Dynamic channel sharing strategies through game-theoretic reinforcement learning
View/ Open
Date
2014-03Author
Ayienga, Eric M
Opiyo, Elisha T
William, Okello-Odongo
Language
enMetadata
Show full item recordAbstract
—Random access protocols are used by multiple nodes in
wireless networks to effectively share a wireless channel for data
transmission. While competing for the channel, the nodes seek to
achieve an individual or group objective. Game theory, can thus
be applied to analyze and model individual or group behavior of
nodes in random access networks. It can also be used as an
‘engineering’ application and subsequently re–engineer the
system. In this paper, the current CSMA/CA mechanism was
analyzed using game theory. Based on the analysis, the strategy
space available to individual nodes was increased so that the
optimal strategies for different situations learnt using
reinforcement learning. From the analysis it was determined that
the Nash equilibrium was not Pareto optimal. Simulation
experiments yielded better results for the modified algorithm
pointing to moving the Nash equilibrium towards being fair and
Pareto optimal.
Publisher
University of Nairobi