Sequential interference subtraction multi-user transceivers: Designs for general bounded channel uncertainty models

We study the design of non-linear multi-user communication systems that implement sequential interference (pre-)subtraction, such as broadcast channels with Tomlinson-Harashima precoding (THP) and multiple access channels with decision feedback equalization (DFE), with an emphasis on the broadcast case. We consider scenarios with partial channel state information in which the channel uncertainty is deterministically bounded. We propose a general channel uncertainty model that embraces many bounded uncertainty regions, and we study the robust design of THP transceivers for the broadcast channel that minimize the maximum MSE over all set admissible channels. We show that the design problem is NP-hard, and we propose an iterative local optimization algorithm that is based on efficiently-solvable convex subproblems. We then generalize the robust designs to the case in which the channel uncertainty is described by the intersection of bounded regions. The robust design framework is also generalized to multiple access channels with DFE and bounded channel uncertainty. Simulation studies demonstrate that in presence of channel uncertainty, the proposed robust design framework can result in considerable improvement in the performance of THP-based transceivers for the broadcast channel.