ANSWER: Adaptive Network Selection in WLAN/UMTS EnviRonment

Abstract

The next generation wireless network is aimed to provide users with anywhere, anytime, and seamless service. With the increasing
demand for the next generation network, many research works have focused on the efficient way to integrate different types
of heterogeneous wireless networks such as cellular systems and wireless LAN. Therefore, the network selection technique plays
a critical role in ensuring quality of service for the next generation network. In this paper, we propose an adaptive network
selection (ANSWER) scheme, which is able to make the better decision about when to switch and choice on which access network.
That is, we want to provide the always-best-connected service as much as possible for the users. Specifically, to achieve
the above goals, the available bandwidths of all possible networks, the location of user’s device and its moving direction
are taken into consideration in the ANSWER approach. We evaluate the performance of the ANSWER scheme through extensive simulations
and the results agree with our goals.

Full text

M. Denko et al. (Eds.): EUC Workshops 2007, LNCS 4809, pp. 413–424, 2007.
© IFIP International Federation for Information Processing 2007
ANSWER: Adaptive Network Selection in
WLAN/UMTS EnviRonment*
Chih-Cheng Hsu1, Ming-Hung Chen1, Cheng-Fu Chou1,
Wei-Chieh Chi1, and Chung-Yi Lai2
1 Dept. of Computer Science and Information Engineering,
National Taiwan University, Taipei, Taiwan, R.O.C.
{Kenneth, mhchen, ccf, hypec}@cmlab.csie.ntu.edu.tw
2 Institute for Information Industry, Taiwan, R.O.C.
laici@iii.org.tw
Abstract. The next generation wireless network is aimed to provide users with
anywhere, anytime, and seamless service. With the increasing demand for the
next generation network, many research works have focused on the efficient
way to integrate different types of heterogeneous wireless networks such as
cellular systems and wireless LAN. Therefore, the network selection technique
plays a critical role in ensuring quality of service for the next generation
network. In this paper, we propose an adaptive network selection (ANSWER)
scheme, which is able to make the better decision about when to switch and
choice on which access network. That is, we want to provide the always-best-
connected service as much as possible for the users. Specifically, to achieve the
above goals, the available bandwidths of all possible networks, the location of
user’s device and its moving direction are taken into consideration in the
ANSWER approach. We evaluate the performance of the ANSWER scheme
through extensive simulations and the results agree with our goals.
Keywords: seamless, heterogeneous wireless network, available bandwidth,
network selection.
1 Introduction
With the increasing demand for high data rate multimedia services, commercial third-
generation (3G) cellular network and handsets are gradually rolling into the market.
Compared to cellular networks, WLAN (IEEE 802.11b offers a data rate up to
11Mb/s) is able to offer higher transmission bandwidth at a lower cost but cover
smaller geographic areas. As a result, the WLAN is regarded as a proper candidate for
high data rate services at certain hotspot areas with low user’s mobility.
* This work was partially supported by the National Science Council and the Ministry of
Education of R.O.C. under the contract No. NSC95-2221-E-002-103-MY2 and NSC95-2622-
E-002-018.

