Content uploaded by Alessandro Bazzi
Author content
All content in this area was uploaded by Alessandro Bazzi
Content may be subject to copyright.
A Softer Vertical Handover Algorithm for
Heterogeneous Wireless Access Networks
Alessandro Bazzi
WiLab, IEIIT-BO/CNR, DEIS-University of Bologna,
V.le Risorgimento 2, 40136 Bologna, Italy.
Email: alessandro.bazzi@cnr.it
Abstract—Future wireless access will involve a variety of
integrated and interworking technologies. Depending on the
conditions (e.g., the specific application, the quality of service
requirements, the user mobility) each session will be served
through the most suited access network; following some event,
there will be a variation of the access technology, which is called
vertical handover. Although the use of one technology at a time is
mostly envisioned, the use of multiple wireless links at the same
time, hereafter called parallel transmission, is also envisioned for
higher throughput. On the basis of parallel transmission, in this
paper the original definition of softer vertical handover is given
and an algorithm for softer vertical handover in heterogeneous
wireless access networks is proposed; the achievable improvement
is shown in two reference scenarios.
I. INTRODUCTION
The evolutions of wireless LANs, cellular systems and other
wireless technologies allow today to have broadband access
anywhere at reasonable costs and without the need of annoying
cables. The next step is the integration and interworking of
these heterogeneous radio access technologies (RATs) with
the “always best connected” paradigm for the final user: the
different technologies will be automatically and optimally
managed by the terminal. How to do this is a large field
of research and includes different issues, from the way to
physically integrate the different RATs to algorithms for multi
radio resource management: to which RAT should a terminal
connect as a function of its needs? When and how should a
terminal modify the technology it is using?
The term used to indicate that a connection between a
terminal and the network moves from the signal exchange to
one source to a different one is handover. If a single RAT
is considered, it corresponds to the variation of the point of
access: From a Node-B to another Node-B in UMTS systems;
from an access point (AP) to another AP in WLAN systems.
In these cases the handover is said horizontal. The mechanism
through which a terminal moves from a RAT to a different one
is called vertical handover [1].
Handovers can be furthermore divided into hard, soft and
softer. In a single RAT network it is a (horizontal) hard
handover when the old radio link is removed before the new
one is established; only hard handovers are possible in most
of actual technologies, including WLANs systems and cellular
technologies that do not use CDMA. In CDMA systems, the
old and the new radio links may be simultaneously active
during the handover phase; this is called either soft or softer
handover. In particular, soft handover refers to links corre-
sponding to different bases; in this case each base decodes the
message independently from the others and multiple versions
of the same packet are sent to a coordinating node, with an
increased probability that one correct copy of the packet is
received. Softer handover refers to links corresponding to the
same base; in this case, diversity combining techniques are
possible at the receiver, before message decoding.
Differently from the case of single RATs, where the terms
hard, soft and softer have an agreed meaning, the terms softer
vertical handover are never used in literature and soft vertical
handover assume different meanings: in most cases they refer
to packets duplicated in all available links in order to increase
the reliability; for example, packets duplication is envisioned
for voice calls in [2], for video streaming applications in [3],
and for data transfers with TCP protocols in [4], [5]; the same
terms correspond to the activation of multiple links with the
use of one of them at a time in [6], [7], [8]; finally, they
mean a smart distribution of the packets over multiple RATs
(hereafter denoted as parallel transmission) in [9].
With the aim to include both single and multiple RATs, we
assume the following definitions, that differ in the number of
simultaneously established links and in their use:
•in hard handovers only one link is present and data is
delivered through that link; the old connection is released
before the new one is established;
•in semi-soft handovers more than one link are simul-
taneously present, although only one is used for data
transmission at a time;
•in soft handovers more than one link are present and
data is delivered through all of them, with resources
independently managed; in this case data duplication is
used to exploit diversity;
•in softer handovers more than one link are present and
data is delivered through all of them, with resources
jointly managed.
