This thesis proposes an adaptive inverse control for multivariable discretetime systems. The adaptive inverse control is well known for its excellent ability in tracking and disturbance rejection. However, previous adaptive inverse control does not apply to multivariable systems because multivariable transfer function matrices do not commute with each other. To resolve this problem, this thesis uses a feedforward compensator to decouple the multivariable systems, and then apply the adaptive inverse control to each channel of the decoupled systems. The proposed adaptive inverse control works for both minimum-phase and non-minimum phase multivariable systems.