Int. J. Communications, Network and System Sciences, 2011, 4, 495-506
doi:10.4236/ijcns.2011.48061 Published Online August 2011 (http://www.SciRP.org/journal/ijcns)
Dynamic Handoff Priority Adjustment Based on
Mobility-Aw areness in Multimedia Cellular Networks
Chow-Sing Lin1, Pin-Jing Huang2
1Department of Computer Science and Information Engineering, National University of Tainan, Taiwan, Chinese Taipei
2Department of Information Management, Southern Taiwan University, Tainan, Taiwan, Chinese Taipei
Received June 1, 2011; revised July 6, 2011; accepted July 15, 2011
Abstract
In multimedia cellular networks, when a Mobile Host requests multimedia services, it may experience hand-
offs to several cells before the request is completely served. If a target cell cannot provide adequate band-
width for a service request, instead of directly dropping the request, the MH is put into the handoff queue and
hopefully the requested bandwidth can be satisfied by later released bandwidth. Obviously, it is important to
properly assign priorities for queued handoff of MHs based on their inborn dynamics to avoid unnecessary
dropping. In this paper, we present a dynamic handoff priority adjustment scheme which applied a handoff
queuing scheme to dynamically adjust handoff priority based on receiving signal strength, service class, and
mobility of Mobile Hosts. In addition, idle bandwidth reserved by inactive MHs is reallocated to urgent
handoff MHs to reduce the call dropping probability. The goal of the proposed dynamic handoff priority ad-
justment scheme is to further reduce call dropping probability while still maintaining high bandwidth utiliza-
tion and acceptable call blocking probability on multimedia cellular networks.
Keywords: Multimedia Cellular Network, Handoff Priority, Handoff Queue, Quality of Service, Mobility
1. Introduction
In recent years, due to rapid advances in networking
technology, providing multimedia service on cellular
network now becomes feasible. The types of accessible
multimedia contents, such as data, voice, music, and
video have grown explosively [1-4]. On multimedia cel-
lular networks, Mobile Hosts (MHs) may access various
types of services, and each requires different Qual-
ity-of-Service (QoS). During the service time, a MH may
generally move across several cells. Due to the unpre-
dictable mobility of MHs, how to maintain a consistent
QoS becomes more difficult and challenging on multi-
media cellular networks.
Three metrics are generally used for measuring QoS:
Call Blocking Probability, Call Dropping Probability,
and Bandwidth Utilization. For a new call, if the origin
cell cannot provide sufficient bandwidth, it will be
blocked [5-6]. The call blocking probability (CBP) de-
notes the probability for a new call to be denied by origin
cell. On the other hand, for an on-going call, if its origin
cell no longer has enough bandwidth to maintain the re-
quested service, it issues a handoff request to hand over
the service to the target cell. However, the on-going call
is dropped if the target cell also has insufficient band-
width for continually providing service. The probability
of rejecting a handoff request due to the insufficient
bandwidth is called call dropping probability (CDP).
From a client’s perspective, it is more intolerable to drop
an on-going service, than to block a service that has yet
to be established [3]. Therefore, with limited bandwidth
in a cell, satisfying handoff requests of on-going calls is
more important and thus gives high priority than new
calls, i.e., the CDP should be kept as low as possible in a
cell [7-8]. Finally, the bandwidth utilization (BU) de-
notes the average bandwidth utilization of cells in a mul-
timedia cellular network.
The typical infrastructure of cellular network is made
of several hexagons called cells, and MHs in a cell are
served by a base station (BS). Between two cells there is
an overlap area, called handoff area or handoff zone.
MHs in this handoff area can receive signals from both
BS. When the Received Signal Strength (RSS) of origin
cell is below a certain threshold, a MH is handed over to
the target cell to continue the service. To avoid disrupt-
ing connections during a handoff process, a seamless