Following these definitions, soft vertical handover is applied
in [2], [3], [4], [5] and semi-soft vertical handover in [6],
[7], [8]. At the best of our knowledge, only in [9] a softer
vertical handover algorithm is discussed. Similarly to here,
in [9] Yang et al. consider the parallel use of resources of
different technologies, defining an active set and an addition
phase. Differently from here, they do not consider any loss in
throughput following a non-perfect packet distribution and do
not discuss a deletion phase. However, the main difference is
that the algorithm proposed in [9] acts following a throughput
request per user, while here we consider a best effort use of
the available resources.
In the next section the background of our research is briefly
discussed, before illustrating the algorithm itself (in Sec.III).
In Sec.IV the analytical model adopted for numerical results is
shown and the adopted figures of merit are defined. In Sec.V
two scenarios will be discussed, highlighting why and when
the proposed solution allows to obtain better performance with
comparison to other choices. Conclusions are drawn in Sec.VI.
II. BACKGROUND:SOFT/SOFTER HORIZONTAL HANDOVER
AND PARALLEL TRANSMISSION
The algorithm we are proposing in this paper aims at
bringing the concepts of cellular systems handovers to hetero-
geneous networks. In cellular systems, soft/softer (horizontal)
handovers allow those terminals that are on cell edges to
perceive a higher signal to noise ratio or a lower packet error
probability. Soft/softer handovers are not performed when the
signal from one base is received with significantly higher
power level; in such a case, in fact, a terminal would occupy
the resources in other cells without gaining any advantage.
When and how these concepts can be of some help in
heterogeneous wireless networks? In [10], we demonstrated
that the use of multiple links allows to reach a throughput
which is the sum of the throughput perceived through each of
the RATs alone. We also demonstrated that this is possible only
with a careful distribution of the packets over the various links;
a not correctly balanced distribution of the packets over the
links does not bring to same results. In Fig.11, the throughput
perceived by a terminal with two active connections (either
UMTS-WLAN or WiMAX-WLAN) is given as a function of
the ratio PWLAN of packets transmitted through the WLAN. It
is shown that the sum of the throughput is reached for a precise
value of PWLAN and that a significant reduction follows even
a small deviation from that value. In [10] we concluded that
the presence of multiple links allows better performance, but
the gain is a function of the distribution of packets over the
RATs; if one RAT allows a throughput which is much higher
than the others, parallel transmission should be avoided.
III. THE SOFTER VERTICAL HANDOVER ALGORITHM
The algorithm we propose hereafter, which focuses on a
best effort service, directly derives from the one adopted in
UMTS systems (a detailed description of UMTS algorithms
can be found for example in [11]).
The algorithm acts in multi-mode terminals through the use
of two dynamic lists containing an updated measurement or
estimation of the throughput each RAT would give if used
alone. The first list is the multi radio active set (hereafter
denoted with A), which includes the RATs the terminal is
currently connected to (with parallel transmission); in this
1Fig.1 is taken from [10], with authors permission.
00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
5
10
15
20
25
30
35
PWLAN
TCP level perceived throughput [Mbps]
WLAN−WiMax, simulation
WLAN−UMTS, simulation
WLAN−WiMax, analitical
WLAN−UMTS, analitical
Fig. 1. TCP level throughput adopting a WLAN connection and a WiMax
or UMTS one, as a function of the probability that the packet is transferred
through the WLAN.
list, the RAT which allows the highest throughput is remarked
as best RAT. The second list is the multi radio monitoring
set (hereafter denoted with M) and includes those RATs,
currently not included in A, to which the terminal is capable
to connect if convenient.
Let us assume that the multi-mode terminal is under cover-
age of RRATs. The union of Mand Acorrespond to these
RRATs. The value of the estimated or measured throughput
Tj(t)at time tper each RAT j∈[1,R]is monitored
periodically by the terminal: throughput measurements update
data in A, while throughput estimations update data in M.