414 C.-C. Hsu et al.
To provide anywhere, anytime, and seamless service, the next generation wireless
network is expected to be a heterogeneous network, which can efficiently integrate
several different characteristic access networks. Mobile users may move among these
heterogeneous networks by seamlessly switching between different serving stations.
How to take advantage of the wide coverage support of cellular network and the high
data rates of WLANs is a challenge.
In cellular network, resource allocation is implemented by properly scheduling
access to wireless channel to provide QoS guarantee. However, there is no user QoS
guarantee in the current IEEE 802.11 WLAN standard. The latest IEEE 802.11e
standard only enhances relative QoS. Besides, different traffic types usually require
different QoS deliveries. Real-time services such as voice and video are sensitive to
end-to-end delay, while the main concern of delay-tolerant data service is throughput.
WLANs are more efficient than cellular networks in serving bursty data traffic, while
it is quite difficult for WLANs to meet the strict delay requirements. Hence, the
differences of QoS support in these two networks need to be considered for resource
management.
While the moving speed of a mobile device is higher, more handoffs may occur
during the lifetime of a call. The handoff procedure would cause extra delay, e.g.,
packet losses or even connection interruption. Moreover, handoff traffic should be
separated from new traffic in terms of call admission. Thus, network selection
algorithms and admission control for different requests are investigated.
In this work, we focus on finding the better access network between IEEE 802.11
WLAN and UMTS (Universal Mobile Telecommunications System) for the user
device with two types of interfaces. Although the capacity of WLAN is larger than
that of UMTS, the current available bandwidth of WLAN may decrease to be lower
than the bandwidth guaranteed in UMTS when a lot of users stay in WLAN. In our
proposed approach, we estimate the current available bandwidth of WLAN and the
user dwell time in WLAN; then we make a handoff decision based on the information
gathered. Since the coverage of UMTS network is much wider than that of WLAN,
we assume that UMTS service always exists. Our contributions are estimating the
network condition, predicting user's moving behavior and deciding if it is beneficial to
make vertical handoff.
The remainder of the work is organized as follows. In section 2, we discuss related
work about the UMTS-WLAN internetworking. In section 3, we describe the overall
network selection algorithm including the basic idea, assumption, and detail
procedures. In section 4, we evaluate the performance of proposed approach. Finally,
we conclude the paper and give future works in section 5.
2 Related Work
In traditional selection methods such as [1], only radio signal strength (RSS) threshold
and hysteresis values are considered. However, they do not take the current condition
or user's preference into account. When more handoff decision factors are considered,
two-dimension cost functions, such as [2], are proposed. In one dimension, the
function reflects the types of service (ToS) requested by the user, and another
dimension represents the cost to network according to specific parameters. The paper

ANSWER: Adaptive Network Selection in WLAN/UMTS EnviRonment 415
in [3] separates cost function factors into different categories: QoS factors, weighting
factors, and network priority factors. QoS factors are defined based on user-specific
requirements. The weight factors stand for the importance of the particular
requirements with respect to the user. The network priority factors present the abilities
of the networks to meet the requirements.
In [4], analytic hierarchy process (AHP) and the grey relational analysis (GRA) are
integrated into the network selection algorithms. AHP is used to derive the weights of
the parameters based on user's preference and service application. GRA takes charge
of ranking the network alternatives. Some research works in [5] and [6] use dwell-
timer to alleviate frequent handoffs and improve performance in transition area. They
also investigate the effect of dwell-timer as the throughput ratio of WLAN and UMTS
is changing. In [7], location information is shown to be beneficial to the accuracy of
handoff decision in multi-service networks.
[8] [9] show that although the WLAN QoS capabilities have been extended with
the introduction of IEEE 802.11e, the WLAN is still unable to support all QoS
features provided by UMTS. Without affecting the service provided to existing
WLAN data users, there is a limited number of UMTS roamers to be accepted in the
WLAN depending on the bandwidth reservation and QoS requirements.
[10] presents a framework for a service provider to perform resource management
in heterogeneous wireless networks. The proposed architecture allows the service
provider to support real-time resource management functions based on Service Level
Agreement (SLA) and seamless service handoff.
The experimental measurements of VoIP on 3G-WLAN internetworking system in
[11] show that in addition to the VoIP connections, the performance of all clients in
WLAN can degrade significantly due to the unfairness, undistinguished real-time and
non real-time traffic of the packet queue in the AP and the inherent property of IEEE
802.11. In [12], they focus on the wireless multimedia distribution applications and
recognize that service continuity is an important quality requirement. It designs and
implements this class of applications on top of a session layer providing download
continuity support when user changes location, network or terminal.
3 ANSWER Framework
In most of previous research works, the bandwidth in the WLAN is assumed to be
greater than that in UMTS. However, for wireless networks, some conditions such as
medium contention, channel fading, and interference, influence the available
bandwidth. Furthermore, at the boundary of the WLAN cell, the received signal
strength (RSS) varies around some thresholds. Such situation might cause frequently
sequential vertical handoffs. This is not desired because the handoff procedure always
demands extra time and no data traffic can be carried to or from mobile users during
the procedure.
Therefore, instead of always switching to WLAN, evaluating the benefit of making
vertical handoff is the main concern. The basic concept of our approach is: estimating
the current available bandwidth of WLAN and predicting moving direction of mobile
users to avoid unnecessary handoff.

