Intelligent power allocation strategies in an unlicensed spectrum

We consider power allocation strategies for radios operating in unlicensed bands. Since each radio\´s power allocation is a source of interference for other radios sharing the same spectrum, we seek to develop "intelligent" power allocation strategies that not only greedily optimize performance, but also tend toward socially optimal solutions. Radio interaction is modeled as a two-player repeated game where players develop "reputations" based on previous actions. Bad behavior can result in punishment meted out by other players while socially acceptable behavior could be rewarded. The space of possible radio strategies is explored using genetic algorithms and we find that there are identifiable strategy features shared by all good policies. We conclude with a brief discussion of the possible applicability of our strategies and algorithms, depending on the operating environment and computational complexity at hand