Robust Multiuser Sequential Channel Sensing and Access in Dynamic Cognitive Radio Networks: Potential Games and Stochastic Learning

This paper studies the problem of multiuser sequential channel sensing and access in dynamic cognitive radio networks, in which the active-user set is randomly changing from slot to slot. Furthermore, each user only has its individual information with no information exchange among users. The goal of the users is to determine their channel sensing order. We first define a generalized interference metric to address the overlapping of channel sensing order and establish two optimization objectives: minimizing the aggregate interference for each active-user set and minimizing the expected aggregate interference for all potential users. It is challenging to solve the two optimization problems, even in a centralized manner, because the active-user set is randomly changing, and the probability distributions of the active-user sets are unknown to the users. We then propose two noncooperative game models to solve the optimization problems: a state-based one-shot game and a robust game. We prove that they are potential games and that the best Nash equilibrium of the two games corresponds to the optimal solutions of the two optimization problems, respectively. To cope with the uncertain, dynamic, and incomplete information constraints in the dynamic networks, we propose a stochastic learning algorithm, which is analytically proven to converge to Nash equilibria of the two formulated games in the presence of a changing active-player set. Finally, simulation results are presented to validate the convergence and superior performance of the proposed learning algorithm.