708 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO.

4, MAY 2012
Seamless Dual-Link Handover Scheme in
Broadband Wireless Communication Systems for
High-Speed Rail
Lin Tian, Member, IEEE, Juan Li, Yi Huang, Member, IEEE, Jinglin Shi, Member, IEEE, Jihua Zhou
AbstractDue to frequent handovers in broadband wireless
communications in high-speed rail, communication interruption
during handover could seriously degrade the experiences of
passengers on the train. Aiming to reduce the interruption time,
this paper proposes a seamless handover scheme based on a dual-
layer and dual-link system architecture, where a Train Relay
Station is employed to execute handover for all users in a train
and two antennas are mounted at the front and rear of a train.
In the proposed scheme, the front antenna executes handover
while the rear antenna is still communicating with BS, so that
the communication can keep non-interruptive throughout the
handover. Moreover, bi-casting is adopted to eliminate the data
forwarding delay between the serving BS and target BS. A
complete handover protocol is designed and the performance
of the proposed scheme is analyzed. It can be seen from
analytical results that the handover failure probability decreases
as cell overlap increases and the communication interruption
probability decreases with the decrease of train handover location
and the increase of cell overlap. The simulation results show
that in the proposed scheme, the communication interruption
probability is smaller than 1% when the handover location is
properly selected and the system throughput is not affected by
handover. In conclusion, both theoretical and simulation results
show that the proposed scheme can efciently perform seamless
handover for high-speed rail with low implementation overhead.
Index TermsHigh Speed Rail, Handover, Dual-link, Dual-
layer, Seamless.
1. INTRODUCTION
R
ECENTLY, high-speed railways has been developed
rapidly around the world [1] [2]. As a fast, safe and green
public transportation system, more and more people choose to
travel by high-speed trains. Due to the high demand for various
mobile wireless services such as online game, videophone and
so on, broadband wireless communication is necessary for
Manuscript received 15 May 2011; revised 28 October 2011. This work was
supported by Beijing Natural Science Foundation Major Project(No.4110001),
Co-building Program of Beijing Municipal Education Commission, Na-
tional Grand Project for Broadband Wireless Communication Networks,
China(2011ZX03001-001), Main Direction Program of Knowledge Innovation
of Chinese Academy of Sciences and Chongqing Science and Technology
Major Project (CSTC-2009AB2146).
L. Tian, J. Li, Y. Huang, J. Shi is with Wireless Communication
Technology Research Center, Institute of Computing Technology, Chinese
Academy of Sciences, and Beijing Key Laboratory of Mobile Computing
and Pervasive Device, P.O.Box 2704, 100190, Beijing, P.R. China (e-mail:
tianlindd@ict.ac.cn).
J. Zhou is with Chongqing Jinmei Communication Co., Ltd., Chongqing,
China.
Digital Object Identier 10.1109/JSAC.2012.120505.
train passengers especially in a long time traveling [3]. How-
ever, the existing broadband wireless communication systems
are mainly optimized for low mobility environments, whose
performance degrades seriously at high mobility [4]. Many
challenges occur when designing broadband wireless systems
for high-speed rail, among which handover is a great challenge
due to the following reasons:
1) High handover frequency. High-speed trains run at a
speed of more than 350km/h, which is expected to be 500km/h
in the future [4]. With the drastically increase of train speeds,
the handover will occur more and more frequently. For exam-
ple, given a cell size of 1-2 km which is typical in GSM-R
(GSM for Railways) networks [5], a high-speed train of 350
km/h experiences one handover every 10-20s.
2) Group handover. A group of mobile terminals in a train
will execute the handover simultaneously when the train enters
a new cell, while a single users handover is common in
existing broadband wireless systems such as Long Term Evo-
lution (LTE) [6]. In general, there are about 1000 passengers
in a train such as China Railways High-speed 3 (CRH3).
Assuming 10 percent passengers mobile terminals are active,
100 handover requests should be handled simultaneously,
which brings heavy implementation overhead to the system.
3) QoS guarantees. The Quality of Service (QoS) of various
mobile services of High Speed Trains (HST) will be degraded
because of frequent handover, especially QoS of real-time
services including passenger entertainment services and train
control services which are critical for the trafc safety [7].
For example, the packet delay of videophone should be
less than 50ms [8] while the handover time usually requires
100ms. If communication is interrupted during handover, QoS
requirements of videophone can not be satised.
In order to solve the above handover problems in broad-
band wireless communication systems for HST, the handover
scheme should minimize the communication interruption time
during handover to guarantee QoS requirements and reduce the
heavy implementation overhead brought by group handover.
Aimed at improving the performance of handover, consid-
erable researches have been carried out. A communication
system using Wi-Fi proposed in [9] can realize an average
handover time of 110ms when the train travels at 270km/h,
and an efcient MAC (Media Access Control) protocol that
reduces the handover delay is designed by [10] for vehi-
cle to roadside networks based on Wi-Fi. LTE-based HST
mobile communication system is proposed in [11], which
0733-8716/12/$25.00 c 2012 IEEE
TIAN et al.: SEAMLESS DUAL-LINK HANDOVER SCHEME IN BROADBAND WIRELESS COMMUNICATION SYSTEMS FOR HIGH-SPEED RAIL 709
adopts the at access network architecture to achieve the
lower control plane latency comparing to GSM-R networks,
so that the handover processing time can be reduced fur-
ther. Moreover, a two-layer cellular architecture is presented
to optimize the handoff blocking probability of high-speed
moving terminals [12] and dynamic antennas systems are
deployed to reduce the handover frequency [7]. Based on
special features of railways such as the predictable speed and
motion direction, a number of handoff decision algorithms
are proposed [13] [14] [15]. The unnecessary handovers
can be reduced by the fuzzy logic based handoff decision
algorithm [13] or random direction-lock based algorithm [14]
to lighten handover implementation overhead. Furthermore,
key parameters of handoff decision algorithms are optimized
to increase the decision accuracy [15]. A predictive handover
method is introduced in [16] with channel borrowing to reduce
the connection blocking probability and handover dropping
probability.
Although the handover problem in broadband wireless
systems for HST has been widely investigated, there is lim-
ited work on how to design a seamless handover scheme
where the communication interruption time is almost zero
and transmitted data is lossless [17] with low implementation
overhead. Therefore, we propose a seamless handover scheme
based on the dual-layer and dual-link system architecture for
HST. Two antennas are employed in this scheme, one at
the front of the train and the other at the rear. The rear
antenna keeps connected to the serving Base Station (BS)
when the front antenna carries out handover to the target BS,
and it switches to the working frequency of the target BS
after the handover of the front antenna is nished. Therefore,
the communication will not be interrupted during handover.
Secondly, the proposed scheme adopts the idea of bi-casting
to eliminate the delay and overhead of data forwarding from
the serving BS to the target BS. Moreover, it is Train Relay
Station (TRS) not each user in a train to implement the
handover procedure, which reduces the signalling overhead
to about half of LTE handover schemes. The performance
of the proposed scheme is analyzed in terms of handover
probability, handover failure probability and communication