Initially (at time t=0), Awill include RAT bwith the highest
estimated throughput and all RATs with a sufficient estimated
throughput, following Eq.1 and Eq.2:
b=argmaxj∈[1,R](Tj(0)) (1)
A={j:Tj(0) >ξ
ATb(0)}(2)
where ξA(0≤ξA≤1) represents the minimum throughput
with respect to that of the best RAT that is needed to gain a
sufficient advantage. If none of the RATs keep to Eq.2, then
the terminal will be connected only to the best RAT.
During the active session, as a consequence of modifica-
tions in the conditions (due, for example, to mobility), the
measured/estimated throughput vary in time and the following
events may occur:
•abest RAT replacement is performed in the terminal any-
time the throughput of the actual best RAT is exceeded
by another one of A; this replacement has an impact only
in the multi radio active set terminal list;
•an addition is performed any time the throughput of RAT
j, which is in M, keeps to the following equation:
Tj(t)>ξ
ATb(t)(3)
where b is the best RAT; in this case a softer vertical
handover is performed to RAT jand RAT jis moved
from Mto A;
•adeletion is performed any time the throughput of RAT
j, which is in A, keeps to the following equation:
Tj(t)<ξ
DTb(t)(4)
where ξD(0≤ξD≤ξA) represents the maximum
throughput with respect to that of the best RAT that
allows a sufficient advantage; in this case the connection
to RAT jis ended and RAT jis moved from Ato M.
The definition of two different thresholds for the add and
delete phase (ξAand ξD, respectively) is provided to allow
some hysteresis.
IV. ANALYTICAL MODEL AND FIGURES OF MERIT
A. Analytical model
In order to evaluate the performance of the proposed algo-
rithm, a simple model is hereafter assumed. Let us consider
the Shannon capacity limit in AWGN channels:
C=Blog2(1 + S
N)(5)
where Cis the maximum capacity of the channel (in bit/s),
Bis the available bandwidth (in Hz), Sis the useful received
signal strength (in W), and Nis the noise power level at the
receiver (in W). Eq.5 also applies if white and gaussian inter-
ference sources are considered: in this case, Ncorresponds to
the sum of noise and interference power at the receiver.
In order to allow transmissions to reach a data rate as
near to the limit as possible, all modern RATs provide a set
of modulation schemes and channel coding rates to tradeoff
throughput with reliability; more robust combinations (lower
code rate and lower modulation level) are selected far from the
signal source, where S/N is lower; combinations with higher
throughput (higher code rate and higher modulation level) are
chosen near to the signal source, where S/N is higher.
Starting from Eq.5, the following assumptions are consid-
ered for throughput calculation:
•the total resource is assumed to be equally distributed
among active users; this means, e.g., that the same num-
ber of OFDMA-slots are given to active users in a Mobile
WiMAX system or that the transmission opportunity is
the same for all users in an IEEE802.11e system;
•a loss in average S/N should be considered if the
scenario differs from the AWGN model;
•a reduction with respect to the maximum achievable
capacity has to be taken into account, due to the overhead
introduced by all protocols involved from the physical
level up to the application;
•a further reduction shall be introduced as the number
of active users increases, due to the medium access
overhead; referring to the same examples as before, an
increasing number of OFDMA-slots are needed to map
the users resources in a Mobile WiMAX system, and
more collisions occur in an IEEE802.11e system.
