Development of smart gripper for identification of grasped objects

Force control is one of the most important schemes in many industrial applications, especially in painting and grasping tasks. In practice, the proper control parameters in force controller is not easy because the performance of entire system does not only depend on the actuator dynamic, but also the environment and grasped object. In addition, several applications need to use the force control system for grasping various types of the object, this results in the dynamic change and poor performance in nature. To deal with this problem, the fast dynamic identification using a pair of input and output data is proposed to identify the plant dynamic of the force control system. The position and force are collected and used for the structured dynamic identification. Predefined clusters determined from several data sets of different objects are evaluated using K-means clustering. As seen in the experimental results, the proposed gripper system can identify the object group correctly.