Repetitive Impedance Learning-Based Physically Human-Robot Interactive Control.

Area of Science:

• Robotics
• Control Systems Engineering
• Human-Robot Interaction

Background:

• Model-based impedance learning control allows robots to adjust impedance online without force sensors.
• Existing methods have limitations, requiring periodic or slowly varying human impedance and only guaranteeing uniform ultimate boundedness (UUB).

Purpose of the Study:

• To propose a novel repetitive impedance learning control approach for physical human-robot interaction (PHRI) in repetitive tasks.
• To address limitations of existing methods by handling iteration-dependent disturbances in human impedance profiles.

Main Methods:

• A control strategy combining proportional-differential (PD) control, adaptive control, and repetitive impedance learning.
• Differential adaptation with projection modification for robotic parameter uncertainty estimation.
• Fully saturated repetitive learning for time-varying human impedance uncertainty estimation.

Main Results:

• Uniform convergence of tracking errors is theoretically proven using Lyapunov-like analysis.
• The approach effectively estimates and compensates for iteration-independent and iteration-dependent disturbances in stiffness and damping.
• Simulations on a parallel robot demonstrate control effectiveness in a repetitive following task.

Conclusions:

• The developed repetitive impedance learning control is effective for PHRI tasks with iteration-dependent disturbances.
• This method enhances robot adaptability and control accuracy in dynamic human-robot collaborations.
• The approach overcomes limitations of prior impedance learning techniques, offering improved system stability and performance.