MSE-based linear transceiver optimization in MIMO cognitive radio networks with imperfect channel knowledge

This paper addresses the robust transceiver optimization in multiple-input and multiple-output cognitive radio network, where primary users (PUs) and secondary users (SUs) coexist in the same spectrum band. In the design of cognitive system, the performance degradation perceived by PU should be strictly restricted even with imperfect channel state information (CSI) at cognitive transmitter and receivers. Therefore, this work aims at minimizing the sum mean square error of secondary downlink network and strictly limiting the interference caused to PUs with imperfect channel knowledge. Two types of CSI error models are considered: the bounded model and the stochastic model. Since the original optimization problems are non-convex for the joint optimization, firstly it is decomposed into two subproblems to optimize the precoding and equalizers separately, then the iterative algorithms are proposed to solve the subproblems in an alternating way. The challenge is to design the efficiently solvable forms of these subproblems. For the bounded model, Schur complement lemma is utilized to convert the subproblems into convex optimization problems. For the stochastic model, the problem is formulated either according to the stochastic rule or derived for the analytical solutions. The effectiveness and robustness of proposed algorithms are evaluated by the numerical results.