Professional Documents
Culture Documents
Johan Hultell
Introduction
We will study the management of resources in radio networks where the decisions
are delegated to the users. While the main benefits of this approach are lower
system complexity (since the intelligence are moved from the network to users),
increased redundancy and possibly less signaling, the major drawbacks are that
users have to base their decisions on partial/imperfect information and that
they may have incentives to act selfishly and ignore how their actions affect
the overall system performance. These concerns have induced research efforts
focusing on quantifying how severe performance degradation the selfishness of
users are likely to cause1 . Examples range from distributed power control,
multi-hopping, dynamic spectrum allocation, to how a packet channel should
be shared amongst a group of users with conflicting interests. In the first part
of the thesis we will add to this body of literature by studying:
• If distributed control of wireless resources where users maximize their per-
formance without considering how the decisions affect overall performance
can result in similar performance as “traditional” methods for sharing a
common resource.
The set of studies that can be related to this question will be presented in
Chapter 2 – Chapter 3 of the thesis. While the former chapter focuses on the
problem in the context of multiple access protocols Chapter 3 will evaluate the
potential loss in performance when selfish users are responsible for prioritizing
between a set of candidate access points.
The second part of the thesis; Chapter 4 – Chapter 6; targets a multi op-
erator scenario where users can roam freely between all providers and we will
refer to it as competitive network sharing. This mode of operation can
be beneficial for both network providers, who could reduce their investments
in infrastructure, and users due to the multiplicity of access alternatives (more
coverage, more services, lower price). Contrary to the situation for contempo-
rary mobile operators though, this type of open systems where users can roam
between the providers may result in competition between operators also in the
very short term perspective (per session, second,...). The overall aim in the
second part of the thesis is to determine:
1 These will be discussed in detail in Section 1.1.
1
• Whether competitive network sharing can be self-sustained and create
win/win situations where both users and providers benefit?
The remaining part of this document is outlined as follows: We start by re-
viewing the related work in the subject of interests. Thereafter we, in Section
1.2, describe the scope of the thesis along with the contributions and overall
research approach. Chapter 2 presents the preliminary conclusions of the re-
spective thesis chapters as well as a more detailed research approach for the
chapters of the thesis where additional work remains. Chapter 3 outlines the
remaining work and a tentative time plan. To keep this document concise
we have only included the related literature, contributions, and preliminary
conclusions of the thesis chapters. The results for each chapter can be
found in the papers and “draft version” of the thesis available at:
www.s3.kth.se/~johanh/PhDProposal/.
information.
2
users can be on par with the one offered by other distributed utility maximizing
protocols.
another access point the one associated with maximum path gain will consume more resources
(transmission time,transmit power, etc.) and therefore generate higher interference levels. In
uplink transmission the reallocated users will, in addition to consumer more resources, also
be closer to the “overloaded” access point.
5 We may note that studies that do not account for interference overestimate the gain from
3
and user assignment they use Gibbs sampling and as in the other references
load based assignment is shown to improve the performance.
4
“end-to-end” price. Since different links in a route can be managed by different
providers, the ISPs are involved in strategic noncooperative game. For an iso-
lated route the authors show that noncooperative pricing where each ISP tries
to maximize its individual profit not only can result in “unfair” profit alloca-
tions but also discourage network upgrades since a provider’s revenue typically
increases if it manages the bottleneck link in a route.6 As an alternative a
revenue-sharing policy, based on the proportionally fair criterion, is proposed
and this is shown to increase profits for all involved providers. The effects of
competitive demand-responsive pricing in the contexts of dynamic spectrum al-
location is the topic studied in [29]. There spectrum is managed by a third party
broker wherefrom the providers can purchase bandwidth in a wholesale market.
Thus operators, in addition to competing for users the network providers also
competes for bandwidth. It is shown that competition between operators, in
general, transfers welfare from network providers to consumers as well as in-
creases the spectrum utilization. A similar study is presented in [30] wherein
a sealed-bid Knapsack auction for allocating shared spectrum amongst a set of
competing providers is proposed.