C.-S. LIN ET AL. 496
cellular network is constructed [6,9]. MHs can move
from one cell to another by soft handoff to keep enjoying
the service without disruption [11]. In order to increase
system efficiency and therefore to serve more users in
cellular networks, nowadays, cells in a cellular network
tends to shrink from macro cells to micro/pico cells [11,
12]. Nevertheless, such an approach inevitably increases
the number of handoffs among cells, leading to an in-
crease of the CDP [13-15]. Figure 1 depicts the rela-
tionship of RSS and handoff process of MHs. When MHa
moves toward the target cell, CellB, its RSS from BSA
changes accordingly. When MHa reaches its handoff
threshold of the RSS equal to A’, it issues a handoff re-
quest to the BS of target cell, BSB. The RSS of MHa from
CellA gets weaker while MHa moving toward the CellB.
Once it reaches the received threshold of B’, the RSS
from BSA will be too weak to establish a workable con-
nection to MHa. At this point, the connection of MHa will
be disrupted if BSB is unable to provide sufficient band-
width for handoff [16,17]. In other words, the handoff to
CellB for MHa is issued at time t1 and must be completed
before the time t2 to avoid service interruption.
Generally, once a MH enters a handoff zone, it issues
a handoff request to the target cell. If the target cell can-
not provide adequate bandwidth for on-going services
either by bandwidth reservation scheme [18] or by guard
channel scheme [19], instead of directly dropping the
request, the MH is put into the handoff queue, and hope-
fully the requested bandwidth can be satisfied later by
released bandwidth. Obviously, properly assigning pri-
orities based on their inborn dynamics to avoid unneces-
sary dropping is important for queued handoffing MHs.
In multimedia cellular networks, MHs have various
characteristics such as requested service class, moving
velocity, mobility, and RSS, etc., which can be used to
Figure 1. Handoff and RSS [16].
determined the handoff priority of an on-going call. Par-
ticularly, the mobility of a MH may change erratically,
and therefore, its priority in handoff queue must be dy-
namically adjusted based on its mobility to reflect the
urgency of handoff. For example, if a MH in handoff
queue stops moving (stands still) or even moving away
from the target cell, its handoff priority should be gradu-
ally decreased. On the other hand, the handoff priority of
a MH in handoff queue should be timely raised if the MH
accelerates toward the target cell.
In this paper, we proposed a dynamic handoff priority
adjustment (DHAP) scheme to order queued handoffing
requests of MHs. To specify whether a MH is continu-
ously moving or rarely moving, the mobility of a MH
was classified into active mode and inactive mode based
on a threshold of the moving speed. In addition, the
DHAP consisted of bi-handoff queues, which were ac-
tive handoff queue and inactive handoff queue. Handoff-
ing MHs in active and inactive mode failing to acquire
adequate bandwidth in target cell were inserted into ac-
tive handoff queue and inactive handoff queue, respec-
tively. The priority of a MH in active handoff queue was
periodically adjusted by its dwelling time which indi-
cates how soon the MH is to be handoffed to the target
cell. On the other hand, the priority of a MH in inactive
handoff queue is not updated or updated at a large time
interval since the MH rarely moves. The migration be-
tween active and inactive handoff queues might happen
when the mobility of a queued handoff MH changes.
Besides, we also classify MHs who successfully reserve
bandwidth in target cells into mobile set and immobile
set based on their mobility. Idle bandwidth reserved by
inactive MHs in immobile set is reallocated to urgent
handoff MHs to further reduce the CDP.
The remainder of this paper is organized as follows. In
Section 2, related works of handoff queuing schemes are
investigated. The proposed DHPA scheme is presented
in Section 3. Section 4 presents simulation results and
analyses. Finally, concluding remarks are given in Sec-
tion 5.
2. Related Works
To reduce the probability of terminating handoff calls
due to the shortage of available bandwidth in target cell,
handoff queuing scheme have been widely studied in
recent years. The priority of a handoff call in a handoff
queue can be determined by single or multiple factors,
which are described as follows.
The simplest strategy of determining priority for a
queued MH is based on its order of entering handoff area,
such as First-In-First-Out (FIFO). A handoff call is added
into FIFO queue if its target cell has no sufficient avail-

