Performance of iCAR Systems: A Simplified

Analysis Technique
Evsen Yanmaz and Ozan K. Tonguz

Hongyi Wu

Chunming Qiao

Department of Electrical and

Center for Advanced Computer Studies Department of Computer Science and
Computer Engineering
University of Louisiana at Lafayette
Engineering
Carnegie Mellon University
Lafayette, LA 70504-4330
SUNY at Buffalo
Pittsburgh, PA 15213-3890
Email: wu@wupc.cacs.louisiana.edu
Buffalo, NY 14260
Email: {eyanmaz, tonguz}@ece.cmu.edu
Email: qiao@computer.org
Abstract In this paper, a simplified analysis technique for
the Integrated Cellular and Ad hoc Relay (iCAR) systems is
presented. First, a simple two-cell system is analyzed using
a multi-dimensional Markov-chain. The performance metric
employed is the call blocking probability of each cell in the
system. To this end, first a closed-form expression for the call
blocking probability in the two-cell system is provided. Then, it is
shown that these closed-form expressions could be used to analyze
more practical systems. The accuracy of the developed simple
analytical expressions is checked and verified by comparing the
results predicted by these analytical expressions with simulation
results. It is shown that there is an excellent match between
analytical and simulation results.
Indexing Terms: Performance Analysis of iCAR Systems, Dynamic
Load Balancing, Cellular Networks

I. I NTRODUCTION
One of the common problems faced by the wireless service
providers worldwide is coping with congestion or hot spots,
where the amount of traffic generated by the users is more
than the capacity of the service providers infrastructure. This
means that the data channels (DCHs) in the congested cells
are less frequently available than the minimum acceptable
level and, hence, the grade of service (GoS) in those cells
may go down to a level below a prescribed threshold (e.g.,
the call blocking probability in those cells exceeds 2%). This
localized congestion might result in a significant number of
calls being blocked or dropped, even though the overall traffic
load of the system has not reached its maximum threshold.
This degradation in service quality is clearly not acceptable
for the cellular customers.
The hot spot problem can be solved by dynamically balancing the load of the hot cells in cellular networks. Many
channel assignment schemes and dynamic load balancing
strategies have been proposed in the past for efficient use of
the frequency spectrum and to handle the hot spot problem
[1]-[11]. A recent approach to dynamic load balancing is the
Integrated Cellular and Ad hoc Relay (iCAR) system [8].
iCAR employs ad hoc relay stations (ARSs), which operate in
the Industrial, Scientific, and Medical (ISM) band, to balance

traffic loads efficiently and to share channels between cells in

the cellular network via primary and secondary relaying [8].
While performance of iCAR systems was previously reported for special cases using extensive simulations, to date
no comprehensive analytical framework was reported which
can cope with very general cases (multi-tier cellular networks,
arbitrary number of hot spots located at arbitrary locations).
In this paper, a simple analytical tool is provided for
studying the performance of iCAR systems. To this end,
first a two-cell system is studied to formulate the dynamic
load balancing and load sharing capabilities of the iCAR
system. It is then shown that the analytical tool developed
using the two-cell system can also be used for computing
the performance of any practical multi-tier coverage area
with multiple cells. The analytical technique developed in
this paper leads to closed-form performance expressions for
iCAR systems. These closed-form expressions, in addition
to providing a simple and accurate closed-form analytical
tool, also provide valuable physical insight into the impact
of main system parameters (such as coverage area of ARSs,
the number of cellular-band and ISM-band channels, etc.) on
the performance of iCAR systems.
II. OVERVIEW OF THE I CAR S YSTEM
The basic idea of the iCAR System is to place a number of
ARSs throughout the geographical coverage area, so that the
signals between the mobile hosts (MHs) and base transceiver
stations (BTSs) can be relayed [8]-[11]. An ARS is a wireless
communication device, which may have limited mobility under
control of a mobile switching center (MSC), and it can
communicate directly with a MH, a BTS, or another ARS
through air interfaces. It can be assumed that each ARS
has two air interfaces, one for communicating with the BTS
(cellular interface) and the other for communicating with the
MH and other ARSs (relay interface). Similarly, each MH
will need to have two air interfaces. We can assume that the
ARSs operate in 2.4 GHz unlicensed ISM band and, hence,
dont cause any interference to cellular band.

