Resource Allocation for OFDMA Based Cognitive Radio Networks with Arbitrarily Distributed Finite Power Inputs

The existing works on resource allocation for OFDMA based cognitive radio networks are based on the assumption of Gaussian inputs whereas in practical systems the inputs are taken from a set of finite symbol alphabets. This paper considers a system with arbitrarily distributed finite power inputs and solve the resource allocation problem by employing the relationship between mutual information and minimum mean-square error. To protect the primary users' links, constraint on interference power of the secondary users (SUs) is imposed. In OFDMA based CR networks, a tone can be assigned to one SU at most (exclusivity constraint), due to which the resource allocation problem becomes combinatorial and its solution becomes prohibitively difficult.In this paper, first, the exclusivity constraint on tones allocation is relaxed, the problem is convexified and an optimal solution is derived that provides an upper bound on the system performance. Then, an integer tone allocation and optimal power allocation (ITA-OPA) algorithm is developed that guarantees the assignment of each tone to a single SU with close-to-optimal performance. Finally, keeping in view the complexity of the optimal solution and ITA-OPA algorithm, a low-complexity suboptimal algorithm is devised that accounts for exclusive tone assignment. Simulation results show that the suboptimal algorithm also achieves near-optimal performance. The proposed algorithms outperforms the algorithms that assume Gaussian inputs.