interruption probability. Theoretical and simulation results
show that comparing with LTE handover scheme, the proposed
scheme can reduce communication interruption probability
and handover failure probability drastically with only half
of LTE schemes signalling overhead. Moreover, simulation
results show that the throughput of the proposed scheme is
not affected by handover.
The rest of the paper is organized as follows. In Section 2,
the dual-layer and dual-link system architecture is introduced.
Section 3 is devoted to new handover scheme design. Section
4 develops an analytical model to derive the performance of
handover schemes for HST wireless communication systems.
Simulation results are given and discussed in Section 5.
Section 6 concludes the paper.
2. SYSTEM ARCHITECTURE
The dual-layer system architecture has been well recognized
in the broadband wireless communications for HST [4] [18].
As shown in Fig.1, the users communicate directly with the
Fig. 1. System Architecture
Access Points (AP) located inside each carriage. All APs
which could support different types of wireless access tech-
nologies (e.g. WLAN, 2G and 3G) are controlled by a TRS.
The data collected by APs will be forwarded to a ground BS
via the TRS. In the dual-layer architecture, the TRS can act as
a single user to execute handover from the serving BS to target
BS, which greatly reduces the handover overhead comparing
to the conventional scheme where all the active mobile users
in a train should execute handover simultaneously.
On the other hand, dual-link means that two antennas with
different transceivers are mounted on the top of a train [19],
which are denoted by front antenna and rear antenna in Fig.1.
When the train camps in a single cell, the TRS will select
one antenna with better signal qualities to communicate with
BS. When the train is approaching the serving BS, the signal
quality of the front antenna is usually better than that of the
rear antenna. Hence, the front antenna will be selected as the
communication link. On the contrary, if the train moves away
from the BS, the rear antenna will be selected to communicate
for its higher signal qualities. When the train moves across
cell edges, the front antenna executes handover while the rear
antenna keeps connected to the serving BS. If the handover
is successful, the rear antenna simply changes its working
frequency to that of the target BS sometime after it enters the
overlap region. However, if the handover of the front antenna
fails, the rear antenna will carry out handover for the second
time.
International Union of Railways has decided that the next
generation of railway wireless communication standard will
be designed based on 3GPP LTE. Therefore, several func-
tional entities in LTE system are employed to implement the
proposed handover scheme. As shown in Fig.2, the network
entities include the Mobility Management Entity (MME) and
Serving Gateway (S-GW). MME holds the functions of non-
access stratum signalling tracking, TRS mobility management,
air-interface security control, authentication, S-GW selection,
etc. The major functions of S-GW include local mobility sup-
port for TRS handover, downlink packets buffering, packets
routing and forwarding, etc. In the proposed scheme, S-GW
is mainly responsible for the data transmission from the packet
core networks to the proper BS. Therefore, S-GW would
710 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 4, MAY 2012
Fig. 2. Network Entity related with Handover
switch the downlink packets routing path and forward the
packets to the target BS after a handover procedure is success-
fully executed. In the proposed scheme, there are two major
modications to the above entities: 1) MME needs to maintain
two different sets of the connection context information about
a TRS at the same time since the TRS may connect with
both the serving BS and target BS during handover. 2) S-GW
should keep two forwarding path within its routing table to
support bi-casting, that is, the service packets are forwarded to
both of the serving BS and target BS. Although the proposed
scheme introduces some special functions in the design of BS,
MME and S-GW entities of LTE, it will not affect mobile
terminals of LTE because the TRS entity represents mobile
terminals to implement handover.
3. HANDOVER SCHEME
3.1 Handover Procedure
How to ensure the transmitted data lossless during handover
is an important problem. There are two types of solutions to
solve this problem in existing broadband wireless systems. As
shown in Fig.2, one is bi-casting [20] and the other is data
forwarding [21]. In bi-casting, S-GW sends two copies of data
to the serving BS and target BS respectively during handover.
The serving BS processes and sends data towards mobile
terminals while the target BS discards all the arriving data
until it takes the serving role. In data forwarding, the serving
BS will duplicate data: one copy is sent to mobile terminals,
and the other is forwarded to the target BS which discards
all the data as same as in bi-casting. Because the target BS
transmits new arrival data from S-GW immediately without
re-ordering the data after handover in bi-casting solution, the
communication interruption time is equal to the handover
process time approximately. However, in data forwarding so-
lution, the target BS should nish sending the data forwarded
from the serving BS rstly and then transmit new arrival data
to avoid the data out of sequence, so the communication
interruption time includes both the handover process time
and data forwarding delay. Therefore, the communication
interruption time of the bi-casting solution is shorter than
that of the data forwarding solution. Because some real-time
services require that the communication interruption time is as
short as possible, the bi-casting solution is more suitable for
Fig. 3. Proposed Handover Procedure
these services. Moreover, the interface between BS and S-GW
for data transmission has higher throughput than the interface
between BSs.
Based on the above analysis, the proposed scheme utilizes
bi-casting to avoid the data forwarding delay and overhead.
The scheme is composed of three stages: handover prepa-
ration, handover execution and handover completion. The
detailed handover process is illustrated in Fig.3.
Stage 1: Handover Preparation
1) The TRS issues a Measurement Report to the serving
BS periodically, which includes necessary information for
handover decision such as Received Signal Strength (RSS) of
the front antenna. The serving BS decides whether to handoff
the TRS based on some handoff decision algorithm.
2) Once the serving BS decides to trigger the handover of
TRS, it sends a handover request message to the target BS to
inform the upcoming handover.
3) The target BS makes an admission control and congure
required resources for TRS to guarantee a successful handover.
The required resources are decided by all the services of the
user terminals and the train. Then a handover request ACK
(Acknowledgement) message is sent by the target BS towards
the serving BS to inform the handover command.
Stage 2: Handover Execution
4) The target BS requests bi-casting downlink packets from
the S-GW. One of the duplicated packets from S-GW is sent to
the rear antenna through serving BS, and the other is discarded
by the target BS until it takes the serving role.
5) When the TRS receives a RRC (Radio Resource Control)
connection recongure message from the serving BS, it makes
the front antenna detach from the serving BS, perform syn-
chronization to the target BS, and establish connection with
the target BS.
6) When the rear antenna enters the overlap region and the
front antenna is selected as the the communicating link, the
TIAN et al.: SEAMLESS DUAL-LINK HANDOVER SCHEME IN BROADBAND WIRELESS COMMUNICATION SYSTEMS FOR HIGH-SPEED RAIL 711
rear antenna changes its working frequency to that of the target
BS.
If the handover of the front antenna fails, step 6) is replaced
by step 6A) as follow:
6A) TRS will report the signal qualities of the rear antenna
to the serving BS according to step 1). When the serving BS