As a consequence of the discussed assumptions, considering a
base jand Mj(t)active terminals connected to it at time t,the
generic terminal kat distance dj
k(t)from the base perceives in
downlink (the equation can easily be obtained for the uplink
case also) a throughput Tj
k(t)equal to:
Tj
k(t)=(1−Δj(t)) Bj
Mj(t)log21+ Sj
k(dj
k(t))
Nj
k
1
ηj(6)
with
Δj(t)=Δ
j
F+Mj(t)Δj
u(7)
Sj
k(dj
k)=(PtjGt
j
kGr
j
k)/(AtjArkPL
j
k(dj
k)) (8)
and
Nj
k=kBT0NFkBeqj
k(9)
where Δj(t)is the total loss due to protocols overhead, Δj
F
is the loss due to protocols overhead independently from the
number of active connections, Δj
uis the loss due to each
active connection, Bjis the available bandwidth, ηjis the
loss in S/N due to non AWGN conditions, Ptjis the power
transmitted by the base, Gt
j
kis the antenna gain at the base in
the terminal direction, Gr
j
kis the antenna gain at the terminal
in the direction of the base, Atjis the power loss at the base
transmitter, Arkis the power loss at the terminal receiver,
PL
j
k(dj
k)is the path loss as a function of the distance, kB
is the Boltzmann constant, T0is the reference temperature,
NFkis the noise figure of the terminal receiver, Beqj
kis the
equivalent noise power bandwidth at the terminal receiver.
When parallel transmission is performed, then a throughput
equal to the sum of the throughput perceived by the single
RATs is assumed, with a loss due to imperfect distribution of
packets as it follows from the considerations in Sec.II. Thus,
if the generic terminal kis connected to RARATs, its total
throughput Tk(t)is:
Tk(t)=(1−δ)ΣRA
j=1Tj
k(t),(10)
where δ∈[0,1] takes into account the loss due to imperfect
distribution.
B. Figures of merit
In the following, the performance of the system will be
investigated in terms of throughput and fairness. Denoting the
number of terminals with M, the following definitions apply:
•average throughput of terminal k
ATk=<T
k(t)>(11)
•average throughput per user
AT =1
M
M
i=1
ATi(12)
•fairness (as defined in [12])
F=1−AD
AT ·M
2(M−1) (13)
where AD =M
i=1 |ATi−AT |
Mis the average distance
between users throughput and the average throughput per
user. Please note that Franges from 1 to 0, where 1 is
achieved when all users perceive the same throughput
(AD =0), while 0 is achieved when one user perceives
a given throughput whereas all other users perceive no
throughput (proof is given in [12]).
V. E XAMPLES OF APPLICATION:SCENARIOS AND RESULTS
Two scenarios of application are considered in our work,
allowing to clarify the softer vertical handover advantages.
They correspond to a scenario with mobility and a scenario
with an increasing number of active users. In both cases,
two bases 100m far from each other (denoted with b1and
b2) are assumed, with different RATs (RAT1and RAT2,
respectively). Among all users, only one user (denoted with
uX) has a dual mode terminal, able to connect to either RAT
and possibly performing parallel transmission over the two
RATs. Four cases are compared:
1) single RAT:uXuses only one RAT; he never performs
vertical handovers;
2) parallel transmission:uXalways performs parallel
transmission;
3) hard VHO:uXis always connected to the RAT which
guarantees the highest throughput, i.e. it performs hard
vertical handover whenever the RAT he is connected to
is not the best solution anymore;
4) softer VHO:uXacts following the proposed algorithm.
In our results, parameters are set as follows:
•for all RATs and terminals, antenna gains and losses
are equal to 0dB (omnidirectional antennas, negligible
losses), NF=10dB at the receiver, Beq j
k=Bj;
•parameters referred to b1:Bb1=20MHz, Ptb1=20dBm,
Δb1
F=0.85,Δb1
u=0.001,ηb1=0dB;
•parameters referred to b2:Bb2=7MHz, Ptb2=30dBm,
Δb2
F=0.7,Δb2
u=0.001,ηb2=0dB;
•for both RATs and all users PL(d) = 32 + 40 log10 (d)
expressed in dB, with din meters;
•the two RATs do not interfere to each other, meaning
they occupy orthogonal frequency bands;
•δ=0.1;
•ξA=ξD=0.5.
Although the scope of the present paper is to show the advan-
tage of softer vertical handover, without considering specific
RATs, still some investigations were performed in order to
have realistic values for the parameters. As an example, taking
the curves shown in [13], where MAC level throughput of
IEEE802.11a systems in indoor environment is considered, a
good match is obtained with ΔF=0.83 and η=2dB for S/N
between 5and 30dB; as a second example, with reference
to the curves of Mobile WiMAX PHY level throughput in
SUI1 channels shown in [14], a good match is obtained with
ΔF=0.25 and η=6dB for S/N between 5and 25dB.