C.-S. LIN ET AL.497
able bandwidth to satisfy the service. When there is
bandwidth released in a cell, the MH in the handoff
queue with the earliest handoff time will obtain the
bandwidth to process its handoff. Obviously, after enter-
ing a handoff area, a MH may stop moving, or even
move backward. Without considering other factors, FIFO
scheme cannot truly reflect the urgency of handoff due to
the erratic mobility of MHs. Qian and Feng [20] present
the minimum-dwelling-time scheme to prioritize handoff
requests. The priority of handoff requests is determined
by its estimated dwelling time in the handoff area. A MH
with shorter dwelling time has higher probability to
handoff to its target cell, and hence should have the
higher handoff priority. A measurement-based prioritiz-
ing scheme (MBPS) [21] has been developed for priori-
tizing the queue instead of FIFO scheme. In MBPS, the
priority is determined only based on the RSS of a MH.
Once a MH enters the handoff area, its RSS may con-
tinuously vary during the handoffing time. The stronger
the RSS of a MH, the lesser urgency of handoff, and
therefore the lower priority in the handoff queue. Com-
pared with the FIFO scheme, the MBPS using the RSS to
determine the handoff priority is more reasonable but
still not accurate for reflecting the urgency of handoff
calls.
To more precisely reflect a MH’s status, instead of us-
ing only one single factor, several studies apply multiple
factors for fine-tuning the handoff priority. Ebersman
and Tonguz [22] proposed a signal prediction priority
queuing (SPPQ) method using not only RSS but also
moving speed, which is interpreted as the change of RSS
(ΔRSS) to determine the priority of a MH. With the same
time interval, a higher value of ΔRS S means faster mov-
ing speed of a MH. Since a MH with high moving speed
is more likely to be handoffed to its target cell than one
with low moving speed, it should have a higher priority
in handoff queue. Therefore, compared with queuing
schemes considering only one factor, the SPPQ assigns a
more proper handoff priority to an on-going MH. How-
ever, in SPPQ only one type of service is considered.
Chang and Leu [16] further extended the SPPQ algo-
rithm, called Signal Strength for Multimedia Communi-
cations (SSMC) to handle multiple service types of mul-
timedia traffic. In SSMC, the service class (Ci), RSSi and
ΔRSSi are considered to determine the priority of a MHi
in handoff queue, Pi, which can be estimated by (1).
1
ii i
i
PC RSSRSS   (1)
The SSMC scheme fails to consider the situation when
a MH stops moving or is even moving backward. For
example, when a MH close to the boundary of handoff
area becomes temporary immobile, as shown in (1), its
handoff priority Pi suddenly becomes zero as the value of
ΔRSS equals zero. If the MH reinstates mobility, its
handoff priority may be too low to acquire available
bandwidth in time, leading to an increase of call drop-
ping probability.
In this paper, we propose a Dynamic Handoff Priority
Adjustment (DHPA) scheme to further extend the SSMC
scheme. In DHPA, we similarly use RSS, ΔRSS, and ser-
vice class to determine the priority of a queued MH for
handoff. In addition, the DHPA scheme further monitors
the mobility of queued MHs handoff to dynamically and
gradually adjust handoff priorities based on their dwell-
ing time. Idle bandwidth reserved by immobile MHs is
also utilized to further reduce the CDP.
3. Dynamic Handoff Priority Adjustment
In this paper, for simplicity of analyses, a precise mov-
ing prediction algorithm [9,23] is assumed so that a
handoffing MH issues bandwidth reservation request to
the target cell once it enters the handoff area. However,
the DHPA could be easily extended to multiple target
cell approach with reserving bandwidth on target cells
based on their handoff probability [18]. In general, a
MH may move in various directions. Then, we assumed
a MH moving toward the target cell is called moving
forward; a MH moving away from the target cell is
called moving backward. In DHPA, depending on ΔRSS,
the mobility of a MH was classified as active mode or
inactive mode. Accordingly, we also denoted bi-handoff
queues in a cell. One was called active handoff queue,
and the other was called inactive handoff queue. A
handoffing MH failing to acquire bandwidth from a tar-
get cell was inserted into the active or inactive handoff
queue based on its mobility. The handoff priorities of
the queued handoff MHs were periodically adjusted to
timely reflect the change of their mobility so that an
urgent MH could have a higher priority to acquire re-
leased bandwidth in target cell to avoid service interrup-
tion. Basically, when there was bandwidth released in
the target cell, MHs in the active handoff queue were
served first, followed by those in the inactive handoff
queue if any bandwidth left. MHs may migrate between
active handoff queue and inactive handoff queue when
their mobility changes with time. Figure 2 shows the
handoff queuing scheme.
3.1. The Classification of the Mobility of MHs
In this paper, the ΔRSS of MHi at time t was primarily
determined by ΔRSS,
1
,
tt
tii
i
RSS RSS
RSS T


  (2)

C.-S. LIN ET AL. 498
Figure 2. Handoff queuing scheme.
where and denoted the RSS of MHi at
time t and time t − 1, respectively, and ΔT equaled the
difference between two consecutive time t and t − 1. If
the ΔRSS was positive, it was moving backward to origin
cell; if negative, it was moving toward target cell; if the
ΔRSS equal to zero, it was immobile. The mobility of a
MH was classified into active and inactive modes, which
can be determined by,
t
i
RSS 1t
i
RSS 
1
1
1
1
active, if
Mobility
inactive, if
tt
ii
tt
ii
RSS wRSSw
RSS wRSSw
2
2




 

 


,
(3)
where δ denoted the threshold of mobility, and w1 and w2
represented the weights of two consecutive ΔRS S. If the
value of 1
12ii
of a MH was lar-
ger than or equal to δ, it means that the MH kept moving
at reasonable speed. Its mobility was in active mode. In
contrast, if the value of
tt
RSS wRSSw

 
1
12ii
of
a MH was less than δ, it means that the MH rarely moved
or even stopped moving, and could be classified as in
inactive mode. In addition, when a MH moved out of a
handoff area, the handoff request was no longer valid
and therefore the MH must be moved out of the handoff
queue.
tt
RSS wRSSw

