The 2nd Workshop of COST Action IC0902
Cognitive Radio and Networking for Cooperative
Coexistence of Heterogeneous Wireless Networks
October 6–7, 2011
Castelldefels, Barcelona-Spain
Contribution to Working Group 2 - Definition of cooperation-based
cognitive algorithms, that take advantage of information exchange
at a local level
Comparison of Selfish versus Altruistic Strategies in Dynamic
Spectrum Allocation for Cognitive Femtocells
Gustavo W.O. da Costa, Luis G.U. Garcia, Andrea F. Cattoni –
Aalborg University
Klaus I. Pedersen, Preben E. Mogensen – Nokia Siemens Networks
Denmark
Gustavo W.O. da Costa/Andrea F. Cattoni – Aalborg University
Dynamic spectrum sharing is of increasing relevance due to the scarce availability of spectrum, and its importance
is further increasing due to the massive growth of traffic demand. In particular, dense femtocell deployment is
expected to take off during the next years in order to meet the demand.
Distributed autonomous spectrum sharing approaches are preferred for the large scale deployment of femtocells.
Yet, designing efficient, stable, fair and scalable distributed algorithms is no easy feat, especially if the cells in a
wireless network tend to act selfishly and independently.
We investigated a practical solution, which consists in letting neighbor Cognitive-enabled femtocells (CFemto) to
establish mutual non-aggression agreements. Under some policies it is possible to devise efficient distributed
channel allocation rules which smoothly reallocate the spectrum.
Game Theory is a powerful toolbox which models the interaction of autonomous agents. In this contribution we
would like to introduce a game theoretic model for a dynamic spectrum sharing framework recently proposed for
femtocells [1].
Using Game theory and, particularly, elements of Network Formation Games, we modeled the establishment of
such agreements as a dynamic game. In addition to that, several strategies for this game were proposed ranging
from purely egoistic to selfless.
Our analysis includes cases where femtocells compete for spectrum as well as cooperative cases towards a
common goal.
In particular, we considered equal rights dynamic spectrum sharing among closed-subscriber group (CSG)
CFemtos. We investigate only the intra-tier interference avoidance [2] and the macrocells are assumed to operate
in a separate band.
The CFemto Access Point (FAP) coordinates the spectrum allocation with the served User Equipments (UEs).
Hence, the spectrum decisions are done on a cell basis. Each FAP is then considered a player of the allocation
game.
The utility of the players is affected not only by the mutual agreements between pair of players, but also by the
way the channels are re-distributed when new players join the game. Ultimately, these two elements will define
the component carrier allocation of each femtocell and the actual SINR in each channel.
The reader interested in the details of the channel allocation algorithm may refer to [1].
In, [1] the spectrum allocation policy states that a new entrant can request coordination at most to two active
players. The analysis in [1] shows that coordinating with two other players is enough to provide considerable
gains, while avoiding spectrum fragmentation and reducing the need for signaling.
In dynamic games, strategies are essentially a contingent plan of how to play the game on each possible
information set. Rational players are typically assumed to select their strategies on a purely selfish manner. Based
on examples from society [3] and nature [4] we are interested in challenging the selfishness assumption and
identify what are the desirable strategies for dynamic spectrum sharing in femtocells.
Action IC0902 – WG2-2011-Doc_YY
1
In this contribution we consider four different ways of selecting strategies:
 Selfish: All players select their strategies according to the canonical GT assumptions, optimizing only
their own instantaneous throughput.
 Selfless new entrant: The new entrant intends to protect the existing players and it is completely selfless.
Other players still play selfishly.
 Max-min: a pair of players will choose to cooperate if this is of benefit of the player with lowest incoming
BIM.
 Max-sum: A pair of players will coordinate transmissions if this decision increases their sum capacity
compared to uncoordinated transmissions.
The performance was evaluated through semi-static system level simulations. We derive our results from a Monte
Carlo performance evaluation and thousands of snapshots have been simulated to ensure statistical reliability.
The scenario consists of a single 5x5 grid of houses assuming CSG femtocells. Each house contains 4 rooms
where both FAPs and UEs are randomly located. Yet, there is at most one femtocell per house. The indoor
propagation is modeled according to the WINNER A1 indoor home scenario [5]. Furthermore, if the UE and the
FAP are located in the same room, we assume line-of-sight propagation and non-line-of-sight otherwise. The
transmit power of femtocells was set to 24dBm. Look-up tables map the SINR to corresponding throughput values
according to a modified Shannon's formula from [6]. The system bandwidth is divided into 6 channels of 15 MHz
each. More details and results are available in [7].
The system level simulation results show that each femtocell should strive for a balance between selfishness and
altruism. It is possible to use this balance in favor of the overall network throughput or in favor of achieving a
minimum performance for each cell. In conclusion, theoretical assumptions should guide the design of practical
solutions, but they should not be a limiting factor.
References
[1] L. Garcia, G. Costa, A. Cattoni, K. Pedersen, and P. Mogensen, “Self organizing coalitions for conflict
evaluation and resolution in femtocells,” in GLOBECOM 2010, 2010 IEEE Global Telecommunications
Conference, 2010, pp. 1 –6.
[2] V. Chandrasekhar, J. Andrews, and A. Gatherer, “Femtocell networks: a survey,” Communications Magazine,
IEEE, vol. 46, no. 9, pp. 59–67, 2008.
[3] R. M. Nelissen, D. S. van Someren, and M. Zeelenberg, “Take it or leave it for something better? Responses to
fair offers in ultimatum bargaining,” Journal of Experimental Social Psychology, vol. 45, no. 6, pp. 1227 –
1231, 2009.
[4] C. Stephens, “Modelling reciprocal altruism,” The British Journal for the Philosophy of Science, vol. 47, no. 4,
pp. 533–551, 1996.
[5] P. et al.. Kysti, “IST-WINNER D1.1.2 , ”WINNER II Channel Models”, ver 1.1,” Winner II, Tech. Rep., 2007.
[Online]. Available: https://www.ist-winner.org/WINNER2-Deliverables/D1.1.2v1.1.pdf
[6] P. E. Mogensen et al., “LTE Capacity Compared to the Shannon Bound,” in VTC Spring, April 2007.
[7] Gustavo W. O. Costa et al., Dynamic Spectrum Sharing in Femtocells: a Comparison of Selfish versus
Altruistic Strategies, in VTC Fall, September 2011.
Action IC0902 – WG2-2011-Doc_YY
2