A. Scenario 1: mobility
In the first scenario, M(t)=3users (u1,u2,uX)are
considered, as follows:
•u1stands still, 10mformb1and 90mfromb2,
•u2stands still, 90mformb1and 10mfromb2,
•uXmoves from the same position as u1to the same
position as u2, at a speed of 1m/s.
In the single RAT case uXis connected to b1.
Results are shown in Fig.2 and Tab.I.
In Fig.2, ATk(t)is shown as a function of time per each
user. The four subfigures refer to the four described cases.
It can be noted that, in the case of single RAT (Fig.2(a))
u2always use the whole bandwidth of b2, whereas u1and
uXshare that of b1; due to the increasing distance from b1,
throughput of uXdecreases in time. In the case of parallel
transmission (Fig.2(b)), both u1and u2share the bandwidth
of their own base with uX.u2perceives thus a throughput
which is half that of the previous case. uXgains a noticeable
advantage from the parallel transmission. Observing the hard
VHO (Fig.2(c)) case, it can be noted that while u1and u2have
an overall advantage with respect to the parallel transmission
case, uXsuffers of a lower throughput when he is located at
mid distance from the two bases.
Finally, the softer VHO case is shown in Fig.2(d). In this
case, uXacts following the algorithm discussed in Sec.III;
initially, Aonly includes RAT1, while RAT2is in M;uX
only perceives throughput from the nearer b1, not reducing the
throughput of u2.Att=13sAof uXvaries, also including
RAT2, thus allowing uXto perceive a throughput never lower
than approximately 10Mbit/s; in this case Mis empty. Finally,
at t=46sRAT1is deleted from Aand included in M;uX
only perceives throughput from the nearer b2, not reducing the
throughput of u1.
In Tab.I, average values are shown in order to allow fair
comparison of the four cases. For all cases, ATkper each user,
AT and Fare shown, highlighting advantages and weaknesses
of the four cases. The maximum AT is obtained in the hard
VHO case, while the maximum Fis reached in the parallel
transmission case. It must be noted, however, that the softer
VHO allows the best compromise between the two parameters:
uXperceives a throughput which is near to his maximum, not
heavily affecting the other two users.
B. Scenario 2: variable number of users
In the second scenario, M(t)=3+t/10users are
considered2(i.e., the number of active users increases of 1
unit each 10s), with no mobility, as follows:
•one user, u1,is40mformb1and 60mfromb2,
•M(t)−1users (uXand all uk, with k≥2)are60m
from b1and 40mfromb2; only uXand u2are active at
time t=0s, while the other users start their connection
at regular intervals: in particular, user uk, with k>2,
starts its session at the instant t= (10k−20)s.
In the single RAT case uXis connected to b2.
In Fig.3, AT1(t),AT2(t), and ATX(t)are shown as a
function of time. Please note that ATk(t)can be easily derived
also for any user k, with k>2, since in that case ATk(t)=0
2xdenotes the highest integer not higher than x.
010 20 30 40 50 60 70 80
0
5
10
15
20
25
30
35
40
45
50
Time [s]
Throughput [Mbit/s]
T1(t)
T2(t)
TX(t)
(a) Single RAT
010 20 30 40 50 60 70 80
0
5
10
15
20
25
30
35
40
45
50
Time [s]
Throughput [Mbit/s]
T1(t)
T2(t)
TX(t)
(b) Parallel transmission
010 20 30 40 50 60 70 80
0
5
10
15
20
25
30
35
40
45
50
Time [s]
Throughput [Mbit/s]
T1(t)
T2(t)
TX(t)
(c) Hard VHO
010 20 30 40 50 60 70 80
0
5
10
15
20
25
30
35
40
45
50
Time [s]
Throughput [Mbit/s]
T1(t)
T2(t)
TX(t)
AX={RAT
1}
MX={RAT
2}
AX={RAT
1,RAT
2}
MX=0
AX={RAT
2}
MX={RAT
1}
(d) Softer VHO
Fig. 2. Throughput in time for the three considered users. Scenario 1: u1near b1,u2near b2,uXmoving from b1to b2. In Fig.2(d), AXand MXdenote
Aand Mof uX.