(access points) are to be deployed while greedy heuristics are used when “several” facilities
should be deployed.
5
require a model that describes how the market is divided between the competing
providers given a certain deployment of facilities. These range from determin-
istic, distance-based models where each facility serves its Voronoi region to sta-
tistical attraction models that support scenarios with heterogeneous facilities.
While the majority of studies have focused on scenarios where one company
enters an already existing market there are also studies where competitors take
turn in deploying new facilities; se e.g. [39, 40]. We highlight that even with
very simplistic models the problem of determining where to, optimally, place
new facilities (“access points”) becomes mathematically intractable.
6
Paper 2 J. Hultell, “Selfish users sharing a multi-rate ALOHA channel”, to be
submitt to some IEEE conference, 2008.
The work presented in Paper 1 is extended to systems in which multiple
data rates are supported. Both a scenario with link adaptation and a
case with hybrid automatic retransmission requests (“incremental redun-
dancy”) are analyzed.
In second part of the chapter we study conflict free protocols in which users
with files to transfer participate in an auction at which the resources (“trans-
mission time”) is sold. Auctions have been frequently been proposed as a means
for letting users, with heterogeneous requirements, share a common resource.
Two papers have been written on the subject and, similar to Paper 1 and Pa-
per 2, the main question addressed is whether such protocols can offer similar
performance as the traditional ones.
Paper 3 M. Blomgren, and J. Hultell, “Decentralized market based radio re-
source management in multi network environments”, Proc. of Vehicular
Technology Conference (VTC-Spring), May, Dublin, 2007.
We develop a market based framework for price based radio resource man-
agement when users can be served by multiple access points. This frame-
work is used in both Chapter 3 and Chapter 5 of the thesis.
Paper 4 J. Hultell, “Market based resource allocation for delay elastic file
transfers”, conference to be decided, 2008.
Market based scheduling is compared with several centralized schedulers.
In particular we consider the effect of accounting for peak data rate, ser-
vice class, and file size when allocating resources to users.
7
larger resource pool) and improved path gain statistics due to lower carrier
frequency, higher macroscopic diversity order, smaller channel bandwidth,
etc. We show that when prioritizing between access points with
skew throughput data users should connect to the offering high-
est peak data rate.
Paper 6 M. Berg and J. Hultell, “On Selfish Distributed Access Selection Al-
gorithms in IEEE 802.11 Networks”, Proc. of Vehicular Technology Con-
ference (VTC-Fall), Montreol, September, 2006.
In this paper we study the potential gains of load based access selection
in noise limited WLAN networks by means of time-dynamical simula-
tion experiments. We show that load based access selection can improve
user data rates and system efficiency compared to methods that only ac-
count for path gain. Furthermore we show that algorithms according
to which users act selfishly offer comparable performance as the
centralized ones and that all (reasonable) centralized load-based
methods, in fact, offer similar performance.
Paper 7 M. Blomgren, J. Hultell, R. Cai, and T. Cai, “Distributed demand-
aware access selection in wireless multi-cell data networks”, Proc. of
PIMRC, Athens, September, 2007.
We study whether load based access point selection can improve capacity
and user data rates in “interference limited” cellular data networks.
As reference case, we use a single frequency system where the spectrum
is reused in all cells and users connect to the access point with maximum
path gain. We show that (i) the gains from load based access selec-
tion in combination with the introduction of a channel plan is
modest, and that (ii) the performance loss arising from that users
act selfishly and maximize their individual performance is small.
8
it has for users and providers. Both a case of pure competition and one
of pure cooperation are discussed as two extremes. Both scenarios are
shown to result in cost reductions and in the competitive scenario this
translates into price reductions (this will be further discussed in Chapter
5). Examples of emerging business models are identified and discussed.
Paper 10 J. Hultell, K. Johansson, and J. Markendahl, “Business models and
resource management for shared wireless networks”, Proc. of Vehicular
Technology Conference (VTC-Fall), Los Angeles, September, 2004.