0-7803-7802-4/03/$17.00 2003 IEEE

949

We can describe the types of relaying in the iCAR system

as follows:
Primary Relaying: If a MH cannot be assigned a data
channel in a congested cell, it can be directly relayed to a
neighboring cell via ARSs if the MH is within the ARS
coverage area [8]. For example, assuming that cell 2 in Fig. 1
is congested, a new user X will not be able to find a cellularband channel in cell 2. However, since it is in the coverage area
of ARS2, it can use a channel of cell 1 via primary relaying
through ARS2.
Secondary Relaying: If the MH requesting service is outside
the ARS coverage area of the congested cell, an ongoing call
within the ARS coverage area can be relayed to a neighboring
cell via ARSs freeing up a data channel in the congested
cell to serve the new call [8]. Note that, when secondary
relaying is employed, it is implied that primary relaying is
also employed. For example, in Fig. 1, a new user Y can
not be assigned a channel in cell 2 and it is not covered by
an ARS; therefore, primary relaying will not work. In this
case, checking if there are any ongoing calls within the ARS
coverage area, one realizes that user Z is in the coverage area
of ARS2. User Z is then relayed to cell 1, and user Y can
then use the channel released by user Z in cell 2.
ISM-band
Cellular Band

ARS1

ARS2

Cell 1

Cellular Band

Cell 2

Base Transceiver Station (BTS)

Ad hoc Relay Station (ARS)

Mobile Host (MH)- ongoing call

Cellular Interface

Mobile Host (MH)- new call

Relay Interface

generalized to study any practical multi-tier coverage area with

multiple cells. Note that, the analysis presented in this paper is
based on perfect channel conditions. Also, issues like handoff,
routing, mobility of the ARSs, etc. are not taken into account.
The transceiver design in the ARSs and other issues due to
ISM-band (such as interference, power, etc.) will be addressed
in future studies.
In the analysis, it is assumed that each cell in the system
has M cellular-band channels and K relay channels in the
ISM-band, where M > K. The ARSs are located across
the shared border of the two cells. The ARS coverage area
is normalized with respect to the base station coverage area
and is denoted by p. For simplicity, we assume that the users
are uniformly distributed throughout a cell. We also assume
that call requests arrive according to a Poisson process and call
arrival rates in cells 1 and 2 are 1 and 2 , respectively; and
service times (i.e., call durations) are exponentially distributed
with parameter (i.e., call termination rate in both cells is ).
When employing iCAR, our objective is to increase the
capacity of the congested cell without decreasing the capacity
of the neighboring cells. Therefore, we assume that whenever a
call using a cellular-band channel is terminated in a congested
cell, one of the relay channel users (if such a user exists)
will be switched to the available cellular-band channel and
the corresponding relay channel will be released. This way
unnecessary usage of the resources and, hence, a degradation
in the capacity of the neighboring cell can be avoided.
When secondary relaying is studied, one should keep track
of the active users in the ARS coverage area. Hence, the
states for the secondary relaying are chosen to be of the
form S(i1 , i2 ; j1 , j2 ), where i1 and i2 are the total number
of active users, and j1 and j2 are the number of active users
within the ARS coverage area in cells 1 and 2, respectively.
A closed-form expression for the state probabilities can
be obtained by solving the state-flow equations. The state
probabilities are then given by:
S(i1 , i2 ; j1 , j2 ) =

Outgoing call

= T1i1 T2i2

Fig. 1. Two-cell system layout. Primary relaying (user X) and secondary

relaying (user Y via user Z) is also illustrated.

pj1 +j2 (1 p)(i1 +i2 )(j1 +j2 )

S(0, 0; 0, 0)
j1 ! j2 ! (i1 j1 )! (i2 j2 )!