triggers the handover, the rear antenna will handoff to the
target BS as the front antenna dened in step 5).
Until now, the handover of the train is nished. A Handover
Complete message is sent to the serving BS.
Stage 3: Handover Completion
7) The serving BS requests to nish bi-casting and the S-
GW makes corresponding conguration. Downlink packets are
only transmitted to the target BS. The serving BS releases the
resources associated to the TRS.
In the above process, the front antenna executes handover
in step 5) while the data is still transmitted/received (Tx/Rx)
between the rear antenna and the serving BS, and then the rear
antenna attaches to the target BS as dened in step 6) after
the front antenna has taken charge of data Tx/Rx. Therefore,
the proposed scheme can ensure that there is no interruption
in the data communication for all the users in a train during
handover.
3.2 Handover Decision
The high-speed handoff decision algorithms [14] [15] whose
main idea is widely used can be employed to trigger the
handover of TRS in the proposed scheme. If the relative
received signal strength by a TRS from different BS exceeds
the pre-dened hysteresis level optimized in [15], the handover
will be triggered. Authors in [14] suggest to take the direction
of train moving into consideration to reduce the unnecessary
handovers. In order to support the above algorithms, each BS
in the proposed scheme should maintain the records dened
as (Previous BS ID, Next BS ID) to show the relationship
between the moving direction and target BS. For example, the
record (1, 2) represents that the direction of train moving is
BS 1the serving BSBS 2. If a TRS executes handover
to the current serving BS from BS 1, the serving BS can decide
that the target BS would be BS 2 because of the linear cell
deployment for high-speed railways.
Moreover, the time when the rear antenna changes its
working frequency should satisfy two conditions. One is that
the rear antenna has entered the overlap region, and the other
is that the front antenna is selected to communicate with
BS to guarantee at least one antenna keep connected during
handover.
3.3 Signalling Design and Implementation Overhead
In the handover signalling design, there are three main
differences between the proposed scheme and the standard
LTE scheme [21]. Firstly, new signalling messages including
bi-casting request and bi-casting request acknowledge are de-
signed in the proposed scheme because bi-casting is adopted.
For example, the feasible design of bi-casting request message
will contain the following elds: TRS ID (identication) and
target BS ID used for building the bi-casting path, all the
active users id and connection information used to create data
TABLE I
HANDOVER SIGNALLING BANDWIDTH
Signalling Message LTE
Scheme(Byte)
Proposed
Scheme(Byte)
Measurement Report 4400 44
Handover Request 24100 15982
Handover Request Ac-
knowledge
7000 2713
RRC Connection Recon-
guration
3600 36
RRC Connection Recon-
guration Complete
900 9
SN Status Transfer 7300 -
Path Switch Request 13600 5977
Path Switch Request Ac-
knowledge
11700 8730
UE Context Release 3000 30
Bi-casting Request - 2977
Bi-casting Request Ac-
knowledge
- 3000
Total 75600(604Kb) 39498(316Kb)
bearers of every user. At the same time, SN status transfer
message used for data forwarding in LTE scheme is canceled.
Secondly, because TRS represents all the active users in a
train to execute handover, some signalling messages, such as
measurement report, only need to include the information of
TRS. Finally, some signalling messages should be modied to
contain necessary information of all the active users in a train.
For example, a target BS executes admission control for every
handover user based on the users context information included
in a handover request message. In the proposed scheme, only
one handover request message is sent to a target BS for all the
active users. Therefore, the handover request message should
contain all the active users context information. Based on the
above analysis of the signalling design in the proposed scheme
and LTE scheme, the occupied bandwidth of every handover
signalling message in the proposed scheme and LTE scheme
with 100 active users can be derived as shown in Table I. Due
to the space limit, the complete handover signalling of the
proposed scheme, which can be designed based on the above
analysis and the references [20] [22] [23] [24], will not be
illustrated one by one in this paper.
We use the occupied bandwidth of handover signalling, the
average data forwarding delay between the serving BS and
target BS, and the number of simultaneous handover requests
to measure the implementation overhead of handover schemes.
Table II shows the overhead of the proposed scheme, the
standard LTE handover scheme and the dual-antenna scheme
dened in [19] with 100 active terminals in a train. First of all,
the TRS executes handover as a single user in the proposed
scheme, so the handover signalling information including
handover request/ACK message, RRC connection recongure
712 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 4, MAY 2012
TABLE II
HANDOVER IMPLEMENTATION OVERHEAD
Proposed LTE Dual-
antenna
[19]
Signalling Bandwidth 316Kb 604Kb -
Forwarding Delay 0 18ms 18ms
Simultaneous
Handover Request
1 100 1
message etc. dened above only need to be transmitted one
time for all the active terminals in a train. Therefore, the
proposed scheme signicantly reduces the signalling overhead
of LTE handover scheme. It can be seen clearly from Table II
that the signalling overhead of the proposed scheme is only
half of LTE schemes [22] and LTE handover scheme needs to
handle 100 handover requests simultaneously while the pro-
posed scheme only handle one request from the TRS. Because
authors of [19] have not dened the handover signalling, the
signalling bandwidth can not be calculated. However, there is
also a central control entity dened in [19] which is similar as
TRS, so the number of simultaneous handover requests of the
dual-antenna scheme [19] is 1. Secondly, as described in step
4), it is the S-GW but not the serving BS to send a copy of
data to the target BS in the proposed scheme, which can avoid
the data forwarding delay between the serving BS and target
BS. In the LTE scheme, data is transferred to the target BS
by serving BS through X2 interface [23]. The latency of X2
interface is about 18ms [25], so the data forwarding delay is
18ms in the LTE scheme if the datas processing and queuing
delay is negligible. The data forwarding method in [19] is
similar as that of the LTE scheme except that the interface
between BSs is X2 interface in the LTE scheme which is
optimized to reduce data transfer delay. Thus, the forwarding
delay of the dual-antenna scheme [19] is no less than 18ms.
4. PERFORMANCE ANALYSIS
In order to evaluate the performance of the proposed scheme
in minimizing the communication interruption time during
handover, handover probability, handover failure probability
and communication interruption probability will be analyzed.
Because the above performance measures are dependent on
the received signal strength, we will analyze the received
signal strength rstly. As shown in Fig.4, d
s
is the distance
between BS and railway track, L
t
is the length of train,
v is the velocity of the train, t is the handover time, D
is the distance between two adjacent BSs such as BS i
and BS j, D
overlap
represents the length of overlap be-
tween the adjacent cells whose radius can be dened as
Rd. The train location x is measured by the distance be-
tween the front antenna and origin in X-axis. Therefore,
the distance from the antenna to BS can be expressed as
follows: D
f,i
=
_
x
2
+ d
2
s
, D
f,j
=
_
(D x)
2
+ d
2
s
, D
r,i
=
_
(x L
t
)
2
+ d
2
s
, D
r,j
=
_
(D + L
t
x)
2
+ d
2
s
, where
D = 2RdD
overlap
. OFDM(Orthogonal Frequency Division
Multiplexing)-based air interface is employed [26] [27]. Sup-
Fig. 4. Analysis illustration
pose there are totally M sub-carriers. In one symbol duration,
the baseband signal received by the front antenna from BS i
can be given by
r
f,i
(t) = pl
f,i
sh
f,i
h
f,i
(t)
M1