Positioning paper that discusses research areas for resource management in
shared wireless networks. Specific requirements that are posed on resource
management in shared radio networks are highlighted.
It should be stressed that the these papers we will extend considerably. As a
considerable part of this chapter have not been published a draft version of the
chapter is available at the website.
9
Paper 12 J. Hultell, M. Blomgren, and J. Zander, “Demand responsive usage
based pricing in wireless access markets”, to be submitted International
Journal on Mobile Network Design and Innovation, Spring, 2008.
We analyze how the individual interests and strategic interactions of users
and mobile operators affect retail price in an oligopolistic and competitive
access market. Focus in the paper is to determine how many providers
users should include and evaluate in their candidate set and what effects
this choice has for providers’ revenues and users’ monetary expenditures.
The study is performed for concave user demand and we show, by example,
that compared to the contemporary situation where users only can connect
to one operator, competitive infrastructure sharing can result in win win
situations where both the users and providers benefit.
Another peculiarity of open system is that the providers may have incentives to
provide users with “false” information in order to attract more customers; see
e.g. Paper 9 which discuss this issue. One example of how this could be realized
in wireless systems is a scenario where providers need to determine how to split
their power between a data and pilot channel (users base their selection of on
the received signal strength associated with the pilot channel). This is studied
in Paper 13.
Paper 13 J. Hultell, “Non cooperative power games in multi provider scenar-
ios”, internal report, 2008.
We study whether operators would have incentives to “advertise” false in-
formation in order to attract more customers in open systems. This work
is basically completed (90%) and the expected conclusion is that as long as
users revisit the same provider, incentives for providing false information
is modest.
10
study the related problems we have adopted a game theoretic mindset. Game
theory is a multidisciplinary branch that combines applied mathematics and mi-
croeconomics and it is suitable for distributed multi agent decision problems. In
the thesis we will often combine analytical treatment with computer based
simulation experiments.
11
Chapter 2
In this chapter we present the preliminary conclusions of the chapters that are
based on already completed material. Focus is, however, devoted to the chapters
where remaining work exists. A more detailed description of the completed
contributions can be found at www.s3.kth.se/~johanh/PhDProposal/ where
the original papers and a draft version of the thesis are available.
12
We furthermore studied the case where the access point supports multiple
data rates as well as retransmissions of unsuccessfully received packets. Com-
pared to single rate systems the efficiency gains by letting users select trans-
mission probability and rate (as compared to path gain oblivious protocols)
becomes even more pronounced. At low transmission costs, the efficiency in sys-
tems where only link adaptation is supported is higher than when also HARQ is
supported. This is an effect of that all users will have a positive expected payoff
when the transmission costs is low and HARQ is supported. In the prolonging
this increases interference levels and therefore lowers system efficiency. As the
transmission costs increases, though, the system efficiency with HARQ exceeds
the one associated with link adaptation. Both modes of operation, however,
offers considerably higher system efficiency than a distributed ALOHA protocol
where all users transmit with the same probability.
In the second part of the chapter we developed a framework for price based
rate selection in contexts where users with delay elastic files competed for wire-
less resources. The price was settled through a proportionally fair divisible
auction. The problem for users, whose preferences was modeled as weighted
sum of the file transfer delay and the associated monetary expenditure, was to
determine how much resources they should purchase. Our results show that
selfish users outperforms the common resource input and output fair protocols
and offer similar performance as preemptive priority based schedulers.
13
2.3 Cooperative and competitive wireless access
(Ch4) – Partly remaining work
The purpose of this chapter is to analyze the rationale, advantages and draw-
backs for sharing infrastructure and wireless resources. In addition to quan-
tifying how much operators are likely to gain by cooperating (e.g. sharing
infrastructure spectrum) the chapter will provide a discussion regarding the
characteristics of the supply and demand for wireless access. Moreover we pro-
pose an alternative method for implementing infrastructure sharing; wherefrom
both operators and users are likely to benefit. This will be further analyzed in
the following two chapters. In the following we will outline the major parts of
this chapter.