 
3.2. Call Admission Control
In order to avoid disruption of on-going services, the
target cell provides the released bandwidth to on-going
connections prior to new calls. Consequently, the DHPA
provides different call admission control to new calls and
handoff calls. When a cell, Ci, received a request from a
new call MHi, it first determined whether there was suf-
ficient available bandwidth (BWa) to accept the requested
bandwidth () of the new call. If the available band-
width was sufficient, the cell accepted the request, and
this new call became an on-going call. On the other hand,
if the cell had no sufficient bandwidth, the new call was
inserted into a new call queue waiting for available
bandwidth. Figure 3 shows the flow chart of call admis-
sion control for new calls.
r
i
BW
Figure 3. Call admission control for new call.
When an on-going call entered a handoff zone at time
t, i.e., the current of MHi was less than the
pre-defined signal threshold of issuing handoff,
, it issued a handoff request to the target cell.
Its mobility was changed to active, and then the MH tried
to reserve available bandwidth in the target cell to avoid
service interruption. If the required bandwidth was less
than the available bandwidth of the target cell, the target
cell granted the bandwidth reservation request. The MH
was then added to the mobile set, which is defined in the
later section of bandwidth reservation. If the target cell
had insufficient available bandwidth for the handoff re-
quest, the target cell denied the bandwidth reservation
request, and the handoff request was then put into the
handoff queue with the initial priority calculated by the
DHPA algorithm. Figure 4 shows the flow chart of call
admission control for handoff calls.
t
i
RSS
issuehandoff
i
RSS
3.3. Bandwidth Reservation
Generally, the mobility of a MH might change after suc-
cessfully reserving bandwidth. The DHPA algorithm
classified those handoffing MHs into two sets, which are
the mobile set and the immobile set. The mobile set (Um)
is a set of MHs with the mobility in active mode; the
immobile set (Uim) is a set of MHs with the mobility in
inactive mode. A MH with successful bandwidth reser-
vation is initially put into the mobile set. Furthermore, a
MH may migrate between mobile set and immobile set
based on its mobility which might be changed with time.
When a MH successfully reserved bandwidth in a target
cell, DHPA would initially put those MHs in the mobile
set. These MHs are monitored periodically to check if a
handoff is required by examining the strength of its re-
ceived signal. If the RSSt was less than the RSShandoff, the
MH was handoffed to the target cell. In DHPA, before
handoffed to the target cell, the mobility of MHs was
continuously monitored and updated. If the MH was still

C.-S. LIN ET AL.499
Figure 4. Call admission control for handoff call.
in active mode, it stays in mobile set; otherwise, it mi-
grates to the immobile set. Figure 5 shows the process of
periodically updating MHs in mobile set.
On the other hand, when a MH successfully reserving
bandwidth in the target cell changed to inactive, it was
added into immobile set, Uim. A MH in Uim might be
asked to release the reserved bandwidth in order to avoid
dropping handoff calls, because its reserved bandwidth
in the target cell might not be used for a while. In this
paper, the bandwidth of inactive MHs was aggregated as
a reserved bandwidth pool (RBpool), which was used to
estimate mobility ?
RSSi
t≤RSSi handoff
at each time interval T,
for each MHiin mobile set
insert MHito
immobile set
handover the callyes
no
inactive
remain intact
active
Figure 5. Process of periodically updating MHs in mobile
set.
serve those MHs that require urgent handoffs but had
failed to reserve bandwidth previously. Again, before
handoffed to the target cell, the mobility of MHs was
continuously monitored and updated.
If a MH became mobile, before it changed to active
mode, we checked whether there was enough bandwidth
in RBpool to be allocated to the MH. If so, the requested
bandwidth in RBpool was allocated to the MH; otherwise,
the MH was inserted into the inactive handoff queue.
Finally, the mobility of the MH was changed to active.
Figure 6 shows the process of periodically updating
MHs in immobile set in immobile set.
3.4. Handoff Priority Adjustment
Recall that when a MH entered the handoff area, it issued
a handoff request to the target cell. If the target cell could
not provide adequate bandwidth for requested service,
the MH was inserted into the handoff queue, and hope-
fully the requested bandwidth could be satisfied by later
released bandwidth. Before inserting the MH i into the
handoff queue, the initial priority, , was computed
by (4), where SCi, ΔRSSi, and RSSi denoted the priority
of requested service class, the change in received signal
strength, and received signal strength of MHi, respec-
tively.
init
i
P
init 1
ii i
i
PSCRSS
RSS
  (4)
A higher class of services had a higher value of SC
and should be served first. ΔRSS is the change of signal
strength between two moves. The faster a MH moved,
the larger the ΔRSS and th higher handoff priority. e

C.-S. LIN ET AL. 500
Figure 6. Process of per i odically updating MHs in immobile set.
When a MH moved further away from its origin cell, the
signal receiving from the origin cell became weaker,
resulting in a larger 1/RSS and hence leading to an in-
crease of the handoff priority.
1
1
1
1+, 0
, 0
i
tt
i
tt
i
tt
ii
PD
PDT
PD



