Using a stereo camera mounted on the robot’s head, 3D information was precisely calculated for grasping the object. The grasping position had been already set in modeling and was obtained when the location estimation of the object was assumed. When the robot moved its hand to this position, it could grasp an object such as a milk cup or toast. When the robot did so, it was essential to control the grasp strength. The strength of grasp had been set experimentally according to the object type.