(1)

where T1 = 1 / and T2 = 2 / are the traffic intensities

of cells 1 and 2, respectively and the normalization factor
denoted by SUM is given as:

III. A NALYSIS OF AN I CAR S YSTEM WITH TWO - CELLS

To evaluate the performance of iCAR systems, one can
model the dynamic load balancing in a multi-tier cellular
network as a multi-dimensional Markov chain. Unfortunately,
as the number of tiers increases, such analysis becomes
intractable. Therefore, in this paper, first a simple two-cell
system, depicted in Fig. 1, is analyzed via a four-dimensional
Markov chain [12]-[13]; it is then shown that the simple
analytical tool developed for the two-cell system can be

SU M =

S(i1 , i2 ; j1 , j2 ) = 1.

(2)

(i1 ,i2 ),(j1 ,j2 )

In Eq. (2), if 0 i1 M , then 0 i2 M , 0 j1 i1 ,

and 0 j2 i2 ; or if M < i1 M + K, then 0 i2
2M i1 , i1 M j1 i1 , and 0 j2 i2 .

950

In Eq. (1), S(0, 0; 0, 0) is the idle state probability and is

given by:
S(0, 0; 0, 0) =
i1
M
+K 2M

M

M


T1i1 T2i2
+
i !i !
i =0 i =0 1 2
1

i1

i1 =M +1 i2 =0 j1 =i1 M

1

T1i1 T2i2 + T1i2 T2i1 pj1 (1 p)i1 j1

j1 !(i1 j1 )!i2 !

(3)

Next, we will find an expression for the call blocking

probabilities of the two cells. The call blocking probability
of cell 1 will be derived. The call blocking probability for cell
2 can be calculated similarly.

and the last term corresponds to the states where secondary

relaying cannot be done, although there are available cellularband channels in cell 2.
Observe that when there is negligible coverage by ARS and
no ISM-band channels, i.e., when K = 0 and p = 0, Eq. (4)
reduces to:
PB =

Assuming that the number of cellular-band channels M is

50 and T2 30 Erlangs, for 0 < p 0.5, after some algebra,
Eq. (4) can be further simplified to:

T1M (1p)M +1
M!

A new call in cell 1 will be blocked with probability (1p),

if:
all cellular-band channels in cell 1 are being used
and there are no ongoing calls within the ARS coverage
area to employ secondary relaying, i.e., the current state is
S(i1 , i2 ; i1 M, j2 ), where i1 M , i2 and j2 .
Using these facts one can show that the call blocking
probability in cell 1, denoted by PB , is given as:
PB = S(0, 0; 0, 0)

M
K
M +K

T2i2
pj1 (1 p)M +Kj1
+
T1M +K

i2 !
j1 !(M + K j1 )!
i =0
M
+K

i2

i2 =M +1 j2 =i2 M
M +K1

i1

T2
T1

T1
T2

i2

T12M pj2 (1 p)i2 j2

+
j2 !(i2 j2 ) !(2M i2 ) !

i!

i=0

T1M +K (1p)M +K
(M +K) !

T1i
i!

M +K

p
1p

i=K
M +K

T1i (1p)i
i!

i=M +1

i 

j=iM

p
1p

i 

M +K
i

j  
i

(6)

The first term in the numerator corresponds to the case

where secondary relaying cannot be done (i.e., the new MH
cannot be primary relayed and there are no ongoing calls
within the ARS coverage area). The second term in the
numerator corresponds to the case where all M + K cellularband and relay channels available to cell 1 are being used. The
denominator is the normalization factor and is approximately
equal to 1/S(0, 0; 0, 0) where S(0, 0; 0, 0) is the idle state
probability. Note that since cell 2 is chosen to have a very
small blocking probability (i.e., 0.02%), this expression does
not depend on the traffic intensity of the second cell, T2 .