14
9 6
10th percentile user throughput 10th percentile user throughput
8 Average user throughput Average user throughput
5
7 10% indoor usage 10% indoor usage
50% indoor usage
Throughput [Mbps]
Throughput [Mbps]
6 50% indoor usage 4
90% indoor usage
5 90% indoor usage
3
4
3 2
2
1
1
0 0
0 10 20 30 40 50 60 0 5 10 15 20 25
2 2
Area throughput [Mbps/km ] Area throughput [Mbps/km ]
Figure 2.1: Achievable data rate as a function of the area throughput and the
percentage of indoor users. In the example the intersite distance is 500 m and
a channel bandwidth of 5 MHz is available to the operators.
networks and Siemens as well as the one Lucent and Alcatel. These mergers are a combined
effect of that the consolidation amongst operators and that new low cost vendors, e.g. Huawei,
has entered the markets.
3 Current examples are Hutchison (Operator 3) and T-mobile in the UK which have agreed
to share mobile mast and access network for the 3G infrastructure. Another current example
is the radio access network (including mast, antennas, sites, power supply, transmission links,
3G base stations, and the radio network controllers) sharing agreements between Orange and
Vodafone. and mobile. During 2008, they have also announced plans for sharing network in
Romania and Spain.
15
Network sharing may be necessary for wireless data access
As we mentioned the value proposition of personal communication services is
strongly connected to the coverage. As these services only require modest data
rates (and therefore also investments) and consumer demand is well documented
the potential cost savings from shared infrastructure have not outweighed the
reduced differentiation possibilities.
With the introduction of wireless data services, though, wireless infrastruc-
ture may become an necessary complement to traditional expansion strategies.
This is because: (i) users willingness to pay is strongly dependent on the con-
tent and not only on converge (as for voice)4 , (ii) most usage will be generated
by stationary indoor users who will not be willing to pay a coverage and mobility
premium, and (iii) providing indoor broadband access with outdoor solutions
may not be economically viable and a wide uptake of specialized indoor solu-
tions may require substantial subsidization (which implies that operators need
to free up capital).
16
1500 1500
1000 1000
500 500
0 0
−500 −500
−1000 −1000
−1500 −1500
−1500 −1000 −500 0 500 1000 1500 −1500 −1000 −500 0 500 1000 1500
Figure 2.2: Illustration of the layout when the cooperating systems are almost
cosited and when they are widely separated.
11 6
10
Widely seperated Widely seperated
Almost cosited Almost cosited
9 5
No cooperation No cooperation
8
Throughput [Mbps]
Throughput [Mbps]
4
7
6
3
5
4
2
3
2 1
1
0 0
0 10 20 30 40 50 60 0 5 10 15 20 25 30
Area throughput [Mbps/km2/operator] Area throughput [Mbps/km2/operator]
(a) Intersite distance 500m (DL) (b) Intersite distance 500m (UL)
Figure 2.3: Data rates as a function of the area throughput. Red curves corre-
sponds to a scenario where the access points of the cooperating operators are
“widely separated”. Blue curves represents a case where the access points is
“almost cosited”, and the black curves a case where there is no cooperation. In
the figures we assume that all users are outdoors, and that there is a standard
deviation of 8dB.
17
Relative number of APs
1,6
1,4
1,2
10
4000
20 2000
30 1000
500
2
Offered traffic [Mbps/km /operator] Data rate requirement [kbps]
Figure 2.4: The access point ratio for a case where operators do not cooperate
and a case where they instead cooperate for different data rate requirements
and traffic volumes. In the specific example focus is on downlink transmission
and the access points of the two operators are “almost cosited”.
