In this study, an idea to the self-tuning sliding mode controller design is proposed. The proposed controller not only reduces the chatting phenomenon, but also needs less the system bounds than traditional sliding mode controllers do. The proposed self-tuning sliding mode controller can be viewed as an improved hitting controller. Besides, in order to resolve the parameter selection problem, the Lyapunov stability theory is considered and a self-tuning machine is employed to reduce the conjecture of parameter selection. Finally, the proposed controller is employed to control a two-link robot arm, which is a class of MIMO nonlinear control system. The result is promising.