i
T
T






(6)
For those MHs residing in the active handoff queue, at
each time interval, first we checked if a MH was moving
out of the handoff area. If a MH in active mode kept
moving toward the target cell and was ready to handoff,
i.e., RSSi
t <= , the MH was handoffed to the
target cell if there was enough bandwidth in the target
cell or in the RBpool; otherwise, the MH was dropped.
On the other hand, if a MH moved backward to the ori-
gin cell, i.e., RSSi
t >= , the MH was removed
from the active handoff queue. If a MH still stayed in the
handoff area, we next determined its mobility. If it
changed to inactive, it was migrated to the inactive hand-
off queue; otherwise, its priority in the active handoff
queue should be periodically updated. In this paper, the
DHPA periodically updated handoff priority of a hand-
offing MH according to its dwelling time, which indi-
cated how soon a MH was expected to be handoffed to
the target cell. Intuitively, a MH which had shorter
dwelling time should have higher probability to be hand-
offed into the target cell, and thus should be given a
higher handoff priority. Assume DTi
t denoted the dwell-
ing time of MHi in the handoff area, denoted
the threshold of receive signal strength to handoff, w1
and w2 denoted the relative weights of RSSi
t and RSSi t−1,
then the dwelling time of MHi in the active handoff
queue could be estimated as (5),
handoff
i
RSS
handoff
i
RSS
handoff
i
RSS
where Pi
t and Pi t−1 denoted the handoff priority for MHi
at time t − 1 and t, and DTi
t indicated the dwelling time
at time t. Figures 7 and 8 show the process and pseudo
code of updating MHs’ statuses in active handoff queue.
On the other hand, if MHs rarely moved in the handoff
area, they became inactive and thus were inserted into
inactive handoff queue. The priority of a MH for insert-
ing into the inactive queue, , could be calculated
by (7)
inactive
i
P

inactive
issuehandoff
issuehandoff
1
1
i
t
ii
t
i
i
it
i
P
SCRSS RSS
RSS
RSS
SC RSS
 
 
i
(7)
Because the RSS of a MH in the inactive queue was
seldom changed, its priority would not to be changed
periodically, or could be updated at a large time interval.
However, similar to MHs in active handoff queue,
they had to periodically check their residence and mobil-
ity in handoff area. If a MH in inactive mode was ready
to be handoffed, i.e., RSSi
t <=, the MH was
handoffed to the target cell if there was enough band-
width in the target cell or in the RBpool for this handoff;
otherwise, the MH was dropped. On the other hand, if a
MH gradually moved backward to the origin cell, i.e.,
RSSi
t >=, the MH was removed from the inac-
tive handoff queue. Furthermore, if a MH still stayed in
the handoff area, we next estimated its mobility. If the
mobility of a MH changed to active, it would be mi-
handoff
i
RSS
handoff
i
RSS
handoff
1
12
.
i
t
tii
itt
i
RSS RSS
DT RSS wRSSw


  (5)
Based on the dwelling time of MHi, we used (6) to pe-
riodically update its handoff priority, Pi
t,

C.-S. LIN ET AL.501
Figure 7. Process of updating MHs’ statuses in active handoff queue.
At each interval, T
For each handoff MH in active handoff queue, MHi
If (RSSit <= ) // move back to target cell
handoff
i
RSS
If there is adequate available bandwidth in the target cell or in RBpool
handoff the MH;
else drop the MH;
Else If (RSSit >= RSSiissuehando ff) //handoff to origin cell
dequeue the MH from the active handoff queue;
Else // still in handoff area
Update the mobility of MHi by Equation (3);
If (MHi is active)
Update MHi handoff priority by Equations (5) and (6);
Else // MHi is inactive
Compute the initial priority, Piinit_inactive by Equation (7);
Migrate the MHi to the inactive handoff queue;
Figure 8. Pseudo code of updating MHs in active handoff queue.
grated to the active handoff queue with the initial priority
calculated by (4); otherwise it remained intact. Figures 9
and 10 show the process and pseudo code of updating
MHs’ statuses in inactive handoff queue.
3.5. Reallocation of Released Bandwidth
Once a target cell finished providing the service for a
MHl, the released bandwidth was returned back to the
available bandwidth pool for reallocation. If there were
any MHs in the handoff queues and the new call queue,
the latest available bandwidth was allocated to queues in
order of priority, which was active handoff queue first,
then inactive handoff queue, and finally the new call
queue. First, if there were pending MHs in active handoff
queue, we searched for the MH with the highest priority
and its requested bandwidth was less or equal to the
available bandwidth. If we found one, its requested
bandwidth was reserved, and it was removed from active
handoff queue and joined the mobile set. The available
bandwidth was then updated accordingly. Such a process
was repeated until no MH satisfied the above criteria
could be found. If there was any available bandwidth left,
the above process was again applied to MHs in the inac-
tive queue. After applying the above two steps, if there
was available bandwidth left, it was allocated to pending
new calls if there was any. The process of reallocation of
released bandwidth detailed process flow is shown in