416 C.-C. Hsu et al.
Assumptions
Mobile IP allows users in an environment with two wireless networks.
Mobile host is aware of its current position, speed, and direction by GPS.
Coordinates of AP can be broadcast.
UMTS network is always achievable and guarantee a certain amount of
bandwidth.
Table 1. The definition of notations used in this work
Variable Description
BWWLAN Available bandwidth in WLAN
BWUMTS Available bandwidth in UMTS
Pr Receiving power of access point
RXTresh Receiving threshold in WLAN
Vertical handoff delay
T Sojourn time of the mobile host in WLAN cell
υ
Velocity of mobile host
R Transmission range of WLAN
Table 1 gives the definition of notations used in this work, BWWLAN is assessed by
using the approach described in section 3.1. BWUMTS is a constant value since we
assume that certain bandwidth is reserved for the mobile host. Pr is measured from the
packets sent from the base station. RXTresh is the minimal necessary power level for
a packet to be successfully received. comes from the vertical handoff procedure as
the mobile host switches between different access networks. In section 3.2, we will
introduce how we measure the value of T.
υ
is the current speed of the mobile host.
3.1 Available Bandwidth Estimation
The bandwidth estimation approach in [13] [14] is adopted in this article. Figure 1
shows the packet transmission sequence in IEEE 802.11. They measure the
throughput TP of transmitting a packet as TP = S / (tts), where S is the size of
the packet, ts is the time-stamp that the packet is ready at the MAC layer, and tr is the
time-stamp that an ACK has been received. The time interval trts includes the
channel busy and contention time. It is clear that the measured throughput of a packet
depends on the size of the packet. A larger packet has higher measured throughput
because it sends more data once it grabs the channel. To make the throughput
Fig. 1. IEEE 802.11 unicast packet transmission sequence

ANSWER: Adaptive Network Selection in WLAN/UMTS EnviRonment 417
measurement independent of the packet size, they normalize the throughput of a
packet to a pre-defined packet size.
In Figure 1, Td = S / BWch is the actual time for the channel to transmit the data
packet, where BWch is the channel's bit-rate. Here they assume channel's bit-rate is a
pre-defined physical layer parameter. The transmission times of any two packets
should differ only in their times to transmit the DATA packets. Therefore, the
following equation holds:
12
11 2 2
() ( )
rs r s
ch ch
SS
tt tt
BW BW
−− = − − (1)
122
11
2
()
rs
ch ch
SSS
tt BW TP BW
⇒−− = − (2)
Where S1 is the actual data packet size and S2 is a pre-defined standard packet size.
We can use Equation (2) to calculate normalized throughput TP2 for the standard size
packet and use the normalized throughput to represent the available bandwidth of
wireless link.
3.2 Sojourn Time Estimation
We predict MH's sojourn time in WLAN by its current speed and its distance to the
edge of WLAN cell. In the hotspot, MH simply moves with some velocity from one
waypoint to another. As illustrated in Figure 2, the solid line represents the actual
moving path of the MH, and the dotted line stands for the path we predict the MH will
move along. Since we know the moving direction of MH and the coordinates of both
MH and AP, two vectors, a
r and b
r
, are obtained. By the definition of inner product:
cosab ab
θ
⋅=
rr rr , (3)
where is the angle between a
r and b
r
, cos is derived. Then based on the cosine
rule:
2
22
2cosRbL Lb
θ
=+−
r
r (4)
The distance to the edge of WLAN cell, namely L, is obtained by equation (4).
Therefore, the dwell time in WLAN is T = L /
υ
, and even if MH is not in the AP
coverage, T represents the time to meet the boundary of WLAN.
3.3 Network Selection Algorithm
With the information gathered in 3.1 and 3.2, we describe the network selection
procedure in Algorithm 1. Normally, the procedure should be executed for each
probing interval t. However, in order to reduce overhead caused by frequently
probing, we introduce a sleep time t into the algorithm (Step 5 and 21). Step 1 to 5
gather the necessary information and stay in UMTS if no WLAN service is available.
Once MH enters the AP coverage, the handoff decision is made based on the value
cost.

418 C.-C. Hsu et al.
Fig. 2. Moving behavior of a mobile host
The idea of cost is: Is making handoff profitable for the throughput in next t
seconds? If MH is currently connected to UMTS and continues using UMTS service,
no handoff is needed when MH leaves the WLAN. On the contrary, if MH switches to
WLAN, two handoffs are made when MH switches back to UMTS. Similarly, if MH
is currently connected to WLAN, only one handoff is made throughout the procedure
no matter MH switches to UMTS or not. Thus, different situations result in different
objective functions (Step 11 and 13). Note that we add α in the objective function to
reduce oscillation. We prefer to stay in UMTS network when the bandwidth of
WLAN approximates that of UMTS because WLAN network no longer has high
bandwidth, which is its main advantage over UMTS network.
4 Performance Evaluation
We will show and discuss the performance of our approach compared with the
WLAN-first and UMTS-first approach. Section 4.1 describes our simulation
environment and performance metrics.

