Cognitive radio networks allow unlicensed users to access licensed spectrum opportunistically without disrupting primary user (PU) communication. Developing a distributed implementation that can fully utilize the spectrum opportunities for secondary users (SUs) has so far remained elusive. Although throughput optimal algorithms based on the well-known maximal weight scheduling (MWS) algorithm exist for cognitive radio networks, they require central processing of network-wide SU information. In this paper, a new distributed algorithm is introduced that asymptotically achieves the capacity region of the cognitive radio systems. The proposed algorithm achieves the full SU capacity region while adapting to the channel availability dynamics caused by unknown primary user activity. Extensive simulation results are provided to illustrate the efficacy of the algorithm.

Cognitive radio networks (CRNs) allow unlicensed users to opportunistically access the licensed spectrum without causing disruptive interference to the primary users (PUs). One of the main challenges in CRNs is the ability to detect PU transmissions. Recent works have suggested the use of secondary user (SU) cooperation over individual sensing to improve sensing accuracy. In this paper, we consider a CRN consisting of multiple PUs and SUs to study the problem of maximizing the total expected system throughput. First, we study the sensing decision problem for maximizing the system throughput subject to a constraint on the PU throughput, and we design a Bayesian decision rule-based algorithm. The problem is shown to be strongly NP-hard and solved via a greedy algorithm with time complexity O([(N5)/(log2[1/(1-ε)])]), where N is the total number of SUs. The algorithm achieves a throughput strictly greater than 1/2(1-ε) of the optimal solution and results in a small constraint violation that goes to zero with ε. We then investigate the more general problem with constraints on both PU throughput and the sensing time overhead, which limits the number of SUs that can participate in cooperative sensing. We illustrate the efficacy of the performance of our algorithms and provide sensitivity analysis via a numerical investigation.

In Cognitive Radio Networks (CRNs), secondary users (SUs) are allowed to opportunistically access the unused/under-utilized channels of primary users (PUs). To utilize spectrum resources efficiently, an auction scheme is often applied where an operator serves as an auctioneer and accepts spectrum requests from SUs. Most existing works on spectrum auctions assume that the operator has perfect knowledge of PU activities. In practice, however, it is more likely that the operator only has statistical information of the PU traffic when it is trading a spectrum hole, and it is acquiring more accurate information in real time. In this paper, we distinguish PU channels that are under the control of the operator, where accurate channel states are revealed in real-time, and channels that the operator acquires from PUs out of its control, where a sense-before-use paradigm has to be followed. Considering both spectrum uncertainty and sensing inaccuracy, we study the social welfare maximization problem for serving SUs with various levels of delay tolerance. We first model the problem as a finite horizon Markov decision process when the operator knows all spectrum requests in advance, and propose an optimal dynamic programming based algorithm. We then investigate the case when spectrum requests are submitted online, and propose a greedy algorithm that is 1/2-competitive for homogeneous channels and is comparable to the offline algorithm for more general settings. We further extend the online algorithm to an online auction scheme, which ensures incentive compatibility for the SUs and also provides a way for trading off social welfare and revenue.

Cognitive Radio Networks allow unlicensed users to opportunistically access the licensed spectrum without causing disruptive interference to the primary users (PUs). One of the main challenges in CRNs is the ability to detect PU transmissions. Recent works have suggested the use of secondary user (SU) cooperation over individual sensing to improve sensing accuracy. In this paper, we consider a CRN consisting of a single PU and multiple SUs to study the problem of maximizing the total expected system throughput. We propose a Bayesian decision rule based algorithm to solve the problem optimally with a constant time complexity. To prioritize PU transmissions, we re-formulate the throughput maximization problem by adding a constraint on the PU throughput. The constrained optimization problem is shown to be strongly NP-hard and solved via a greedy algorithm with pseudo-polynomial time complexity that achieves strictly greater than 1/2 of the optimal solution. We also investigate the case for which a constraint is put on the sensing time overhead, which limits the number of SUs that can participate in cooperative sensing. We reveal that the system throughput is monotonic over the number of SUs chosen for sensing. We illustrate the efficacy of the performance of our algorithms via a numerical investigation.

In this paper, the power control problem is considered for access-point based wireless networks under the SINR interference model. This problem is NP hard in terms of the number of APs in the network, and if not designed properly can have high polynomial complexity in terms of the power levels. We first separate the overall problem into two sub-problems using the primal-dual method. We then develop an efficient algorithm under the SINR-interference model that uses only two power levels to solve a subproblem of the overall power control problem, which we call the utility-independent power control (UIPC) subproblem. This approach allows the optimality gap of the UIPC subproblem to be bounded. Moreover, a two-stage iterative algorithm is developed that solves the overall power control problem within a small neighborhood of the global optimum. This offline algorithm has a significantly lower computational complexity than the globally optimal algorithm. Further, based on the structure of the iterative algorithm, an efficient heuristic two-stage greedy algorithm is proposed with low polynomial time complexity. Finally, numerical results are provided to demonstrate the efficacy of our solution.