i1

T22M pj1 (1 p)i1 j1

+
j1 !(i1 j1 ) !(2M i1 ) !
i1 =M j1 =i1 M

i1 1 i2
M +K1

T i1 pi1 M (1 p)M +1 2M

T2
1
,
(4)
M !(i1 M ) !
i2 !
i =0
i1 =M

T1i pi

Also, from Eq. (6), one can observe that as p is increased,

the first term in the numerator of the call blocking probability
expression approaches zero and second term loses its dependence on p. Therefore, when p is increased beyond a point,
the call blocking probability does not change with p and the
call blocking probability for large ARS coverage, i.e., when
p > 0.5, becomes:

j1 =K

K1

i=0

OR
all 2M cellular-band channels in cell 1 & 2 are being used,
i.e., the current state is S(i1 , i2 ; j1 , j2 ), where i1 + i2 = 2M ,
j1 and j2 .

(5)

which is the well-known Erlang B formula [1].

First, we will determine the blocking states. A new call in

PB
cell 1 will be blocked with probability 1, if:
all cellular and ISM-band channels in cell 1 are being
used, i.e., the current state is S(M + K, i2 ; j1 , j2 ), i2 , j1 ,
and j2 ;

T1M
M!
M T1i
i=0 i !

where S(0, 0; 0, 0) is given by Eq. (3).

The first three terms in Eq. (4) correspond to complete

blocking states, i.e., all available channels in cell 1 and
2 (including the ISM-band relay channels) are being used,

PB

T1M +K
(M +K) !
M +K T1i
i=0
i!

(7)

In this case, the system behaves as if each cell can reach a

total of M + K channels at a time instead of M channels and
a new user can be assigned one of those M + K channels.
If all the cellular-band and relay channels in the hot cell are
being used, then the request will be blocked. Note that, this
is the Erlang B formula for the call blocking probability in a
cell with M + K channels and traffic intensity T1 .

951

IV. R ESULTS AND D ISCUSSIONS

In this section, the accuracy of the developed simple analytical expressions is checked and verified by comparing
the results predicted by these analytical expressions with
simulation results.
First, a simple two-cell system is studied. The traffic intensities of cell 1 and 2 are chosen to be 50 and 30 Erlangs,
respectively. The call blocking probabilities corresponding to
these traffic intensities are 10% and 0.02%, respectively. We
assume that the number of cellular-band channels for each cell
is 50. Fig. 2 shows the call blocking probability in cell 1 for a
fixed number of ISM-band relay channels and for different p
values. When K = 5, the minimum call blocking probability
of cell 1 is around 5%. Further decrease in this probability
can be achieved by increasing the number of relay channels.
When K is increased to 15, the call blocking probability
of cell 1 drops below 2%. Observe that the call blocking
probabilities obtained from the approximate equations are in
excellent agreement with the exact call blocking probabilities
for both cases. When K = 15, for large ARS coverage there
is a slight deviation from the exact values. However, even for
this case the approximate results are in good agreement with
the exact results.

Call blocking probability with secondary relaying

C3,2

C5,2
C2,1

C2,2
C1,1

C6,2
C3,1
C7,2

C0,0

C1,2

C4,1

C0,1
C5,1

C0,2
C11,2

C8,2
C9,2

C10,2

Fig. 3. System layout considered in the analysis and simulation. C0,0 denotes
the center cell, Ci,1 , where i = 0, 1, ..., 5, denotes tier-1 cells, and Cj,2 ,
where j = 0, 1, ..., 11, denotes tier-2 cells, while the circles denote the ARSs
placed on the boundary of each cell.

corresponds to a normalized ARS coverage of 0.06. In the

second case, transmission range of each ARS is 500 m, which
corresponds to a normalized ARS coverage of 0.23. Note that
the normalized ARS coverage is calculated according to the
following formula:

Call blocking probability in cell 1

0.12
Approximate, K = 5
Exact, K = 5
Approximate, K = 15
Exact, K = 15

0.1

C4,2

0.08

=
=

0.06

T otal area covered by the ARS  s in one cell

Area of one cell
r2
r2 3
= 3.63 2
(8)
2
R
1.5 R 3

0.04

where r and R are the transmission range of each ARS and

BTS, respectively.