18
competing network providers may be economically sustainable. Typically users
obtain much cheaper service and experience higher average data rates. The gains
are a combined effect of increased competition and improved path gain statistics
since users have more access points to chose from. Although providers will earn
less revenue as compared to a oligopolistic pricing regime the difference is limited
(around 10 percent) when one also account for the user behavior. Compared to
the current situation where users only can connect to infrastructure managed
by the provider with whom they subscribe, however, provider revenue can even
be increased and a win/win situation where both providers and users benefit
are realizable. We have also showed that the difference between a oligopolistic
and competitive pricing regime reduce with concavity of user demand. This
corresponds to a scenario where the vast majority of the users have similar
price preferences.
19
present. All operators have an equal market share as well as the same amount
of spectrum. To increase data rates operators now want to complement their
network with micro base stations at some of the finite number of candidate site
locations. In the following we will outline the models used in the chapter.
Other models
We will assume that the macro and micro base stations have their antenna
placed 30 m and 10 m above street level. The height for houses are assumed
to be 25 m and with a street width of 15 m, building separation of 30 m and
mobile height of 1.5 m. This results in that the path gain values at 1 m are 35.2
and 22.2 dB, respectively, while the path gain exponents are 3.8 and 4.7 (these
values correspond the COST Walfisch-Ikegami model).
The peak data rate achieved for users experiencing a SIR equal to Γi is
modeled as
ri = min (W log2 (1 + βΓi ) , rmax ) , (2.1)
where β is an offset factor (“SIR gap”) that describes the difference between
information theoretic and practically implementable results and rmax the max-
imum supportable peak data rate.
To model the average throughput that users experience, we use a Processor
sharing formula according to which users at demand node i, associated with a
20
peak data rate ri and who are connected to a base station with load ρ, obtain
an average throughput of
log(1 − ρ)
si = −ri (1 − ρ) . (2.2)
ρ
traffic Λinit or the number of micro base stations required to support the same area throughput
at the given data rate.
21
2.6 Preliminary conclusions (Ch6)
The main conclusions in the thesis will be:
• Letting users decide their transmission parameters can increase the system
efficiency even when users act selfishly and ignores how their actions affect
the other users in the system as compared to traditional path gain and
service oblivious protocols. This gain comes at the expense of increased
user performance variations.
• Also accounting for the current demand when prioritizing between a set of
candidate access points can increase user data rates and system capacity
if a channel plan is used. All methods for ranking the candidate cells
that account for load offer similar performance. Hence the potential loss
from delegating the responsibilities to terminals is small. Compared to the
option of reusing all spectrum in all cells, the capacity gains are modest.
• A scenario where users freely can roam between networks who compete by
means of prices (and service quality) can result in win-win situations where
operators earn more revenue concurrently as users obtain better service
at lower cost. Sharing the infrastructure results in resource efficiency
gains (improved path loss statistics, trunking efficiency, etc.) while the
competition between providers ensures that part of this efficiency gain is
transferred to users.
22
Chapter 3
This chapter outlines the work that remains in order to complete the thesis.
23
Table 3.1: Preliminary thesis outline
Chapter Pages % completed
Introduction 25
Background 5 50
Related literature 10 90
Scope of thesis and contributions 5 75
Research methodology 5 75
Selfish users sharing a common channel 50 85
Energy reluctant users sharing packet channel 20 95
Energy reluctant users sharing multi-rate packet channel 15 95
Market based scheduling 15 85
Conclusions 1 95
Load based access selection for data traffic 20 95
Load based selection in traffic hotspots 10 95
Load based selection in cellular networks 10 95
Conclusions 1 95
Cooperative and competitive wireless... 25 80
Demand and supply for wireless access 10 80
Cooperative wireless access 10 80
Competitive wireless access 5 80
Discussion 1 80
Price based access selection... 35 80
Price based access selection 20 95
Time dynamic pricing and the effects of demand 10 95
Effects of false information 10 60
Conclusions 1 95
Noncooperative network deployment 25 20
Conclusions 5
Total 180
24
Bibliography
[1] Tansu Alpcan, Tamer Baser, R Srikant, and Eitan Altman, “CDMA uplink
power control as a noncooperative game”, Wireless Networks 8(6):659-670,
November, 2002.
