example, in a street-side, obstacles can be AP1 AP2

trees lining the avenue, passing vehicles, pedestrians; in a


campus/courtyard environment, the primary obstacles are the 20m
3m
pedestrian. The primary sources of handoff due to user terminal O2 O3
O1 UE 3Km/h
orientation are: user hand movements, user rotation etc., which
can happen in all of the deployment scenarios considered t=0s t=16.5s t=42.4s t=63.5s t=89.4s t=110.6s t=120s
above. The different handoff events in a mmWave access Figure 2 Handoff events for 3 fixed obstacles
network can be categorized into the following four In order to get an analytical estimate of the handover rate,
fundamental types: we consider a deployment scenario of obstacles of
1. Fixed obstacles: Handoffs due to users moving past infinitesimally small width (Figure 3) so that outage does not
fixed obstacles occur. It is assumed that, whenever the user’s LOS to its
current AP is obstructed, it rapidly switches to another
2. Moving obstacles: Handoffs caused by passing vehicles supporting AP and then eventually return to the original AP
obstructing the AP
when it encounters an obstacle to the supporting AP.
3. Rotation: Handoffs caused by user hand movement and
rotations
4. Human obstacles: Handoffs due to blocking of APs by
pedestrians.
In the rest of this paper, we analyze these handoff events and Figure 3 Handoff rate estimation for fixed obstacles
provide an estimate of the handoff rates (or its inverse metric The number of obstacles encountered by the user which
handoff interval) for each of them. The methodology used in trigger handoffs, depends on the obstacle position in the street
and the obstacle spacing, defined as the center-to-center gap Assume that the user is initially connected to the access point
between adjacent obstacles. Based on the observation that AP2. The truck TR1 blocks the LOS of the user to AP2 at time
each obstacle generates two handoff events, the average t=t1 forcing the user to handoff to AP1. Then at time t=t1+t2,
handoff interval, Ihandoff, can be estimated by the following TR1 blocks the LOS of the user to AP1. Thus, it can be
equation. observed that, two handoff events occur for every passing
truck (except at the initial ramping up period from an empty
I handoff =
1 ….. (1) street in cases with very high truck arrival rate). Thus it can
2v  h
* 1 −  be concluded that the average handoff rate is 2/T handoff/sec
s  W where T is the truck inter-arrival time in sec.
where,
v = UE speed
s = center to center obstacle spacing
h = position of the obstacle
W = street width
Using the Equation (1), the handoff rates at two different user
speeds, 3Km/h (typical pedestrian speed) and 25Km/h (user in
a moving vehicle) for various obstacle spacing are computed
and shown in Figure 4. It is observed that, at pedestrian user
speed handoff intervals are of the order of several seconds; for
example, when the obstacles are spaced 15m apart, a handoff
occurs about every 11.6sec. Sub-second handoffs can happen
only at vehicular speed or very closely spaced obstacle
Figure 5 Handoff events for a moving obstacle
scenario. For obstacles closer to the access points, the handoff
In Figure 6 the handoff rate is shown for various average
rates are lower. The analytical estimation of the handoff rates
residual density of trucks in the street block when the truck
matches well to simulation results as observed in Figure 4
speed is 30Km/h. It can be observed that, in a typical scenario
5 of one truck in every 100m block the average handoff interval
3Km/h, analytical 25Km/h, analytical
3Km/h, simulation 25Km/h, simulation
is 6sec.
1.8
handoff rate (#handoff s/sec)

4 3Km/h, sim, obs. width=5m 3Km/h,analyt,obs. pos=10m


handoff rate (#handoffs/sec)

