Variable feedback configuration control of a robotic finger to produce anthropomorphic motions

A robustly stable index finger control algorithm is presented which allows a robotic hand to learn from experience. The proposed algorithm considers a specific case of environment interaction that requires simultaneous motion and posture control of the manipulator. In preliminary experiments, hand motion profiles of ten human test subjects were recorded while running their hands over a contoured surface. It was observed that participants simultaneously adapted to the contour of the surface while maintaining consistent hand posture. Results from these experiments were used to create a human response model that the manipulator learned to mimic by varying the feedback configurations to the motor controllers of the index finger. Results verify that the variable feedback configuration (VFC) control algorithm produces more biomimetic motion by reducing the difference between the response of the robotic hand and the human model. Three trials were conducted with the VFC controller, with sequential improvements in human model error reduction and robotic hand posture seen in each trial.