0.02

The ARSs are located at the shared border of each cell

(shown as circles in Fig. 3). We assume that each cell has 50
cellular-band channels. Since the number of non-overlapping
channels in the 2.4 GHz ISM-band is limited to 3 for IEEE
802.11 standard to meet FCC regulations [14], we assumed
that the number of relay channels that each ARS can reach is
3. Therefore, the center cell can reach K = 18 relay channels.
The traffic intensity of the cells in the same tier is assumed to
be the same. The simulation was performed using GloMoSim
[8], [15].

0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Normalized ARS coverage in a cell (p), M = 50, T1 = 50, and T2 = 30 Erlangs

Fig. 2.
Call blocking probability of cell 1 obtained from exact and
approximate equations versus normalized ARS coverage p.

Next, the validity of the simplified expressions for more

practical systems is shown by studying a 3-tier, i.e., 19-cell,
system with a single hot cell in the center (this is exactly the
same scenario simulated in [8]). The results obtained from the
analytical expressions are compared with simulation results.
The system under investigation is shown in Fig. 3. In the
simulation, the transmission range of each BTS is 2 km. The
system is simulated for two different ARS coverage areas. In
the first case, transmission range of each ARS is 250 m, which

While generating the analytical results we treated the 3-tier,

i.e., 19-cell, system conceptually as a two cell system. The
first cell is assumed to be the center cell, i.e., cell (0,0), which
is assumed to be the hot spot, and second cell is assumed to
be the other 18 cells in the system lumped together. Fig. 4
shows the call blocking probability of cell (0,0) for different
traffic intensities for two different ARS coverage values. The
traffic intensity of cell (0,0) varies between 40 and 51 Erlangs.

952

Call blocking probability with secondary relaying

0.12

0.1

tigate such systems as well.

Erlang B
Analysis, p = 0.06
Simulation, p = 0.06
Analysis, p = 0.23
Simulation, p = 0.23

ACKNOWLEDGMENT
This work was supported in part by the National Science Foundation under the contract ANIR-ITR 0082916 and
Carnegie Mellon University.

0.08

0.06

R EFERENCES
0.04

0.02

0
40

42

44

46

48

50

Traffic Intensity in center cell, M = 50, K = 18

Fig. 4. Call blocking probability of the center cell for the 3-tier system with
secondary relaying versus traffic intensity in the center cell.

The traffic intensities of tier-1 cells are equal to 0.8 Tc0,0 ,

and the traffic intensities of tier-2 cells are equal to 0.64
Tc0,0 , where Tc0,0 is the traffic intensity of the center cell. For
comparison with the conventional system, the call blocking
probability without relaying is also shown in Fig. 4 (see the
dash-dotted curve).
Observe that the simulation and analysis results are in
excellent agreement for both cases. These results suggest that
the call blocking probability expressions obtained for a simple
two-cell system can be used to analyze the performance of
multi-tier systems with multiple cells and hot spots.
V. C ONCLUSIONS
In this paper, a simple analytical tool is developed to study
the performance of iCAR systems. To this end, a closedform expression for the call blocking probability of the hot
cell for a two-cell system is provided. For a wide range of
parameters, this expression provides valuable physical insight
into the impact of main system parameters on the performance
of iCAR systems. It is also shown that one can use the same
analytical tool to study more practical multi-tier systems with
multiple cells. Clearly, the exact analytical expressions for
such cases will be more complicated, but it is shown that the
analytical tool developed in this paper can be used to inves-

