This paper describes a modeling method based on multibody dynamics (MBD) and autonomous control of a dual-arm robot with a stereo camera. Each arm of the robot has seven degrees of freedom, similar to a human arm. Even a robot with a lot of links, a model of the robot should be expressed fully that a check experiment can be done. One of MBD method can express the motion equation of a robot with a lot of links compactly. A model for our robot is derived by MBD. The robot controller consists of an LQI controller and gravity compensation. Gravity compensation is designed using the MBD model. The positions of objects to be manipulated are obtained using 3D color data obtained using a stereo camera. We tested the effectiveness of our proposed method by experiment. As a result, the tracking performance of the LQI controller with gravity compensation was about 21% better than that without gravity compensation.