A new iteration to coupled discrete-time generalized Riccati equations (Q382460)

The authors consider a quadratic optimal control problem concerning a time discrete Riccati system with Markovian jumps and try to find its maximal and stabilizing solution by an iterative procedure based on works of \textit{V. Dragan} and coworkers, see [Int. J. Control 83, No. 4, 837--847 (2010; Zbl 1209.93091)]. They decouple their procedure by proposing a Gauss-Seidel type iteration and investigate its convergence under new (compared to Dragan et al. [loc. cit.]) conditions on the elements of the transition matrix. They also provide results of a series of numerical experiments in which the basic matrices of the Riccati system are random and the weighting matrices are not positive definite but singular, negative definite or indefinite. The outcome is that the new iteration is not always but mostly better than the old one.