ANSWER: Adaptive Network Selection in WLAN/UMTS EnviRonment 419
4.1 Network Environment and Performance Metrics
To conduct our experiment, we use the NS-2 (Network Simulator 2) with the UMTS
extension from [15]. We set up our WLAN-UMTS networks according to the
parameters exposed in Table 2. Total simulation time is 500 seconds, and as to the
data traffic destined to MH, performances of CBR/UDP traffic are assessed
respectively. Two metrics are used for our performance evaluation:
z Goodput: (total bits received - retransmitted bits) / measurement interval
z Number of handoffs: The quantity of handoffs MH made in its lifetime.
Table 2. Parameters used in performance evaluation
Parameter Value
Topology 1000m×1000m
WLAN Capacity 11 Mbps
UMTS Capacity 384 Kbps
RXThresh 3.6526e-10
Vertical handoff delay 0.1 sec
Number of Poisson background traffic
in WLAN 10 to 60
Packet size of Poisson traffic 1000 byte
Average data rate of a Poisson traffic 100 Kbps
Probing packet size 64 byte
4.2 Mobility Model
Three most popular mobility models are used in our simulation, and the brief
descriptions are the following:
 Random Waypoint Model: the Random Waypoint model is the most
commonly used mobility model in research community. In the current NS-2,
the implementation of this mobility model is as follows: at every instant, a
node randomly chooses a destination and moves towards it with a velocity
chosen uniformly randomly from [0,Vmax], where Vmax is the maximum
allowable velocity for every mobile node. When reaching the destination, the
node stops for duration. After this pause time, it again chooses a random
destination and repeats the whole process again until the simulation ends.
 Freeway Mobility Model: this model emulates the motion behavior of mobile
nodes on a freeway. Maps are used in this model. There are several freeways
on the map and each freeway has lanes in both directions. The differences
between Random Waypoint and Freeway are the following: (1) Each mobile
node is restricted to its lane. (2) The velocity of mobile node is dependent on
its previous velocity. (3) If two mobile nodes on the same freeway lane are
within the safety distance, the velocity of the following node cannot exceed
the velocity of preceding node. It also imposes strict geographic restrictions on
the node movement by not allowing a node to change its lane.

420 C.-C. Hsu et al.
 Manhattan Mobility Model: Manhattan model emulates the movement
pattern of mobile nodes on streets defined by maps. It can be useful in
modeling movement in an urban area. The map of the Manhattan model is
composed of a number of horizontal and vertical streets. The mobile node is
allowed to move along the grid of horizontal and vertical streets on the map.
At an intersection of a horizontal and a vertical street, the mobile node can
turn left, right or go straight with certain probability. Except the above
difference, the node relationships involved in the model are very similar to the
Freeway model. However, it differs from the Freeway model in giving a node
some freedom to change its direction.
The maximum speed and pause time in random waypoint model are 10 m/s and 10
seconds respectively in our simulation. The speed in Freeway mobility model ranges
from 10 to 60 m/s, and the maximum speed in Manhattan model is 10 m/s.
4.3 Goodput Comparison
From Figure 3 (a) to (c), we can see that our approach ANSWER greatly outperforms
WLAN-first and UMTS-first method in CBR traffic, no matter network load in
WLAN is heavy or not. Our approach will mostly stay in UMTS network when
the WLAN is too crowded (number of background traffic is 30 to 50) and efficiently
utilize the high bandwidth of WLAN when network load of WLAN is light (number
0
50
100
150
200
250
123456789
Number of background traffic
Goodput (Kbps)
WLAN first
ANSWER
UMTS first
0
50
100
150
200
250
123456
Number of background traffic
Goodput (Kbps)
WLAN first
ANSWER
UMTS first
(a) Random Waypoint model (Number of
background traffic: 1 ~ 9)
(b) Random Waypoint model (Number of
background traffic: 10 ~ 50)
Number of background traffic
Goodput (Kbps)
WLAN first
ANSWER
UMTS first
0
50
100
150
200
250
300
350
1234 5
Number of background traffic
Goodput (Kbps)
WLAN first
ANSWER
UMTS first
(c) Manhattan model (Number of
background traffic: 10 ~ 50)
(d) Freeway model (Number of background
traffic: 10 ~ 50)
Fig. 3. CBR goodput comparison in different mobility model

