In this paper, a sub-optimal controller is studied for a class of time-invariant systems subject to bounded disturbances. The controller is based on the singular perturbation and simultaneous perturbation methods to achieve stable online self-tuning while requiring a little information about the plant model. The suggested controller is an extension of a self-tuning PID controller of Davison by adopting the gradient search scheme and the simultaneous perturbation method to tune the controller parameters. Further optimization effort is added by extending the self-tuning PID controller to a high-order controller with the stable self-tuning property. Both simulation and experimental results are provided to verify the feasibilities of the proposed controller.