Reversible Position Domain Notch Filter for Suppressing Vibrations Caused by Kinematic Errors in Flexible Joint Robots

Position domain control is an effective way to suppress position-dependent vibrations caused by kinematic errors in harmonic drives for flexible joint robots. However, existing position domain methods require motor rotation to be irreversible, while the motor usually needs to reverse in robotic applications. In this article, a reversible position domain notch filter (RPDNF) is proposed to suppress vibrations caused by kinematic errors for flexible joint robots. First, a RPDNF is designed to overcome the limitation of irreversible motor rotation, by defining a novel monotonically increasing angle position variable. Second, the position domain notch filter is transformed into time domain form and is implemented in a time sampled system to reduce implementation complexity. Third, a state reinitialization scheme is proposed to synchronize the phase of the notch filter output, thus improving the notch filter dynamic performance when the motor reverses rotation. Finally, the proposed notch filter is combined with a cascade controller with full-state feedback to control the flexible joint robots. Simulation and experimental results demonstrated the effectiveness of the proposed method.