ANSWER: Adaptive Network Selection in WLAN/UMTS EnviRonment 421
of background traffic is 10 to 30). In Figure 3(d), the performance of freeway
mobility model is not good as that in the other two mobility model when the
background traffic in WLAN increases, because MH rushes in and out as its moving
speed is fast.
4.4 Number of Vertical Handoffs
Vertical handoff causes call intermission, so in this section we focus on how many
handoffs MH makes throughout the whole course. Figure 4 show that number of
handoff in our approach is much less than that in WLAN-first method. Our selection
algorithm efficiently reduces many unnecessary handoffs by predicting MH's sojourn
time.
0
1
2
3
4
5
6
7
8
1234567891020304050
Number of background traffic
Number of handoff (lg)
WLAN first
ANSWER
0
1
2
3
4
5
6
7
10 20 30 40 50
Number of background traffic
Number of ha ndoff (lg)
WLAN first
ANSWER
(a) Number of handoff in Random
Waypoint model
(b) Number of handoff in Manhattan model
0
1
2
3
4
5
6
7
10 20 30 40 50
Number of background traffic
Number of handoff (lg)
WLAN first
ANSWER
(c) Number of handoff in Freeway model
Fig. 4. Number of handoff in different mobility model
4.5 Accuracy of Available Bandwidth
To measure the accuracy of the estimated available bandwidth, we first assume the
total usable bandwidth in WLAN to be the maximum WLAN utilization achieved in
our simulations, which is about 3Mbps. Thus, the bandwidth unused by the Poisson
background traffic is the approximation of actual available bandwidth. The comparison
of the approximation and the estimated value by the approach in section 3.1 is
in Figure 5.
When the network load is light, the bandwidth estimated is about 1700 Kbps at
most. The reason is that tts in Figure 5(a) has a minimum value since the

422 C.-C. Hsu et al.
0
500
1000
1500
2000
2500
3000
3500
1234567891020
Number of back ground traffic
Kbps
Max WLAN
utilization for all
cases
Poisson goodput
BW unused
AVBW
(a) Comparison when Number of background traffic: 1 to 20
0
500
1000
1500
2000
2500
3000
3500
25 26 27 28 29 30 31 32 33 34 35 40 50
Number of background traffic
Kbps
Max WLAN
utilization for all
cases
Poisson goodput
BW unused
AVBW
(b) Comparison when Number of background traffic: 21 to 50
Fig. 5. Comparison between estimated and actual bandwidth
contention period in IEEE 802.11 can not be reduced unlimitedly. When the load
increases, the two curves become closer to each other. Although the estimated value is
not highly precise, the two curves go in the same trend.
4.6 Dynamic Sleep Time
We adjust the sleep time t in our algorithm based on speed and available bandwidth
on the instant, as equation (5) shows.
1
WLAN
UMTS
BW
tcBW
υ
=× × (5)
CBR traffic destined to MH in Random Waypoint model is simulated and the
number of background traffic is set to 30. We can see that there is a tradeoff between
goodput/jitter and number of handoff/probe in Figure 6. As c increases, MH would
miss some opportunities to switch to the better network in sleep time but it would
send fewer probing packets.