m=0
d
m
e
j2mft
+
f,i
(t) (1)
where pl
2
f,i
= AD

f,i
stands for the path loss, A is a constant,
D
f,i
is the distance from the front antenna to BS i, is
the path loss exponent, sh
2
f,i
is the log-normally distributed
shadowing fading, h
f,i
(t) =

L1
l=0
h
f,i,l
(t
f,i,l
) is the
microscopic multipath Rayleigh fading, L is the number of
resolvable multipath components, d
m
is the transmitted data,
P
d
= E[|d
m
|
2
], f is the sub-carrier spacing, t [0, T
g
+T
s
]
T
g
and T
s
are the time duration of Cyclic Prex (CP) and
effective OFDM symbol respectively, and
f,i
(t) represents
the zero mean additive white Gaussian noise.
When Doppler frequency exists, the autocorrelation func-
tion of the l
th
path is dened as [28], [29]

l
(t) = E[h
f,i,l
(t +t) h

f,i,l
(t)] =
2
l
J
0
(2f
D
t) (2)
where
2
l
= E[|h
f,i,l
(t)|
2
], f
D
= vf
c
/c represents the
maximum Doppler frequency shift, f
c
is the carrier frequency,
c is the velocity of light and J
0
() is the correlation coefcient
(zeroth-order Bessel function of the rst kind).
Assuming that the length of CP is longer than the maximum
delay of the equivalent multipath channel, the output signal on
the m
th
sub-carrier of the front antenna in the k
th
time chip
duration can be obtained as
Y
m
f,i
(k) =
1
T
s
_
kT+T
s
kT
r
f,i
(t) e
j2f
m
(tkT)
p(t kT)dt
= pl
f,i
sh
f,i

m
m
(k)d
m
+
M1

m=0,m= m
pl
f,i
sh
f,i

m
m
(k)d
m
. .
ICI
+ n
f,i
(k)
(3)
where n
f,i
is the noise component with a variance of
2
noi
,
T = T
g
+T
s
, H
f,i
(k) = pl
f,i
sh
f,i

m
m
(k) is the channel fading
TIAN et al.: SEAMLESS DUAL-LINK HANDOVER SCHEME IN BROADBAND WIRELESS COMMUNICATION SYSTEMS FOR HIGH-SPEED RAIL 713
factor on the useful signal with

m
m
(k) =
1
T
s
L1

l=0
e
j2
ml
M

_
T
s
0
h
f,i,l
(t + kT)dt (4)
and a
f,i,m
(k) = pl
f,i
sh
f,i

m
m
(k) which represents the inter-
carrier interference (ICI) contributed from the m
th
sub-carrier
has a factor of

m
m
(k) =
1
T
s
L1

l=0
e
j2
ml
M

_
T
s
0
h
f,i,l
(t + kT)e
j2
m m
T
s
t
dt
(5)
Thus,
E[|
m
m
(k)|
2
] = E[|
1
T
s
L1

l=0
e
j2
ml
M

_
T
s
0
h
f,i,l
(t + kT)dt|
2
]
=
1
T
2
s

_
T
s
0
_
T
s
0
L1

l=0

2
l
J
0
(2f
D
(t
1
t
2
))dt
1
dt
2
=
L1

l=0

2
l

_
1
1
J
0
(2f
D
T
s
x)(1 |x|)dx
(6)
Moreover, a tight upper bound can be obtained for the power
of ICI [30] as follows
I = P
d
(pl
f,i
sh
f,i
)
2
(2f
D
T
s
)
2
/12 (7)
In LTE, the received signal strength is measured by mobile
terminals and determined as the average power of the reference
signals. Each reference signal occupies one resource element
(one sub-carrier one OFDM symbol) [31] [32]. Assuming
that small-scale fading, ICI and noise component could be
averaged out and have no effect on the nal average signal
strength, the signal strength received from BS i by the front
antenna, dened as R
f,i
in dB, can be calculated as
R
f,i
= 10log
10
(P
d
|pl
f,i
sh
f,i
|
2
E[|
m
m
(k)|
2
])
= 10log
10
(P
d
A ) 10log
10
D
f,i
+ 10log
10
sh
2
f,i
(8)
where =