[2] Cem U. Saraydar, Narayan B. Mandayam, and David J. Goodman, “Effi-
cient power control via pricing in wireless data networks”, IEEE Transac-
tions on Communications, 50(2):291-303, February, 2002.
[3] Cem U. Saraydar, Narayan B. Mandayam, and David J. Goodman, “Pric-
ing and power control in a multicell wireless data network, IEEE Journal
in Selected Areas in Communications, 19(10):1883-1892, October, 2001.
[4] Hongbin Ji and Ching-Yao Huang, “Noncooperative Uplink Power Control
in Cellular Radio Systems”, Wireless Netwroks 4(3):233-240, March, 1998.
[5] J. Mitola and G.Q. Maguire Jr, “Cognitive radio: Making software radios
more personal”, IEEE Personal Communications, vol. 6, no. 4, pp. 13-18,
Aug. 1999.
[6] Y.E. Sagaduyu, A. Ephremides, “A Game-theoretic look at throughput and
stability in random access”, In Proceedings of the Military Communications
Conference (MILCOM), pp. 1-7, 2006.
[7] D. Fudenberg and J. Tirole, “Game Theory”, MIT Press, 1991.
[8] M. Cagalj, S. Ganeriwal, I. Aad, and J-P. Hubaux, “On Selfish Behavior in
CSMA/CA Networks”, In Proceedings of the 24th Annual Joint Conference
of the IEEE Computer and Communication Societies (INFOCOM), Vol 4,
pp. 2513-2524, Miami, USA, March 2005.
[9] H. Inaltekin and S. Wicker, “The Analysis of Game Theoretic MAC Proto-
col for Wireless Networks”, In Proceedings of the Annual IEEE Communi-
cation Society on Sensor and Ad Hoc Communications and Networks, vol.
1, pp.296-305, 2006.
[10] S. Rakshit and R.K. Guha, “Fair Bandwidth Sharing in Distributed Sys-
tems: A Game-Theoretic Approach”, IEEE Transactions on Computers,
Vol. 54, No. 11, November, 2005.
[11] H. Inaltekin and S. Wicker, “A One-Shot Random Access Game for Wire-
less Networks”, In Procedings of the International Conference on Wireless
Networks, Communication and Mobile Computing, vol. 2, pp. 940-945,
2005.
25
[12] A.B. MacKenzie and S.B. Wicker, “Selfish users in ALOHA: A Game-
Theoretic Approach”, In Procedings of the 54th Vehicular Technology Con-
ference, Vol. 3, pp. 1354-1357, Atlantic City, USA, October, 2001.
[13] Y. Bejerano, S. J. Han, and E. Li, “Fairness and load balancing in wireless
LANs using association control, In Proceedings of ACM MobiCom, pp.
315-329, September 2004.
[14] O. Yilmaz, A. Furuskär, J. Pettersson, and A. Simonsson, “Access Selection
in WCDMA and WLAN Multi-Access Networks”, In Proc. of VTC 2005-
Spring, Stockholm, June, 2005.
[15] A. Tölli, P. Hakalin, and H. Holma, “Performance Evaluation of Common
Radio Resource Managment”, In IEEE International Conference on Com-
munications (ICC), vol. 5, pp. 3429-3433, May, 2002.
[16] J. Chen, T. Rappaport, and G. de Veciana, “Iterative Water-filling for
Load-balancing in Wireless LAN or Microcellular Networks” In Proc. of
IEEE VTC Spring, Melbourne, May 2006.
[17] B. Kauffmann et al., “Self Organizing of Interfering 802.11 Wireless Access
Networks” INRIA Report no. RR-5649, August 2005.
26
[26] L. Badia, M. Lindström, J. Zander, and M. Zorzi, “Demand and Pricing
Effects on the Radio Resource Allocation of Multimedia Communication
Systems”, In Proceedings of Global Telecommunication Conference, pages
4116–4121, December, 2003.