ANSWER: Adaptive Network Selection in WLAN/UMTS EnviRonment 423
0
500
1000
1500
2000
2500
3000
3500
4000
4500
0
20
40
60
80
100
300
500
700
900
Value of c
Number of probing
ANSWER
0
5
10
15
20
25
0
20
40
60
80
100
300
500
700
900
Value of c
Number of probing
ANSWER
(a) Number of probe (b) Number of handoff
0
50
100
150
200
250
0
20
40
60
80
100
300
500
700
900
Value of c
Number of probing
ANSWER
UMTS firs t
0
1
2
3
4
5
6
7
0
20
40
60
80
100
300
500
700
900
Value of c
Number of probing
ANSWER
(c) Goodput (d) Average jitter
Fig. 6. Simulation results for different value of c
5 Conclusion and Future Works
In this work, we have proposed an approach dealing with network selection from
WLAN and UMTS network. The approach estimates current available bandwidth and
sojourn time in WLAN to choose the better network. The simulation in this paper
shows that our algorithm achieves better performance and has fewer handoffs than
UMTS/WLAN-first methods do in different mobility models. In addition, we
investigate the accuracy of available bandwidth estimation, and the tradeoff between
goodput and number of handoff.
It is noted that only two networks, WLAN and UMTS, are considered in our
scheme. More types of network, such as WiMAX and WCDMA, will be integrated in
our scheme. Moreover, we would like to adjust our approach for different traffic with
different QoS.
References
1. Tripathi, N.D., Reed, J.H., Van Landingham, H.F.: Adaptive handoff algorithms for
cellular overlay systems using fuzzy logic. In: Vehicular Technology Conference, 49th,
vol. 2, pp. 1413–1418. IEEE, Los Alamitos (1999)
2. Park, H., Yoon, S.H., Kim, T.H., Park, J.S., Do, M.S., Lee, J.Y.: Vertical Handoff
Procedure and Algorithm between IEEE 802.11 WLAN and CDMA Cellular Network. In:
Mobile Commun. 7th CDMA Intl. Conf.

424 C.-C. Hsu et al.
3. Roveri, A., Chiasserini, C.F., Femminella, M., Melodia, T., Morabito, G., Rossi, M.,
Tinnirello, I.: The RAMON Module: Architecture Framework and Performance Results.
In: Proceedings of the Second International Workshop on Quality of Service in
Multiservice IP Networks (2003)
4. Song, Q., Jamalipour, A.: Network selection in an integrated wireless LAN and UMTS
environment using mathematical modeling and computing techniques. Wireless
Communications 12(3), 42–48 (2005)
5. Ylianttila, M., Pande, M., Makela, J., Mahonen, P.: Optimization scheme for mobile users
performing vertical handoffsbetween IEEE 802.11 and GPRS/EDGE networks. In:
GLOBECOM 2001. Global Telecommunications Conference, vol. 6, IEEE, Los Alamitos
(2001)
6. Ylianttila, M., MÄakelÄa, J., Pahlavan, K.: Analysis of handoff in a location-aware
vertical multi-access network. Computer Networks (2005)
7. Makela, J., Ylianttila, M., Pahlavan, K.: Handoff decision in multi-service networks.
Personal, Indoor and Mobile Radio Communications. In: PIMRC 2000. The 11th IEEE
International Symposium (2000)
8. Salkintzis, A., Skyrianoglou, D., Passas, N.: Seamless multimedia QoS across UMTS and
WLANs. In: Vehicular Technology Conference, VTC 2005-Spring. 2005, 61th, vol. 4,
IEEE, Los Alamitos (2005)
9. Salkintzis, A.K., Dimitriadis, G., Skyrianoglou, D., Passas, N., Pavlidou, N.: Seamless
continuity of real-time video across umts and wlan networks: challenges and performance
evaluation. In: Wireless Communications, vol. 12(3), pp. 8–18. IEEE, Los Alamitos
(2005)
10. Yang, X., Bigham, J., Cuthbert, L.: Resource management for service providers in
heterogeneous wireless networks. In: Wireless Communications and Networking
Conference, vol. 3, IEEE, Los Alamitos (2005)
11. Rajavelsamy, R., Jeedigunta, V., Holur, B., Choudhary, M., Song, O.: Performance
evaluation of VoIP over 3G-WLAN interworking system. In: Wireless Communications
and Networking Conference, vol. 4, IEEE, Los Alamitos (2005)
12. Ghini, V., Salomoni, P., Pau, G.: Always-best-served music distribution for nomadic users
over heterogeneous networks. Communications Magazine 43(5), 69–74 (2005)
13. Shah, S.H.H., Chen, K.H., Nahrstedt, K.H.: Dynamic Bandwidth Management in Single-
Hop Ad Hoc Wireless Networks. Mobile Networks and Applications 10(1), 199–217
(2005)
14. Shah, S.H., Chen, K., Nahrstedt, K.: Available Bandwidth Estimation in IEEE 802.11-
based Wireless Networks. In: Proceedings of 1st ISMA/CAIDA Workshop on Bandwidth
Estimation (BEst)
15. EURANE. Enhanced UMTS Radio Access Network Extensions for NS-2,
http://www.ti-wmc.nl/eurane/