L1
l=0

2
l

_
1
1
J
0
(2f
D
T
s
x)(1|x|)dx. Similarly,
R
f,j
, R
r,i
and R
r,j
can be derived based on (8).
4.1 Handover Probability
The proposed scheme performs a handover to the adjacent
BS if the received signal strength of the front antenna from
the adjacent BS exceeds that of the serving BS by a hysteresis
level U (dB). The handover probability is usually dened as
the probability that there is a handoff at a certain time [33].
However, whether a handover takes place in a proper location
is more important in the performance of handover schemes
for HST [19]. Therefore, the handover probability is dened
in this paper as the probability that handoff occurs at position
x, which can be expressed as
P
ho
= P{R
f,j
R
f,i
U}
= P{10log
10
D
f,j
+ 10log
10
sh
2
f,j
+ 10log
10
D
f,i
10log
10
sh
2
f,i
U}
(9)
Let SD
f,j
= 10log
10
sh
2
f,j
and SD
f,i
= 10log
10
sh
2
f,i
.
Then SD
f,j
and SD
f,i
are both zero mean Gaussian dis-
tributed with the standard deviation
f,j
and
f,i
respectively.
Since the sum of Gaussian distributed variables still follows
Gaussian distribution [34] [35], SD
f
= SD
f,j
SD
f,i
is zero
mean Gaussian distributed with the standard deviation given
by
f
=
_

2
f,j
+
2
f,i
. Therefore, the handover probability
can be derived as follows:
P
ho
= P{10log
10
D
f,i
/D
f,j
+ SD
f
U}
= Q((U 10log
10
_
x
2
+ d
2
s
_
(D x)
2
+ d
2
s
)/
f
)
(10)
It can be seen that the handover probability is determined by
the trains current location x. In normal cases, the handover
will be completed before x D. Therefore, the handover
probability increases when x becomes larger, that is, the train
moves towards cell edge.
4.2 Handover Failure Probability
After an antenna is triggered to handover, if the average
received signal strength of the antenna from the target BS
is lower than a threshold T which is the minimum received
signal strength to maintain the communication, the antenna
can not connect to the target BS and the handover of the
antenna is failed. We dene handover failure probability,
denoted by P
ho fail
, to measure it. In the proposed scheme, if
the handover of the front antenna fails, the rear antenna will
execute handover as the front antenna. Therefore, handover
failure probability of the proposed scheme can be expressed
as
P
ho fail
= P
f
ho fail

1
Rd + L
t
x
r
_
Rd+L
t
x
r
P
r
ho fail
dx
(11)
where P
f
ho fail
and P
r
ho fail
represent the handover failure
probabilities of the front antenna and rear antenna respectively,
x
r
is the location where the rear antenna has entered the
overlap region and the front antenna has completed handover.
If there is only one antenna, P
ho fail
= P
f
ho fail
. Based on
the above analysis, P
f
ho fail
can be obtained by
P
f
ho fail
= P{R
f,j
T|R
f,j
R
f,i
U}
=
P{R
f,j
T, R
f,j
R
f,i
U}
P{R
f,j
R
f,i
U}
=
1
P
ho
_

f

P{SD
f,i
10log
10
D
f,i
/D
f,j
+ U
|SD
f,j
= }P{SD
f,j
= }d
=
1
P
ho
_

f

(1 Q(
10log
10
D
f,i
/D
f,j
+ U

f,i
))

2
f,j
e

2
2
2
f,j
d
(12)
where
f
= (T 10log
10
(P
d
A ) +
10log
10
_
(Dx
0
)
2
+ d
2
s
) and x
0
is the location where
the handover of the front antenna is triggered. If x
0
is xed,
P
f
ho fail
will decrease when D becomes smaller. Because
D = 2RdD
overlap
, it can be derived that P
f
ho fail
decreases
with the increase of D
overlap
.
714 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 4, MAY 2012
Similarly, P
r
ho fail
can be given by
P
r
ho fail
= P{R
r,j
T|R
r,j
R
r,i
U}
=
1
P
ho
_

r

(1 Q(
10log
10
D
r,i
/D
r,j
+ U

r,i
))