1.5
3
1.2

2 0.9

0.6
1
0.3

0 0
0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1
obstacle spacing (m) truck density (#trucks/meter)
Figure 4 Handoff rates for fixed obstacles Figure 6 Handoff rates for moving obstacles in a street
In the above analysis we have not accounted for
deployment of APs on the same side of the street as the user, C. Rotation
which could trivially eliminate all handoffs due to these The rotational motions which may cause link disruption
obstacles; the main objective of the analysis in this paper was and handoff can fall into one of the following four classes:
to obtain first order estimate of handoff rates without 1. Reflexive Movement
considering scenario-specific strategies to optimize the 2. Live action
system. 3. Incidental Motion
4. Unattented
B. Moving obstacles
In the following each of these classes of handoff events are
In this category the primary obstacles which cause discussed in details.
handoffs are moving vehicles, such as large trucks that are Reflexive Movement:
high enough to block the LOS between the AP and user; cars This type of movements occurs when the user reacts to
and bikes are not likely to interrupt the link because the APs external events. The user may make a turn to interact with his
are typically mounted at an elevated position. The handoff virtual or augmented world when surprised by an event. If the
events due to passing trucks are illustrated in Figure 5. angle of rotation is large enough, it may result in a handoff.
The typical application scenario can be shared experience. D. Pedestrian obstacles
These are occasional events and the angle of rotation can Since the APs are most likely to be mounted at a position
range from 90deg to 180deg, resulting in one or two handoffs. significantly higher than typical human height, pedestrians
Live Action: only close to the user are likely to block the link between the
Users filming live actions can generate frequent handoff terminal and the AP and may cause a handoff. It is a common
events. Typical application scenario is shared experience. For perception that pedestrian movements can be random;
example, viewers of various sports events, such as tennis however, depending on the deployment location,
match, little league game, cricket or a soccer match in a characteristics of handoff occurrences can be different. In this
stadium may use his/her 5G-enabled handheld device or paper the handoff events generated due to pedestrian blockings
Google glass to relay the live events in the stadium to their are classified into three categories: single encounter, group or
remotely located peers. Here we will estimate the handoff rate pack and dense crowd.
for two representative events: a tennis rally and soccer goal Single encounter:
kicks and volleys. These types of events commonly occur in the street-side
The length of a standard tennis court is 120ft. Considering deployments. Although, random movements of pedestrians are
the case when a viewer is seated about 65ft distance from the conceivable in a street as illustrated in Figure 7a and 7b, in the
center of the court and at the midpoint of the court length, it most common scenario, movements of the pedestrians are
can be estimated that, when the viewer follows a shot from streamlined and restricted to the street-side, as shown in
one extreme of the court to the other extreme, the angle of Figure 7c. In the most likely scenario a user may be obstructed
rotation is about 85deg. From the Youtube video clip of a by a pedestrian passing from the opposite direction or possibly
tennis rally in the Australian Open tournament[8], it was found by someone walking in the same direction and passing him/her
that 72 volleys in the rally were completed in 95sec. Thus the at a faster speed. The handoff rate in this scenario depends on
average handoff interval was about 1.3sec, assuming that one the pedestrian density in the sidewalk and can be estimated in
handoff is required to cover each volley. a way similar to the moving obstacles scenario.
The length of a standard soccer field is 288ft and the
typical distance of a viewer from the center of the field is
about 132.5ft. The angle of rotation to follow a goal kick from
one end of the field to the other is about 94deg. The speed of
the ball can range from 75km/h to 103Km/h. based on these
data, it can be estimated that for following long goal kicks and
return volleys, a burst of handoffs may be required at an
average interval in the range of 3.1sec-4.2sec.
Incidental Motion:
These types of rotational motions are associated with either
calm movements during low-stress activities, e.g. fidgeting, or
Figure 7 Pedestrian obstacle scenarios in a street
purposeful movements while performing a task. Typical In Figure 8, the handoff rates for various pedestrian
applications scenario in which such motion can happen are densities are shown. It is observed that handoff rate exceed 1
shared experience, augmented reality. A typical turn has an
handoffs/sec when the pedestrian density exceeds 0.6 (i.e. 3
angle of rotation of 90deg. Based on a few measurements an pedestrians every 5 meter stretch of the street), which
estimated rotational speed was found to be 20 rotations/sec. represents an overcrowded scenario; for a typical pedestrian
Assuming that a handoff is required for every 90deg turn, the density of 0.2 (one pedestrian every 5 meter) average handoff
estimated handoff interval is about 0.75 sec on average. interval is about 3sec.
Unattended: 1.8
A link failure can happen due to inadvertent movements of
handoff rate (#handoffs/sec)

