碩士 國立清華大學 通訊工程研究所 101 In this paper, we study a spectrum sharing based cognitive radio network (CRN) where $N$ secondary users (SUs) share the same spectrum with a primary user (PU) over block fading channels. We assume complete perfect channel state information (CSI) at the secondary transmitters while only local instantaneous CSI is assumed at the primary transmitter. The PU is assumed to adopt an ON-OFF power control policy and convey this one-bit side information to all secondary transmitters. Based on these assumptions, we investigate the optimal probabilistic power allocation that seeks to maximize the system utilities for SUs subject to the primary interference power constraint, the secondary rate outage constraints and the average power constraints. The probabilistic power allocation problem was first reformulated as a deterministic power allocation problem through defining a set of weighting variables based on the conditional probability and conditional mean of the probabilistic power random variables. The resulting optimization problem, however, is still non-convex in general. To handle the non-convex constraints, we applied a conservative convex first-order approximation technique. Furthermore, by the successive convex approximation (SCA), we proposed an algorithm that provides high-quality approximate solutions via solving a sequence of convex approximation problems. Our theoretical analysis further demonstrated that the limit point generated by our proposed SCA algorithm is indeed a stationary point of the original optimization problem. To further reduce complexity, a decentralized version of the SCA algorithm was proposed, where only a limited amount of information exchange between the secondary transmitters is required. The convergence analysis was also provided for the decentralized counterpart. Extensive simulations validated our analyses and demonstrated that near-optimal performance is indeed achieved by both our proposed algorithms.