[27] J. Hou, J. Yang, and S. Papavassiliou,“Integration of Pricing with Call Ad-
mission Control to Meet QoS Requirements in Cellular Networks”, IEEE
Transactions on Parallel and Distributed Systems 13(9):898–910, Septem-
ber 2002.
[28] L. He and J. Walrand, “Pricing and Revenue Sharing Strategies for Internet
Service Providers”, IEEE Journal on Selected Areas in Communications
24(5):942–951, May 2006.
[29] Ö. Ileri, and N.B. Mandayam, “Dynamic Spectrum Access Models: To-
wards an Engineering Perspective in the Spetrum Debate”, IEEE Commu-
nication Magazine, 2007.
[30] S. Sengupta, M. Chatterjee, and S. Ganguly, “An economic framework for
spectrum allocation and service pricing with competitive wireless service
providers”, In Proceedings of IEEE New Frontiers on Dynamic Spectrum
Access Networks, Dublin, May, 2007.
[31] N. Nierbert, A. Schneider, H. Abramowicz, G. Malmgren, J. Sachs, and
U. Horn, “Ambient networks: An architecture for communication networks
beyond 3G, IEEE Personal Communications 11(2):14-22, April, 2002.
[32] K. Tutschku, “Demand-based Radio Network Planning of Cellular Mobile
Communication Systems”, In Proceedings of Seventeenth Annual Confer-
ence of the IEEE Computor and Communications Societies, pp. 1054-1061,
May, 1998.
[33] K. Johansson, “Cost Efficient Deployment Strategies for Heterogeneous
Wireless Networks”, PhD. Thesis, December, 2007.
[34] J. Current and S. Ratick, “Introduction: Facility Location Modeling”, Jour-
nal of the Regional Science Association International, pp. 193-197, 1992.
[35] F. Plastria, “Static competitive facility location: An overview of optimisa-
tion approaches”, European Journal of Operational Research, pp. 461-470,
2001.
[36] R. Church and C. ReVelle, “The maximal covering location problem”, Pa-
pers on Regional Science 32:101-118, 1974.
[37] T. Drezner, Z. Drezner, and S. Salhi, “Solving the multiple competitve
faciliteis location problem”, European Journal of Operational Research 142,
pp.138-151, 2002.
[38] R. Aboolian, O. Berman, and D. Krass, “Competitive facility location and
design problems”, European Journal of Operational Research 182: 40-62,
2007.
[39] H-K. Ahn, S-W. Cheng, O. Cheong, M. Golin, and R. van Oostrum, “Com-
petitive Facility Location: The Voronoi Game”, xxxx, 2003.
27
[40] S. Teramoto, E. Demaine, and R Uehara, “Voronoi game on graphs and its
complexity”, 2006.
[41] A. Barlett and N. Jackson, “Network planning considerations for network
sharing in UMTS, In Proceedings of IEEE International Conferernce on 3G
Mobile Communication Technologies, pp. 17-21, May, 2002.
[42] C. Beckman and G. Smith, “Shared Networks: Making wireless communi-
cation affordable”, IEEE Wireless communications 12(2):78-85,2005.
[43] J.A Village, K.P. Worrall, and D.I. Crawford, “3G Shared Infrastructure”,
In Proceedings of 3G Mobile Communication Technologies, May, 2002.
[44] M. Dillinger, M.K. Pereirasamym, J. Lou, and C. Hartman, “An ap-
proach for inter-operator spectrum sharing for 3G systems and beyond”,
In Proceedings of IEEE Wireless Communication and Network Conference,
March, 2004.
[45] M. Dillinger, M.K. Pereirasamym, J. Lou, and C. Hartman, “Dynamic
inter-operator spectrum sharing for UMTS FDD with displaced cellular
networks”, In Proceedings of IEEE Wireless Communication and Network
Conference, March, 2005.
[46] Oliver Whyman Research, “Who will manage the telecom network?”, 2005.
[47] Ericsson white paper, “Managed services’ impact on the telecom industry”,
March, 2007.
28