Many recent studies on autonomous driving have focused on model-based control. A number of studies has addressed that simple models such as the Kinematic Bicycle Model are easier to design controls for autonomous driving systems. However, such a simple vehicle model has a weakness in that it is subject to modeling errors. This is because it does not take into account the nonlinear characteristics due to road conditions and driving conditions (environmental disturbances: road friction coefficient, large steering, acceleration, sideslip, etc.) Therefore, the purpose of this study is to identify vehicles with high accuracy and in real time, adapting to environmental disturbances. This study propose a vehicle model based on the Kinematic Bicycle Model. The nonlinear characteristics of the vehicle are represented by the deviation of the front wheel steering angle of the Kinematic Bicycle Model. This deviation is trained and estimated online using a three-layer Neural Network. In other words, the AI is adaptive learning of modeling errors caused by nonlinear characteristics of the vehicle. This paper presents an example of model-based control using model predictive control.