Power Allocation in Multiuser Parallel Gaussian Broadcast Channels With Common and Confidential Messages

We consider a broadcast communication over parallel channels, where the transmitter sends K+1 messages: one common message to all users, and K confidential messages to each user, which need to be kept secret from all unintended users. We assume partial channel state information at the transmitter, stemming from noisy channel estimation. Our main goal is to design a power allocation algorithm in order to maximize the weighted sum rate of common and confidential messages under a total power constraint. The resulting problem for joint encoding across channels is formulated as the cascade of two problems, the inner min problem being discrete, and the outer max problem being convex. Thereby, efficient algorithms for this kind of optimization program can be used as solutions to our power allocation problem. For the special case K=2 , we provide an almost closed-form solution, where only two single variables must be optimized, e.g., through dichotomic searches. To reduce computational complexity, we propose three new algorithms, maximizing the weighted sum rate achievable by two suboptimal schemes that perform per-user and per-channel encoding. By numerical results, we assess the performance of all proposed algorithms as a function of different system parameters.