C.-S. LIN ET AL. 502
Figure 9. The process of updating MHs’ status es in inactive handoff queue.
At each interval, T
For each handoff MH in in active handoff queue, MHi
If (RSSit <= RSSihando ff) // move back to target cell
If there is adequate available bandwidth in the target cell or in RBpool
handoff the MH;
else drop the MH;
Else If (RSSit >= RSSiissuehando ff) //handoff to origin cell
dequeue the MH from the inactive handoff queue;
Else // still in handoff area
Update the mobility of MHi by Equation (3);
If (MHi is active)
Compute the initial priority,
Piinie by Equation (4);
Migrate the MHi to the active handoff queue;
End If
Figure 10. Pseudo code of updating MHs’ statuses in inactive handoff queue.
Figure 11.
3.6. The State Transition of MHs in Handoff
Area
Figure 12 shows the state transition of a MH in handoff
area. In general, once a MH entered the handoff area, it
was in the “Enter Handoff Area” state and a request of
bandwidth reservation in the target cell was issued. If the
bandwidth reservation was granted, the MH was added
into the mobile set and changed to “Mobile Set” state;
otherwise, the MH was inserted into the active handoff
queue and changed to “Active queue” state. MHs in
“Mobile Set”/“Immobile Set” state might change to
“Immobile Set”/“Mobile Set” state if their mobility be
came inactive/active. Similarly, MHs in “Active Queue”/
“Inactive Queue” state might change to “Inactive Queue”/
“Active Queue” state if their mobility became inactive/
active. A MH in “Immobile Set” state might change to
“Inactive Queue” state if its reserved bandwidth in
RBpool was allocated to other MHs for saving call drop-
ping. Finally, when a MH in “Inactive Queue” was allo-
cated the requested bandwidth in the target cell, it
changed to the “Immobile Set” state and its reserved
bandwidth was then added to the RBpool.
4. Simulation Results
The simulation model was assumed to be on a multime-
dia cellular network, which consisted of 25 cells. The

C.-S. LIN ET AL.503
Figure 11. Process of reallocation of released bandwidth.
Figure 12. State transition of a MH in handoff area.
coverage of a cell was 4*4 km2. Each cell had 50 chan-
nels, and each channel had the network capacity of 64
kbps. There were three service classes in the simulation
listed as in Table 1 [16]. The call arrival rate followed
Poisson distribution, and MHs were evenly distributed to
25 cells. Each MH randomly selected a moving direction
[0, 360] degree and speed [30, 90] km. The distribution
of speeds and directions of MHs is listed in Tabl e 2. The
w1 and w2 was equal to 0.7 and 0.3, and the threshold of
mobility, δ, was 0.1. The time interval of updating
statuses of MHs, T, was one second. The transmit power,
μ, was 105 units, and the path-loss exponent, γ, was 30

C.-S. LIN ET AL. 504
Table 1. Multimedia service types.
Service
class
Required
bandwidth (kb/s)
Connection
duration (s)
Handoff
priority
Class 1 64 60 1
Class 2 64*2 60*5 4
Class 3 64*3 60*15 8
Table 2. Distribution of speeds and directions of MHs.
Velocity and direction The amount of MHs
Constant, forward 75%
Constant, backward 5%
Accelerated 30%, forward 5%
Decelerated 30%, forward 5%
Accelerated 60%, forward 5%
Decelerated 60%, forward 5%
units. Finally, the RSShandoff and RSSissuehandoff were 2.061
and 1.569 units.
In simulations, we had investigated the performance of
DHPA in terms of the call blocking probability (CBP),
call dropping probability (CDP), and bandwidth utiliza-
tion (BU) by comparing with other ordering strategies,
first-in-first-out (FIFO), and signal strength for multime-
dia communication (SSMC), and No Priority (NP)
schemes. In the No Priority scheme, new calls and hand-
off calls had the same priority and were served arbitrarily.
There was no queue implemented in this scheme. In
FIFO scheme, the available bandwidth was allocated
MHs based on their arriving order. If there were no
channels left, new calls were simply blocked and handoff
requests were queued in the handoff queue. In the SSMC
scheme, a new call was directly blocked without queuing
if there was no available bandwidth in it origin cell.
Figure 13 shows the CBP of the NP, FIFO, SSMC,
and DHPA ordering schemes with respect to the increase
of call arrival rate. The SSMC and FIFO schemes both
had higher CBP than the DHPA scheme as they favored
providing bandwidth to handoff calls. Without the design
of new call queue, new calls were blocked once a cell
had no sufficient bandwidth. Similar to the SSMC and
FIFO schemes, the DHPA also prioritized serving hand-
off calls, but it further implemented a new call queue to
avoid immediately blocking new calls when a cell was
temporary out of available bandwidth. The NP scheme
had the lowest CBP because it did not differentiate be-
tween new calls and handoff calls. Available bandwidth
was equally shared between new and handoffs calls,
leading to the lowest CBP of all four schemes.
Figure 13. Call blocking probability.
Figure 14 shows the CDP of NP, FIFO, SSMC, and
DHPA schemes with the increase of call arrival rate. As
shown in the figure, the CDP increases along with the
increase of arrival rate for all four schemes. Obviously,
the CDP of the proposed DHPA was much lower than
that of NP, FIFO, and SSMC on average by 192%, 151%
and 100%, respectively. Two key properties revealed the
superiority of DHPA in CDP. First, the priority of active
handoff queue was adjusted mostly based on the dwell-
ing time which was more suitable to reflect the urgency
of handoff, and most importantly, it was updated peri-
odically, unlike the SSMC which statically assigned pri-
ority to each queued MH. Second, in DHPA scheme
bandwidth successfully reserved in target cell by inactive
handoff calls which was assumed not to be used for a
while was released to save dropping urgent handoff calls.
The NP scheme did not provide with handoff queue so
that any MH unable to successfully reserve bandwidth in
target cell was directly dropped, resulting in high CDP.
In the FIFO scheme a handoff priority of a MH was es-
timated without considering its mobility, and urgent
handoff calls might not acquire bandwidth in the target
cell for handoff, resulting in a higher CDP than the
SSMC and DHPA as well. It was noted that both the
DHPA and the SSMC schemes considered SC, RSS, and
ΔRSS in calculating initial handoff priority. However,
ignoring urgent handoffing MHs caused by the change of
mobility and the utilization of bandwidth reserved by
immobile handoff MHs led to higher CDP in the SSMC
than the DHPA.
Figure 15 shows the bandwidth utilization of the NP,
FIFO, SSMC, and DHPA schemes with respect to the
increase of call arrival rate. The bandwidth utilization of
our proposed DHPA obviously was higher than that of
FIFO and SSMC on average by 10% and 9%, and was
lower than the NP scheme. With the DHPA scheme,
bandwidth successfully reserved by inactive handoff
MHs was released to urgent handoff MHs to avoid ser-
vice interruption. As a result, the idleness of bandwidth
was reduced, thus bandwidth utilization of the DHPA