TAB L E I
SUMMARY OF NUMERICAL RESULTS IN THE TWO SCENARIOS
Case AT F AT1AT2ATX
[Mbps] [Mbps] [Mbps] [Mbps]
Single RAT 21.112 0.619 19.16 37.22 6.96
Sc.1 Parallel tr. 17.574 0.927 19.16 18.55 15.02
Hard VHO 22.890 0.758 31.62 25.23 11.83
Softer VHO 20.978 0.830 27.55 21.54 13.84
Single RAT 8.672 0.642 14.87 5.57 5.57
Sc.2 Parallel tr. 7.245 0.884 7.39 5.57 8.78
Hard VHO 7.501 0.826 10.11 5.87 6.53
Softer VHO 7.498 0.874 8.75 5.61 8.14
per t<(10k−20)s and ATk(t)=AT2(t)per t≥(10k−20)s.
The four sub-figures refer to the four described cases. Looking
at Fig.3(d), which refers to the softer VHO case, the effects
of the proposed algorithm can be noted: initially, Aof uX
only includes RAT2, while RAT1is in M;uXonly perceives
throughput from b2, not reducing the throughput of u1.After
20s, when 4users are sharing b2bandwidth, the contribution
of RAT1becomes relevant for uXand is thus included in
A.Att=90s the contribution of RAT2is not significant
anymore to uXand is thus removed from A, leaving some
more bandwidth to all other users connected to b2.
In Tab.I AT ,F,AT1,AT2, and ATXare shown in the
four investigated cases. Both AT and Fare calculated without
taking into account users uk, with k>2.
Similar conclusions to those of scenario 1 can be drawn
observing Fig.3 and Tab.I. Although Single RAT gives the
highest AT and parallel transmission allows the highest F,
softer VHO outperforms the other solutions if results are
observed as a whole.
VI. CONCLUSIONS
In this paper, a definition for softer vertical handover
was proposed in order to solve the ambiguity that can be
observed in literature on the terms soft vertical handovers.
A softer vertical handover algorithm is then proposed and its
performance is evaluated in two reference scenarios.
ACKNOWLEDGMENT
This work was carried out in the framework of the PEGA-
SUS Project, funded by MSE. The author would like to thank
Prof. O. Andrisano for motivating and supporting this research
activity and colleagues at WiLab for helpful discussions.
020 40 60 80 100
0
5
10
15
20
25
30
Time [s]
Throughput [Mbit/s]
T1(t)
T2(t)
TX(t)
(a) Single RAT
020 40 60 80 100
0
5
10
15
20
25
30
Time [s]
Throughput [Mbit/s]
T1(t)
T2(t)
TX(t)
(b) Parallel transmission
020 40 60 80 100
0
5
10
15
20
25
30
Time [s]
Throughput [Mbit/s]
T1(t)
T2(t)
TX(t)
(c) Hard VHO
020 40 60 80 100
0
5
10
15
20
25
30
Time [s]
Throughput [Mbit/s]
T1(t)
T2(t)
TX(t)
AX={RAT
2}
MX={RAT
1}
AX={RAT
1,RAT
2}
MX=0
AX={RAT
1}
MX={RAT
2}
(d) Softer VHO
Fig. 3. Throughput in time for the three considered users. Scenario 2: u1nearer to b1, all other users nearer to b2; other users include uX,u2and a new
user each 10s(u3from t=10s, u4from t=20s and so on). In Fig.3(d), AXand MXdenote Aand Mof uX.
