The University of British Columbia

This paper addresses the secondary user rate adaptation problem in cognitive radio networks. By modeling primary user activities and secondary user block fading channels as finite state Markov chains, the transmission rate adaptation problem of each secondary user is formulated as a general-sum dynamic Markovian game with a delay constraint. Assumptions are given so that the Nash equilibrium transmission policy of each user is a randomized mixture of pure threshold policies. We also present a stochastic approximation algorithm which can adaptively estimate the Nash equilibrium policies and track such policies for non-stationary problems where the statistics of the channel and user parameters evolve with time.