2
r,j
e

2
2
2
r,j
d
(13)
where
r
= (T 10log
10
(P
d
A ) +
10log
10
_
(D + L
t
x)
2
+ d
2
s
) and x represents the
location where the rear antenna starts attaching to the target
BS and x x
0
+ vt. Same as P
f
ho fail
, P
r
ho fail
also
becomes smaller when D
overlap
is increasing if x is xed.
Therefore, the handover failure probability will decrease
with the increase of D
overlap
when the handover location is
decided.
4.3 Communication Interruption Probability
In order to study the performance of the proposed handover
scheme, a metric Communication Interruption Probability
(Com Interrupt Probability), denoted as P
ho out
, is dened
as follows:
Communication Interruption Probability: The probability of
communication interruption during the handover procedure. If
the communication is interrupted at all times during the han-
dover procedure, Com Interrupt Probability is 1. The smaller
Com Interrupt Probability is, the higher seamless performance
the handover scheme can achieve.
Based on the denition, P
ho out
of the proposed scheme
can be given by
P
ho out
= (1 P
f
ho fail
) P
f
ho out
+ P
f
ho fail
P
r
ho out
(14)
where P
f
ho out
represents the outage probability of the link
between the rear antenna and the serving BS when the
front antenna is accessing to the target BS, and P
r
ho out
is
1 because the communication is interrupted when the rear
antenna executes handover if handover of the front antenna
fails. Based on the analysis, P
f
ho out
is given by (15) (see
next page) where z
r,i
= (T 10log
10
(P
d
A ) +
10log
10
_
(x L
t
)
2
+ d
2
s
)/
r,i
. Therefore,
P
ho out
= (1P
f
ho fail
)(1
1
vt
_
x
0
+vt
x
0
Q(z
r,i
)dx)+P
f
ho fail
(16)
Because x
0
vt in HST scenarios, the approximation of
P
ho out
can be expressed as
P
ho out
(1 P
f
ho fail
) (1 Q(z
0
r,i
)) + P
f
ho fail
= 1 Q(z
0
r,i
) + P
f
ho fail
Q(z
0
r,i
)
(17)
where z
0
r,i
= (T 10log
10
(P
d
A ) +
10log
10
_
(x
0
L
t
)
2
+ d
2
s
)/
r,i
. It can be seen that
Com Interrupt Probability is decided by the handover
location x
0
and the overlap D
overlap
between adjacent cells.
Based on the analysis of P
f
ho fail
, it can be derived that
Com Interrupt Probability decreases with the increase of
D
overlap
and the decrease of x
0
.
TABLE III
SIMULATION PARAMETERS
Parameters Value
Number of Subcarriers M 1024
Guard Interval M 144/2048
Bandwidth of Subcarrier 15kHz
Carrier Frequency 2GHz
Symbol Duration 1/14ms
Total Transmit Power P
tx
86dBm
Path loss model 31.5 + 35log
10
(d)
Shadow fading deviation 4dB
Power density of background noise -145dBm/Hz
Signal threshold T -58dBm
Target BER 10
6
Cell radius 1500m
Overlap 400m
Length of Train 200m
Distance between BS and Railways d
s
100m
BS Antenna Height 32m
Train Antenna Height 1.5m
Handover Time 100ms
5. SIMULATION RESULTS
By means of simulation, the performance of the proposed
handover scheme based on the dual-layer and dual-link system
are investigated in the high-speed scenario. The speed range of
the train is assumed from 180km/h to 540km/h. Suppose there
are totally 1000 users in a train and ten percent of them are
active. According to the HST channel model recommended by
3GPP [36], a single-path fading channel is considered in the
simulation. The overlap area of the adjacent cells is measured
by length of the area as indicated in Fig.4. Other parameters
used in the simulation are shown in Table III.
Firstly, the handover probability with the different hysteresis
level U are plotted in Fig.5 as a function of the train location.
It can be seen that the handover probability with U = 1dB
will exceed 50% when the train location is larger than 1300m.
Moreover, it also can be seen that the theoretical results are
close to the simulation results. If the measurement period
of the received signal strength is 400ms in the proposed
scheme, the handover decision will occur six times at most
with the train location from 1300m to 1500m. The handover
occurrence probability is dened to describe the probability
that handover is triggered before a certain position, which is
calculated based on the handover probability. Therefore, the
handover occurrence probability between 1300m and 1500m
is larger than 1 (0.5)
3
(0.6)
3
= 99.2% with U = 1dB and
Speed = 360km/h. Fig.6 shows the handover occurrence
probability with the different train speeds. In the same train
location, the handover occurrence probability decreases with
the increase of train speed. Therefore, the measurement period
TIAN et al.: SEAMLESS DUAL-LINK HANDOVER SCHEME IN BROADBAND WIRELESS COMMUNICATION SYSTEMS FOR HIGH-SPEED RAIL 715
P
f
ho out
=
1
vt
_
x
0
+vt
x
0
P{R
r,i
T}dx =
1
vt
_
x
0
+vt
x
0
(1 Q(
T 10log
10
(P
d
A ) + 10log
10
D
r,i

