Adaptive Collaborative Position Control of a Tendon-Driven Robotic Finger

This paper presents a collaborative control scheme to optimize the trajectory-tracking performance of the motor that is installed at the metacarpophalangeal (MCP) joint of a robotic finger. A dynamically compensated adaptive-proportional-derivative (APD) tracking controller is used to nullify the tracking errors during free-space motion. An APD compliance controller is used to alleviate disturbance caused by the application of contact-force during the grasping operation. It achieves this objective by altering the reference trajectory, based on the magnitude of applied force. The PD gains of both controllers are adaptively tuned in order to quickly respond to the changes in the system dynamics. Two different intelligent self-tuning mechanisms are used and comparatively analyzed to adaptively adjust the PD gains of these controllers; namely, fuzzy inference system (FIS) and iterative learning algorithm (ILA). An adaptive-neuro-fuzzy-inference-system (ANFIS) is used as an inverse model to transform the reference trajectory into joint-angle dynamics of the finger. It also acts as a feed-forward controller and supervises the trajectory tracking. The feed-forward and tracking controller outputs are beneficially combined via a linearized feedback control law to deliver optimal motor torque commands. The results of real-time experiments are presented to validate the robustness of the proposed controller.