Sensing Constrained Multiuser Learning Strategy In Cognitive Radio Networks

With the rapid development of new wireless applications, there is a greater demand of wireless radio spectrum. However, the fixed spectrum assignment policy has restricted the utilization of the limited spectrum. The limited spectrum are assigned to licensed users, also known as primary users. However, a large portion of the assigned spectrum remains under utilized. To alleviate the inefficient usage of the limited spectrum, dynamic spectrum access techniques are developed, where users who have no spectrum licenses, also known as secondary users, are allowed to opportunistically access the temporarily unused spectrum. Cognitive Radio(CR) provides an effective solution to the realization of the flexible dynamic access techniques.In a cognitive radio network, users are intelligent and have the ability to observe and learn. Users are aware of the dynamic environment and adaptively adjust their operating parameters based on the interactions with the environment and other users in the network. The paper mainly considers the problem of learning the parameters of the unknown environment and efficiently sharing of the idle spectrum among multiple secondary users with limited sensing capability.Firstly, the paper briefly introduced the basic principle of cognitive radio and the model of spectrum access. The limited sensing capability of secondary users and the potential problem of spectrum sharing are also discussed.Then we studied the sensing and access strategy under the scenario where the number of secondary users is less than the number of primary channels. Existing dynamic access approaches make the first order statistical estimation based on the past sensing experiences and calculate the average reward, which would lead to prohibitively high complexity. We make an analysis of the exploration and exploitation tradeoff in the single user case. When the exploration factor is added to the learning process, the rate of convergence is efficiently accelerated. And we also extend the exploration and exploitation into the multiuser case. With multiple secondary users contending for spectrum opportunities, we must take into account the collisions among them. The mixed strategy based channel selection mechanism is proposed to alleviate the multiuser collisions. It is very important to adopt mixed strategy. When the number of secondary users is less than that of the channels, each secondary user would settle down to the orthogonal channels with high probability. Also the learning method would converge to the optimal strategy compared with that of centralized case.When the number of secondary users is large, a multiuser learning algorithm based on evolutionary dynamics is proposed. It takes into consideration of multiuser competition, and achieves fair sharing of the idle spectrum among secondary users. Meanwhile, even if some users play arbitiry strategies, the learning method can recover to the stable state efficiently.