Sparse Output Feedback Synthesis via Proximal Alternating Linearization Method

Abstract: We consider the co-design problem of sparse output feedback and row/column-sparse output matrix. A row-sparse (resp. column-sparse) output matrix implies a small number of outputs (resp. sensor measurements). We impose row/column-cardinality constraint on the output matrix and the cardinality constraint on the output feedback gain. The resulting nonconvex, nonsmooth optimal control problem is solved by using the proximal alternating linearization method (PALM). One advantage of PALM is that the proximal operators for sparsity constraints admit closed-form expressions and are easy to implement. Furthermore, the bilinear matrix function introduced by the multiplication of the feedback gain and the output matrix lends itself well to PALM. By establishing the Lipschitz conditions of the bilinear function, we show that PALM is globally convergent and the objective value is monotonically decreasing throughout the algorithm. Numerical experiments verify the convergence results and demonstrate the effectiveness of our approach on an unstable system with 60,000 design variables.