TMR online optimization using quasi-Newton method for HDD servo systems

Presents a direct online parameter optimization method using a quasi-Newton approach to find compensators, which can minimize the measured track mis-registration (TMR) for hard disk drive (HDD) servo systems, without prior knowledge of disturbance/noise models. An optimal controller for a plant with uncertainty can be obtained within a pre-found robust stable region by using the gradient information of TMR with respect to controller parameters based on nominal plant model. The searching time is shorter compared with other non-gradient optimization methods such as random neighborhood search and genetic algorithms. Both simulation and implementation results show the effectiveness of the proposed method. In addition, it is also proved that if the measurement noise is white noise, an optimal controller that minimizes the measured TMR also minimizes the true TMR.