the user that results in change in orientation of the terminal. It 1.5


is also possible that the user inadvertently stores it away in his
1.2
pocket, brief case or just set it on the table after initiating a
communication. Typical application scenario in which these 0.9
types of movements can happen are seamless cloud storage,
grab’n go purchases etc. It is possible to have prolonged 0.6
blockages in the scenario when the terminal is in an
orientation or location such that it is out of the radio link 0.3
coverage of all nearby APs; thus handoff may not happen at
all. The expected scenario in such a condition is that, the very 0
high data rate and ultra-low latency of the 5G network ensures 0 0.2 0.4 0.6 0.8 1
the service is completed before the terminal goes out of pedestrian density (#pedestrian/meter)
coverage. Figure 8 Handoff rates for pedestrian obstacles
Group or pack: obstacles, such as trucks in a street, can be about 6sec. In a
Users in a group (e.g. tourists) can often obstruct other dense crowd scenario, handoff intervals due to pedestrians
users’ (in the group) access to an AP. In this scenario, users blocking the LOS to the APs can exceed 1 handoff/sec only
and obstacles can be moving at similar speed (~1-3Km/h) in when the residual crowd density is very high. Among all the
the same direction. In the rare situation persistent blocking is handoff events, certain types of rotational motions appear to
possible. The most likely application scenarios are shared be most demanding in terms of handoff overhead. For
experience and augmented reality. Typical deployment example, users following a live action, such as a tennis rally,
scenarios where these handoff events may occur are: street- the average handoff interval can be about 1.3sec. Incidental
side, courtyard or tourist spot. However these handoff events motions, such as calm fidgeting, can even generate handoff
are rare if the APs are mounted at high enough positions. events at sub-second intervals (about 0.75sec). These high
Dense crowd: rates of handoff events in a mmWave air-interface can be
In this scenario, users are blocked from their AP handled by properly designing the handoff architecture and the
connectivity by movements of pedestrians, which can MAC layer.
potentially be semi-random. The deployment scenarios where
such handoff events may occur are: fairgrounds, college Table 1 Summary of estimated handoff intervals
campuses, tourist hotspots etc. In Figure 9 an example Handoff trigger type User action Handoff
courtyard scenario in a university campus is shown which interval
consists of walkways around and across a lawn. Pedestrians Fixed obstacle in a street (e.g. Pedestrian moving 11.6sec
are most likely to use the walkways, although occasionally trees every 15 m) along sidewalk
straying into the lawn. Since pedestrians only close to the user User in a vehicle 1.3sec
can block the LOS to the AP, users close to the walkways are moving at 25Km/h
likely to experience frequent handoffs. Based on the results in Moving obstacle in a street User in sidewalk 6 sec
Figure 8, it can be concluded that, handoff rate in such a (e.g. trucks at 30km/h and 1
scenario can exceed 1 handoff/sec in cases of high density of truck in every 100 m block)
pedestrians, exceeding 0.6 pedestrian/meter. More complex Rotation User following a live 1.3sec
models of pedestrian behavior in such deployments will be action (e.g. tennis rally)
studied in a future work. Incidental motion 0.75sec
Pedestrian Single User sitting or standing 3sec
obstacle: encounter on a sidewalk
Dense crowd User sitting or standing 1-1.5sec
in a college courtyard
lawn

REFERENCES
UE2 [1] M. Cudak, et. al., “Moving Towards Mmwave-Based Beyond-4G (B-
UE1 4G) Technology,” in Proc. IEEE VTC Spring 2013, June 2013.
[2] T. Rappaport, “Millimeter Wave Mobile Communications for 5G
Cellular: It Will Work!” IEEE Acces Journal, Vol 1, No. 1, May 2013
[3] Y. Atesal, B. Cetinoneri, M. Chang, R. Alhalabi, G. Rebeiz,
“Millimeter-Wave Wafer-Scale Silicon BiCMOS Power Amplifiers
UE3 Using Free-Space Power Combining”, IEEE Transactions on Microwave
Theory and Techniques, Vol. 59, No. 4, Apr 2011, pp. 954-965.
[4] M. Marcus and B. Pattan, “Millimeter Wave Propagation: Spectrum
pedestrian Management Implications,” IEEE Microwave Magazine, June 2005.
Figure 9 Pedestrian obstacles in a typical college campus [5] E. Ben-Dor, et. al., “Millimeter-wave 60 GHz Outdoor and Vehicle
AOA Propagation Measurements using a Broadband Channel Sounder,”
V. CONCLUSION in Proc. IEEE Globecom 2011, December 2011.
[6] S. Larew, et. al., “Air Interface Design and Ray Tracing Study for 5G
In a mmWave access network handoffs may occur due to Millimeter Wave Communications,” to be published in Proc. IEEE
obstacles between the user and the access points, user Globecom 2013, December 2013
movements and change in user terminal orientation. In this [7] M. Kyrö, et. al., “Long Range Wideband Channel Measurements at 81-
paper we have attempted to make a first order estimate of the 86 GHz Frequency Range,” in Proc. EuCAP 2010, April 2012.
handoff rates in a mmWave 5G access network for typical [8] Youtube video clip, “The Longest Grand Slam Rally Ever? Australian
deployment scenarios. The estimated average handoff Open 2013”, https://www.youtube.com/watch?v=rLDm254jtZA.
intervals are summarized in Table 1. Results from the analysis [9] Google map at address 1401 W Green St, Urbana, IL 61801, USA,
https://maps.google.com/maps?q=1401+W+Green+St,+Urbana,+IL+%E
show that, for pedestrian users in a tree-lined street, the 2%80%8E&ie=UTF8&hq=&hnear=1401+W+Green+St,+Urbana,+Illino
average handoff event intervals due to the fixed obstacles are is+61801&ll=40.107748,-
about 11.6sec. Typical handoff interval due to moving 88.226588&spn=0.004758,0.004823&t=h&z=18&vpsrc=6.

Вам также может понравиться