A gradient flow algorithm models developed for the on-line robust pole assignment is proposed for solving
Sylvester equations. The algorithm shows to be capable of synthesizing linear feedback control systems via on-line
computing feedback gain matrix and desired closed-loop poles. Meanwhile, the close-loop system matrix is least sensitive to
perturbation or uncertainty，and uniformly asymptotically stable in largely range. Simulation results show that the proposed
approach is suitable to the problem of robust stabilization for nonlinear system and on-line robust pole assignment.