Bandwidth partition and allocation for efficient spectrum utilization in cognitive communications

Conventional cognitive communications rely heavily on the smartness of secondary (unlicensed) users (SUs) to achieve high spectrum utilization, which involves the optimization of the SUs' policies and behaviors for dynamic spectrum access, power allocation among multiple channels, etc. Due to the inherent randomness of the primary users' (PUs') transmission, those efforts inevitably increase the implementation complexity and sensing overheads of the SUs, and in turn lower the spectrum utilization efficiency. In this paper, we try to change the focus from SU to PU. A cooperative traffic allocation strategy for PU, together with the non-uniform bandwidth partition, is employed to regularize the PU's resource occupancy pattern without compromising its performance, and to maximize the spare bandwidth for the SU at the same time. We first study the capacity based optimization problem (COP) together with the fully polynomial time approximation scheme (FPTAS) for an approximation guarantee of the global optimum. Then we analyze the subcarrier based optimization problem as the surrogate problem of COP, which can be solved by a greedy algorithm exactly. Both the theoretical analysis and the numerical simulations demonstrate the effectiveness of those methods to achieve the performance that almost identical to that of the global optimum solution.