C.-S. LIN ET AL.505
Figure 14. Call dropping probability.
Figure 15. Bandwidth utilization.
was better than that of FIFO and SSMC schemes. As far
as the bandwidth utilization was concerned, the NP
scheme was the best solution which utilized any avail-
able bandwidth to satisfy new and handoff MHs. Without
prioritize providing available bandwidth to handoff MHs,
as shown in Figure 14, the CDP of the NP scheme was
much higher than that of the DHPA.
5. Conclusions
When the target cell cannot provide adequate bandwidth
for a handoff call, instead of dropping the request, queu-
ing approach is the major technique to save dropping
handoff MHs. How to effectively and dynamically reflect
the urgency of queued handoff MHs is very crucial to
maintain the QoS of a MH. In general, the mobility of a
MH in the handoff area is varied with time, where a MH
may move toward target cell or backward to origin cell
with various speeds changed with time, or even stand
still. As a result, the urgency of handoffing to target cell
is varied also. It is vital to dynamically adjust the handoff
priority of MHs based on their mobility to timely reflect
the urgency of handoff in order to ensure QoS for
on-going calls without interruption. In this paper, we
proposed the DHPA scheme which not only dynamically
adjusts handoff priority of MHs based on their mobility
inactive MHs in immobile set, to urgent handoff MHs to
reduce the CDP. The simulation results shown that the
proposed DHPA scheme could have high BU and much
lower CDP with acceptable CBP.
but also further reallocates idle bandwidth, reserved by
. References
] M. El-Kadi, S. Olariu and H. Abdel-Wahab, “A Rate-
6
[1
Based Borrowing Scheme for QoS Provisioning in Mul-
timedia Wireless Networks,” IEEE Transactions on Par-
allel and Distributed Systems, Vol. 13, No. 2, 2002, pp.
156-166. doi:10.1109/71.983943
[2] B. M. Epstein and M. Schwartz, “Predictive Qos Based
Admission Control for Multiclass Traffic in Cellular
Wireless Networks,” IEEE Journal on Selected Areas in
Communications, Vol. 18, No. 3, 2000, pp. 523-534.
doi:10.1109/49.840209
[3] A. Mlla, M. El-Kadi, S. Olariu and P. Todorova, “A Fair
Resource Allocation Protocol for Multimedia Wireless
Networks,” IEEE Transactions on Parallel and Distrib-
uted Systems, Vol. 14, No. 1, 2003, pp. 63-71.
doi:10.1109/TPDS.2003.1167371
[4] F. R. Yu, V. W. S. Wong and V. C. M. Leung, “A New
QoS Provisioning Method for Adaptive Multimedia in
Wireless Networks,” IEEE Transactions on Vehicular
Technology, Vol. 57, No. 3, 2008, pp. 1899-1909.
doi:10.1109/TVT.2007.907023
[5] O. T. W. Yu and V. C. M. Leung, “Adaptive Resource
Allocation for Prioritized Call Admission over an
ATM-Based Wireless PCN,” IEEE Journal on Selected
Areas in Communications, Vol. 15, No. 7, 1997, pp.
1208-1225. doi:10.1109/49.622906
[6] D. K. Kim and D. K. Sung, “Characterization of Soft
Handoff in CDMA Systems,” IEEE Transactions on Ve-
hicular Technology, Vol. 48, No. 4, 1999, pp. 1195-1202.
doi:10.1109/25.775368
[7] C. C. Chao and W. Chen, “Connection Admission Con-
y-of-Ser-
trol for Mobile Multiple-Class Personal Communication
Network,” IEEE Journal on Selected Areas in Commu-
nication, Vol. 15, No. 8, 1997, pp. 1618-1626.
[8] S. Chen and K. Nahrstedt, “Distributed Qualit
vice Routing in Ad-Hoc Networks,” IEEE Journal on
Selected Areas in Communication, Vol. 17, No. 8, 1999,
pp. 1488-1505. doi:10.1109/49.780354
[9] V. Li and W. J. Liao, “Wireless Multimedia Networks,”
S. Kang, “An Adaptive Soft Handover
tsuoka, M. Yoshimura and T. Ohya, “A Robust
doi:10.1109/MIC.2003.1189185
Proceedings of International Symposium on Parallel Ar-
chitectures, Algorithms and Networks, Makati City, May
2002, pp. 69-73.
[10] W. I. Kim and C.
Algorithm for Traffic-Load Shedding in the WCDMA
Mobile Communication System,” IEEE Wireless Com-
munications and Networking, Vol. 2, 2003, pp. 1213-
1217.
[11] H. Ma
Method for Soft IP Handover,” IEEE Internet Computing,
Vol. 7, No. 2, 2003, pp. 18-24.