r,i
))dx
= 1
1
vt
_
x
0
+vt
x
0
Q(z
r,i
)dx (15)
Fig. 5. Handover Probability of proposed scheme (Speed=360km/h)
Fig. 6. Handover Occurrence Probability of proposed scheme
should be shortened to reach the same probability when the
speed is increasing.
Fig.7 shows handover failure probability as a function
of handover locations where the front antenna begins to
handover. Both simulation and analytical results are shown.
It can be seen that the analytical results match well the
simulation results. The handover failure probability of the
proposed scheme is smaller than 0.001 while that of the single
antenna scheme is about 0.05. Moreover, the handover failure
probability decreases with the handover location increasing
because the received signal by the front antenna from the target
BS becomes better when the handover location is nearer to the
target BS. Next, Fig.8 shows the handover failure probability
Fig. 7. Handover Failure Probability (Speed = 360km/h)
Fig. 8. Handover Failure Probability as a function of overlap (Speed =
360km/h)
as a function of the overlap. According with the theoretical
analysis, the handover failure probability decreases as the
overlap increases. This is because that, at the same location,
the received signal from the target BS becomes stronger with
the overlap increasing.
In Fig.9, communication interruption probability of the pro-
posed scheme is plotted with the different handover locations.
Again, the analytical results are close to the simulated ones.
Although the handover location is on the edge of the serving
BS, Com Interrupt Probability is only about 0.2. Comparing
with the conventional scheme such as LTE handover scheme
of which Com Interrupt Probability is 1 all the time during
the handover procedure, the proposed scheme decreases the
716 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 4, MAY 2012
Fig. 9. Communication Interruption Probability of proposed scheme (Speed
= 360km/h)
Fig. 10. Communication Interruption Probability with different overlap
(Speed = 360km/h)
probability of communication interruption drastically. It also
can be seen from Fig.9 that Com Interrupt Probability is no
more than 0.1 when the handover location is within 1300m.
Therefore, the proposed scheme can achieve a seamless
handover when the handover location is properly selected.
Moreover, Com Interrupt Probability with different overlap
are shown in Fig.10. At the same train location, Com Interrupt
Probability decreases with the overlap increasing. Therefore,
we can cong the overlap for network deployment by taking
the handover performance such as handover failure probability
and Com Interrupt Probability into consideration.
In Fig.11, the system capacity of the proposed scheme and
single antenna scheme as a function of the train location
are plotted. As shown in Fig.11, the throughput of single
antenna scheme is 0 while that of the proposed scheme is
not changed much during handover (about 1300m-1310m)
because single antenna scheme can not support the seamless
handover. Moreover, the average throughput of the proposed
scheme is higher than that of single antenna scheme when the
Fig. 11. Throughput of proposed scheme and single antenna scheme(Speed
= 360km/h)
train camps on a cell because the TRS can select the antenna
with better signal qualities to communicate with BS. For
example, the throughput of the proposed scheme is decreasing
when the location is from 0 to 100m for the front antenna
moving farther away from the serving BS, and the throughput
starts increasing again when the location reaches about 100m
because the rear antenna is selected to communicate with BS
from 100m and this antenna gradually approaches the serving
BS.
6. CONCLUSION
This paper has investigated the handover problem in broad-
band wireless communication systems for HST. Based on the
characteristics of dual-layer and dual-link system architecture,
a seamless handover scheme is proposed to minimize the
communication interruption during handover. The proposed
scheme includes three main features: 1) two antenna handover
process to keep the communication not interrupted during
handover; 2) bi-casting is utilized to avoid the data forward-
ing delay between serving BS and target BS and 3) TRS
representing all the users in a train to execute handover to
reduce the implementation overhead. The performance of the
proposed scheme is analyzed in terms of handover probability,
handover failure probability and communication interruption
probability. Simulation and theoretical results show that the
proposed scheme can not only achieve seamless handover but
also reduce the handover implementation overhead.
Focusing on the design and performance analysis of the
proposed handover process, a relatively ideal model with
perfect channel estimation is assumed in this paper. However,
it is well known that accurate channel information is difcult
to obtain in broadband wireless systems for HST. Therefore,
in future work, it is necessary to investigate how the channel
estimation errors affect the performance of handover schemes.
Moreover, if error handover decision is made because of the
inaccurate channel estimation results, some recovery methods
should be studied to avoid the communication interruption
during handover.
TIAN et al.: SEAMLESS DUAL-LINK HANDOVER SCHEME IN BROADBAND WIRELESS COMMUNICATION SYSTEMS FOR HIGH-SPEED RAIL 717
REFERENCES
[1] M. Givoni: Development and Impact of the Modern High-speed Train:
A Review, Transport Reviews, vol. 26, no. 5, pp. 593C611, Sep. 2006.
[2] H. Dong, B. Ning, B. Cai, Z. Hou: Automatic Train Control System
Development and Simulation for High-Speed Railways, IEEE Circuits
and Systems Mag., vol. 10, no. 2, pp. 6C18, Secondquarter, 2010.
[3] BBC News: Wi-Fi May Tempt Travellers, London,
United Kingdom, May 20, 2004. [Online]. Available:
http://news.bbc.co.uk/2/hi/technology/3729583.stm.
[4] J. Zhang, Z. Tan, Z. Zhong, Y. Kong: A Multi-mode Multi-band and
Multi-system-based Access Architecture for High-speed Railways, IEEE
72nd Vehicular Technology Conference Fall (VTC 2010-Fall), Ottawa,
Canada, Sep.2010.
[5] ETSI TR 102.281: Detailed requirements for GSM operation on Rail-
ways, v2.0.0, May 2006.
[6] 3GPP TR 25.913: Requirements for Evolved UTRA (E-UTRA) and
Evolved UTRAN (E-UTRAN), v9.0.0, Dec.2009.
[7] S. Lin, Z. Zhong, B. Ai, C. Briso: Transmission Interference Improve-
ment of Railway Communication via Distributed Antennas System,
IEEE 71nd Vehicular Technology Conference Fall (VTC 2010-Spring),
Taipei, Taiwan, May 2010.
[8] ITU-R M.[IMT.EVAL]: Guidelines for evaluation of radio interface
technologies for IMT-Advanced, Oct.2008.
[9] K. Yamada, Y. Sakai, T. Suzuki, Y. Kawahara, T. Asami, H. Aida:
A Communication System with a Fast Handover under a High Speed
Mobile Environment, IEEE 72nd Vehicular Technology Conference Fall
(VTC 2010-Fall), Ottawa, Canada, Sep.2010.
[10] B. Sikdar: Characterization and Abatement of the Reassociation Over-
head in Vehicle to Roadside Networks, IEEE Trans. Commun., vol.58,
pp. 3296-3304, Nov. 2010.
[11] T. Gao, B. Sun: A high-speed railway mobile communication system
based on LTE, International Conference on Electronics and Information
Engineering (ICEIE 2010), vol.1, pp. 414-417, Kyoto, Japan, Aug. 2010.
[12] K. Ioannou, S. Louvros, I. Panoutsopoulos, et al.: Optimizing the
Handover Call Blocking Probability in Cellular Networks With High
Speed Moving Terminal, IEEE Commun. Lett., vol.6, no.10, Oct.2002.
[13] Y. Li, Y. Yan: Fuzzy Logic Based Handoff Decision Algorithm in GSM-
R Network, IET Conference on Wireless, Mobile and Sensor Networks
