This thesis presents an experimental gain optimization scheme of time-delay control (TDC) systems using the Taguchi method with application to robot manipulators. TDC is well known as a simple, robust and decentralized approach for nonlinear systems. The choice of gains in TDC not only determines the stability of the system, but also affects the effect of disturbances and noises. Hence to find optimal gains becomes very meaningful. Since in the discrete domain, TDC and PID are equivalent, this research can be easily extended to the application of PID control systems. Taguchi’s robust parameter design is used and developed as a learning scheme to achieve gain optimization on-line. The integral-squared-error (ISE) is selected as the performance index to minimize both of the error magnitude and duration, while the signal-to-noise (S/N) ratio and analysis-of-variance (ANOVA) are used to analyze the results. Based on the ANOVA and general linear least squares method, an experimental algorithm for detecting the upper limit of the searching range is developed. The detecting algorithm is verified to be safe and effective by experiments. The proposed tuning method is implemented on the obtained searching range. It is shown by both simulations and experiments that the proposed scheme is efficient and satisfied. With experiments on a 2 DOF robot manipulator, the near-optimal values of the controller gains are obtained in about 5 iterations, which correspond to about 90 experiments. The tuning result is also confirmed by the performance index topology on the whole searching range.