REFERENCES
[1] M. Stemm and R. Katz, “Vertical handoffs in wireless overlay networks,”
ACM Mobile Networks and Applications (MONET), vol. 3, p. 335350,
January 1998.
[2] R. Ben Ali and S. Pierre, “On the impact of soft vertical handoff on
optimal voice admission control in PCF-based WLANs loosely coupled
to 3G networks,” IEEE Transactions on Wireless Communications,
vol. 8, pp. 1356 – 1365, March 2009.
[3] G. Cunningham, P. Perry, and L. Murphy, “Soft, vertical handover
of streamed video,” in 5th IEE International Conference on Mobile
Communication Technologies, 2004. (3G 2004). London, UK: IEE,
October 2004, pp. 432 – 436.
[4] H. Naoe, M. Wetterwald, and C. Bonnet, “IPv6 soft handover applied
to network mobility over heterogeneous access networks.” in IEEE 8th
International Symposium on Personal, Indoor and Mobile Radio Com-
munications, 2007. (PIMRC 2007). Athens, Greece: IEEE, September
2007, pp. 1–5.
[5] N.-C. Wang, Y.-Y. Wang, and S.-C. Chang, “A fast adaptive congestion
control scheme for improving TCP performance during soft vertical
handoff.” in IEEE Wireless Communications and Networking Confer-
ence, 2007. (WCNC 2007). Hong Kong: IEEE, March 2007, pp. 3641–
3646.
[6] R. Chakravorty, P. Vidales, K. Subramanian, I. Pratt, and J. Crowcroft,
“Performance issues with vertical handovers - experiences from GPRS
cellular and WLAN hot-spots integration.” in the 2nd IEEE Annual Con-
ference on Pervasive Computing and Communications, 2004. (PerCom
2004). Orlando, Florida, USA: IEEE, March 2004, pp. 155–164.
[7] H. Rutagemwa, S. Pack, X. Shen, and J. Mark, “Robust cross-layer
design of wireless-profiled TCP mobile receiver for vertical handover.”
IEEE Transactions on Vehicular Technology., vol. 56, pp. 3899 – 3911,
Nov. 2007.
[8] H. Huang and J. Cai, “Adding network-layer intelligence to mobile
receivers for solving spurious TCP timeout during vertical handoff.”
IEEE Network, vol. 20, pp. 24 – 31, Nov.-Dec. 2006.
[9] K. Yang, I. Gondal, and B. Qiu, “Context aware vertical soft handoff
algorithm for heterogeneous wireless networks,” in IEEE 68th Vehicular
Technology Conference, 2008. (VTC 2008-Fall). Calgary, Canada:
IEEE, September 2008, pp. 1–5.
[10] A. Bazzi, G. Pasolini, and O. Andrisano, “Multiradio resource manage-
ment: Parallel transmission for higher throughput?” EURASIP Journal
on Advances in Signal Processing, vol. 2008, article ID 763264, 9 pages,
2008. doi:10.1155/2008/763264.
[11] H. Holma and A. Toskala, WCDMA for UMTS, 3rd Ed. Wiley, 2005.
[12] A. Bazzi, N. Dimitriou, and A. Conti, “Adaptive cross-layer techniques
for cellular systems and WLANs: Simulative results within NEWCom
Proj.C,” in IEEE 65th International Vehicular Technology Conference,
2007. (VTC2007-Spring). , Dublin, Ireland, April 2007, pp. 788 – 793.
[13] A. Doufexi, S. Armour, M. Butler, A. Nix, D. Bull, J. McGeehan, and
P. Karlsson, “A comparison of the HIPERLAN/2 and IEEE 802.11a
wireless LAN standards,” IEEE Communications Magazine, vol. 40, pp.
172 – 180, May 2002.
[14] O. Alim and A. El Naggary, “Coverage vs throughput challenges in
Mobile WiMAX.” in National Radio Science Conference, 2009. (NRSC
2009). New Cairo, Egypt: IEEE, March 2009, pp. 1–9.
