Adaptive Tactile Force Control in a Parallel Gripper With Low Positioning Resolution

In this letter we develop an adaptive tactile force controller for a parallel gripper with low positioning resolution. We show both mathematically and experimentally that a standard integral controller is not suited to control the force in a closed-loop fashion as it induces oscillations in the system. Therefore, we devise an adaptive controller that exploits the tactile readings to model the relationship between the gripper position and the tactile force readings and uses it to avoid the oscillations. We provide a mathematical characterization of the proposed controller and extensive experimental results with objects from the YCB model set. The results show that the proposed controller outperforms the integral controller and reduces the oscillations in a substantial way. Moreover, the position/force relationships are estimated with good precision. We also test the proposed controller as part of a cascaded architecture for the control of the sliding velocity of an object. The experiment shows the effectiveness of the controller for tasks where it is required to track a reference force profile from a high-level controller.