[1] T. Rappaport, Wireless Communications: Principles and Practice. Prentice Hall, 1996.
[2] M. Zhang, T.- S. P. Yum, Comparisons of Channel-Assignment Strategies in Cellular Mobile Telephone Systems, IEEE Trans. Vehic. Tech.,
vol. 38, no. 4, pp. 211-215, Nov. 1989.
[3] J. C.- I. Chuang, Performance Issues and Algorithms for Dynamic
Channel Assignment, IEEE J. Sel. Areas Commun., vol. 11, no. 6, pp.
955-963, Aug. 1993.
[4] X. Wu, B. Mukherjee, and S.- H. G. Chan, MACA - An Efficient
Channel Allocation Scheme in Cellular Networks, in Proc. IEEE Global
Telecom. Conf. (GLOBECOM00), vol. 3, pp. 1385-1389, 2000.
[5] T. J. Kahwa and N. D. Georganas, A Hybrid Channel Assignment
Scheme in Large-Scale, Cellular-Structured Mobile Communication
Systems, IEEE Trans. Commun., vol. COM-26, no. 4, pp. 432-438,
Apr. 1978.
[6] B. Eklundh, Channel Utilization and Blocking Probability in a Cellular
Mobile Telephone System with Directed Retry, IEEE Trans. Commun.,
vol. 34, no. 4, pp. 329-337, Apr. 1986.
[7] H. Jiang and S.S. Rappaport, CBWL: A New Channel Assignment
and Sharing Method for Cellular Communication Systems, IEEE Trans.
Vehic. Tech., vol. 43, no.2, pp. 313-322, May 1994.
[8] H. Wu, C. Qiao, S. De, and O. K. Tonguz, Integrated Cellular and
Ad-Hoc Relay Systems: iCAR, IEEE J. Sel. Areas Commun., vol. 19,
no. 10, pp. 2105-2115, Oct. 2001.
[9] S. De, O. K. Tonguz, H. Wu, and C. Qiao, Integrated Cellular and
Ad hoc Relay (iCAR) systems: Pushing the performance limits of
conventional wireless networks, in Proc. 35th Annual Hawaii Intl. Conf.
System Sciences (HICSS02), pp. 3931-3938, Jan. 2002. y 2002.
[10] E. Yanmaz, S. Mishra, O. K. Tonguz, H. Wu, and C. Qiao, Efficient
Dynamic Load Balancing Algorithms using iCAR Systems: A Generalized Framework, in Proc. IEEE Vehic. Tech. Conf. (VTC Fall 2002),
vol. 1, pp. 586-590, 2002.
[11] C. Qiao, H. Wu, and O. K. Tonguz, Load Balancing via Relay in
Next Generation Wireless Systems, in Proc. IEEE Mobile Ad Hoc
Networking & Computing, pp. 149-150, Dec. 2000.
[12] E. Yanmaz, S. Mishra, O. K. Tonguz, H. Wu, and C. Qiao, Impact of
the Number of ISM-band Ad hoc Relay Channels on the Performance of
iCAR Systems, in Proc. IEEE Vehic. Tech. Conf. (VTC Spring 2002),
vol. 3, pp. 1492-1496, May 2002.
[13] O. K. Tonguz and E. Yanmaz, On the Performance of iCAR Systems
with Limited Number of ISM-band Ad hoc Relay Channels, in Proc.
IEEE Global Telecom. Conf. (GLOBECOM02), Taipei, Taiwan, Nov.
2002.
[14] FCC Rules and Regulations. Code of Federal Regulations, Title 47 Part 15. [Online]. Available: http://wireless.fcc.gov/rules.html.
[15] X. Zeng, R. Bagrodia, and M. Gerla, GloMoSim: A library for parallel
simulation of large-scale wireless networks, in Proc. Workshop on
Parallel and Distributed Simulation, pp. 154-161, 1998.

953