On an iterative algorithm to compute the positive stabilizing solution of generalized algebraic Riccati equations

An iterative algorithm to solve a kind of generalized algebraic Riccati equations (GARE) in LQ stochastic zero-sum game problems is proposed. In our algorithm, we replace the problem of solving a GARE with an indefinite quadratic term by the problem of solving a sequence of GARE with a negative semidefinite quadratic term which can be solved by existing methods. Under some appropriate conditions, we prove that our algorithm is globally convergent.