The path tracking controller is one of the most important technology for safety of autonomous vehicle. However, there are lots of disturbances(i.e. path curvature, road bank angle, side wind) that reduce controller tracking performance in driving situation, and therefore we have to consider those disturbances in control system. Among them, the curvature disturbance is the most significant and the curvature can be observed in the planned path, so it can be directly used for the controller. In this paper, we present improved autonomous vehicle path tracking LQR controller by considering exogenous variable curvature disturbance. By adding curvature disturbance in state and cost function, we can get optimal control gain and steering input, and experimental results show the tracking performance of the proposed controller.