C.-S. LIN ET AL. 506
. Xiaoxin and N. Shan-
e Allocation in Fixed
[12] T. Fong, P. Henry, K. Leung, Q
karanarayanan, “Radio Resourc
Broadband Wireless Networks,” IEEE Transactions on
Communications, Vol. 46, No. 6, 1998, pp. 806-818.
doi:10.1109/26.681421
[13] M. Chiu and M. Bassiouni, “Predictive Scheme
Handoff Prioritization
for
in Cellular Networks Based on
Mobile Positioning,” IEEE Journal on Selected Areas in
Communication, Vol. 18, No. 3, 2000, pp. 510-522.
doi:10.1109/49.840208
[14] M. El-Kadi, S. Olariu and H. Abdel-Wahab, “A R
Based Borrowing Schem
ate-
e for QoS Provisioning in Mul-
timedia Wireless Networks,” IEEE Transactions on Par-
allel and Distributed Systems, Vol. 13, No. 2, 2002, pp.
156-166. doi:10.1109/71.983943
[15] B. Epstein and M. Schwartz, “Predictive QoS-Based
Admission Control for Multiclass Traffic in Cellular
Wireless Networks,” IEEE Journal on Selected Areas in
Communication, Vol. 18, No. 3, 2000, pp. 523-534.
doi:10.1109/49.840209
[16] R. S. Chang and S. J. Leu, “Handoff Ordering U
Signal Strength for M
sing
ultimedia Communications in
Wireless Networks,” IEEE Transactions on Wireless
Communications, Vol. 3, No. 5, 2004, pp. 1526-1532.
doi:10.1109/TWC.2004.833515
[17] O. T. W. Yu and V. C. M. Leung, “Adaptive Resource
Allocation for Prioritized Call Admission over an ATM-
Based Wireless PCN,” IEEE Journal on Selected Areas in
Communications, Vol. 15, No. 7, 1997, pp. 1208-1225.
doi:10.1109/49.622906
[18] C. Oliverira, J. B. Kim and T. Suda, “An Adaptive
Bandwidth Reservation Scheme for High-Speed Multi-
media Wireless Networks,” IEEE Journal of Selected
Areas in Communications, Vol. 16, No. 6, 1998, pp.
858-874. doi:10.1109/49.709449
[19] H. H. Liu, “Guard Channel Sharing Strategies in Inte-
G. Feng, “A Minimum-Dwelling—Time
ed Priori-
grated Voice/Data Mobile Networks,” Advanced Infor-
mation Networking and Applications, Vol. 2, 2004, pp.
79-82.
[20] R. Qian and
Prioritization Scheme for Handoff in Mobile Cellular
Systems,” International Conference on Communication
Technology, Beijing, 22-24 October 1998, p. 5.
[21] S. Tekiny and B. Jabbari, “A Measurement-Bas
tization Scheme for Handovers in Mobile Cellular Net-
works,” IEEE Journal on Selected Areas in Communica-
tions, Vol. 10, No. 8, 1992, pp. 1343-1350.
doi:10.1109/49.166761
[22] H. G. Ebersman and O. K. Tonguz, “Handoff Ordering
d-
Using Signal Prediction Priority Queuing in Personal
Communication Systems,” IEEE Transactions on Ve-
hicular Technology, Vol. 48, No. 1, 1999, pp. 20-35.
[23] C. Choi, M. Kim, T. Kim and S. J. Kim, “Adaptive Ban
width Reservation Mechanism Using Mobility Probability
in Mobile Multimedia Computing Environment,” Pro-
ceedings of the 25th Annual IEEE Conference on Local
Computer Networks, Tampa, 8-10 November 2000, pp.
76-85. doi:10.1109/LCN.2000.891011