(CCWMSN), Shanghai, China, Dec. 2007.
[14] Y. Li, Y. Yan: Performance Analysis of Handoff Algorithm Using
Relative RSS Measurements and Random Direction-Lock in GSM-R,
IET International Communication Conference on Wireless Mobile and
Computing (CCWMC), Shanghai, China, Dec. 2009.
[15] Y. Li, Y. Yan: Optimization for Parameters of Handoff Algorithm
with Adaptive Hysteresis in GSM-R, IET International Communication
Conference on Wireless Mobile and Computing (CCWMC), Shanghai,
China, Dec. 2009.
[16] L. Su, F. Liu, et al.: Predictive Group Handover Scheme with Channel
Borrowing for Mobile Relay Systems. International Wireless Commu-
nications and Mobile Computing Conference (IWCMC), Crete Island,
Greece, Aug. 2008.
[17] R. Kim, X. Yang, C. Chou: Advanced Handover Schemes in IMT-
Advanced Systems, IEEE Commun. Mag., vol.48, pp. 78-85, Aug.2010.
[18] Y. Zhou, Z. Pan, J. Hu, J. Shi, X. Mo: Broadband Wireless Commu-
nications on High Speed Trains, Wireless and Optical Communications
Conference (WOCC), Newark, NJ, U.S.A., invited paper, Apr. 2011.
[19] C. Yang, L. Lu, C. Di, X. Fang: An On-Vehicle Dual-Antenna Handover
Scheme for High-Speed Railway Distributed Antenna System, Interna-
tional Conference on Wireless Communications Networking and Mobile
Computing (WiCOM), Chengdu, China, Sep.2010.
[20] 3GPP R3-060741: Inter-RAT Mobility for Real-time ows, Shanghai,
China, May 2006.
[21] 3GPP TS 36.300: Overall description for Evolved UTRA (E-UTRA)
and Evolved UTRAN (E-UTRAN), v9.0.0, Jun.2009.
[22] 3GPP TS 36.331: Evolved Universal Terrestrial Radio Access (E-
UTRA) Radio Resource Control (RRC), v9.0.0, Sep.2009.
[23] 3GPP TS 36.423: X2 Application Protocol (X2AP), v10.0.0, Dec. 2010.
[24] 3GPP TS 36.413: S1 Application Protocol (S1AP), v10.0.0, Dec. 2010.
[25] 3GPP R3-070593: X2 interface delay, St. Julians, Malta, Mar.2007.
[26] Y. Zhou, T. Ng: Performance analysis on MIMO-OFCDM systems with
Multi-code Transmission, IEEE Trans. Wireless Commun., vol. 8, issue
9, pp. 4426-4433, Sept. 2009.
[27] C. Xing, S. Ma, Y. Wu, T. Ng: Transceiver Design for Dual-Hop Non-
regenerative MIMO-OFDM Relay Systems Under Channel Uncertain-
ties, IEEE Trans. Signal Process., vol.58, no.12, pp.6325-6339, Dec.
2010.
[28] Y. Zhou, J. Wang, M. Sawahashi: Downlink transmission of broadband
OFCDM systemsPart II: Effect of Doppler Shift, IEEE Trans. Com-
mun., vol. COM-54, no. 6, pp. 1097-1108, June 2006.
[29] Y. Zhou, J. Wang, M. Sawahashi: Downlink transmission of broadband
OFCDM systemsPart I: Hybrid Detection, IEEE Trans. Commun., vol.
COM-53, no. 4, pp. 718-729, April 2005.
[30] Y. Li, L. Cimini: Bounds on the interchannel interference of OFDM in
time-varying impairments, IEEE Trans. Commun., vol.49, no.3, pp.401-
404, March 2001.
[31] 3GPP TS 36.211: Evolved Universal Terrestrial Radio Access (E-
UTRA); Physical channels and modulation, v9.0.0, Dec.2009.
[32] H. Zhu, J. Wang: Chunk-based resource allocation in OFDMA systems
- Part I: chunk allocation, IEEE Trans. Commun., vol. 57, no. 9, pp.
2734-2744, Sept. 2009.
[33] K. Itoh, S. Watanabe, J. Shih, T. Sato: Performance of Handoff Algo-
rithm Based on Distance and RSSI Measurements, IEEE Trans. Veh.
Technol., vol. 51, no. 6, pp.1460-1468, Nov. 2002.
[34] J. Wang, L. B. Milstein: CDMA overlay situations for microcellular
mobile communications, IEEE Trans. Commun., vol. 43, no. 2/3/4, pp.
603-614, Feb/March/April 1995.
[35] J. Wang, J. Chen: Performance of wideband CDMA systems with
complex spreading and imperfect channel estimation, IEEE J. S. Areas
Commun., vol. 19, pp. 152-163, Jan. 2001.
[36] 3GPP TS 36.521: Evolved Universal Terrestrial Radio Access (E-
UTRA); User Equipment (UE) conformance specication Radio trans-
mission and reception Part 1: Conformance Testing, v9.3.0, Dec.2010.
Lin Tian (M07) is an assistant researcher of Wire-
less Communication Technology Research Cen-
ter, Institute of Computing Technology, Chinese
Academy of Sciences. She received her B.S and
M.S. degree from Beihang University, Beijing,
China in 2002 and 2005 respectively. She is cur-
rently working toward Ph.D. degree in Institute of
Computing Technology, Chinese Academy of Sci-
ences.
Her research interests include wireless resource
management and multimedia multicast scheme in
next-generation mobile communication systems. She has published more than
twenty research papers in IEEE journals and international conferences. She is
also the inventor of more than ten Chinese patents and pending applications.
She currently serves as the Publication Chair of the 7th International ICST
Conference on Communications and Networking in China, and as reviewers
for numerous referred journals and international conferences.
Juan Li received her B.S. degree in computer
science and technology from University of Science
and Technology Beijing, China, in 2005. She is cur-
rently working toward the M.S. degree in Wireless
Communication Technology Research Center, Insti-
tute of Computing Technology, Chinese Academy
of Sciences, Beijing, China. Her research interest
is wireless resource management technology under
high speed moving environment.
718 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 30, NO. 4, MAY 2012
Yi Huang (S07-M10) received the B.S. and M.S.
degree in computer sciences and technology from
the Central South University, China in 2004 and
2007, respectively. In July 2010, he received the
Ph.D. degree in computer architecture from the In-
stitute of Computing Technology, Chinese Academy
of Sciences, China.
He is currently an assistant professor in Wireless
Communication Technology Research Center, Insti-
tute of Computing Technology, Chinese Academy of
Sciences. He is an author of over 10 research papers
and has applied 11 Chinese patents. His research focuses on wireless resource
management mechanisms, multicast services in cellular networks, advanced
air-interface protocol design and implementation. He is in charge of sub
projects in National Grand Project for Broadband Wireless Communication
Networks under grant No.2009ZX03003-007 and 2010ZX03003-004.
Jinglin Shi (M99) is a full professor of Wireless
Communication Technology Research Center, Insti-
tute of Computing Technology, Chinese Academy
of Sciences. He is also the visiting professor of
University of Sydney and Beijing University of
Posts and Telecommunications. He received his B.S
degree from Southwest Normal University, China, in
1993, M.S. degree from Yunnan University, China
in 1996, and Ph.D. degree from Beijing Institute of
Technology, China, in 1999.
His research interests include: network system
of broadband wireless communication, intelligent management and control
signaling system for the next generation heterogeneous network, air interface
architecture of next generation mobile communication network. He has
published two books in the network eld and more than fty research papers
in IEEE journals and international conferences. He is the inventor of more
than forty Chinese patents and pending applications. Dr. Shi serves as a Vice
Director of Key lab of Mobile Computing and Pervasive Device, Beijing,
China.
Jihua Zhou is a vice chief engineer of Chongqing
Jinmei Communications Co. Ltd., Chongqing,
China. He received the Ph.D. degree from the In-
stitute of Computing Technology, Chinese Academy
of Sciences, China in 2008. His research focuses on
broadband wireless communication systems design
and implementation. He has published more than
10 research papers in technical journals and interna-
tional conferences. He is the inventor of more than
15 Chinese patents and pending applications.