Dual Loop Compliant Control Based on Human Prediction for Physical Human-Robot Interaction

In physical human-robot interaction (HRI), human partner's operating force and the performance of the HRI process are the two main criterions. Lower operating force and higher tracking accuracy are demanded during physical HRI. In order to achieve such two goals, we propose a dual loop force assist framework for HRI. In this framework, it includes an outer force loop and an inner position loop. In the outer loop, a model reference adaptive control (MRAC) based compensation method is proposed to assist a human operator to interact with robot. The compensation works as an assist force to minimize the error between human desired position and a robot actual position. Besides, a Sparse Bayesian learning (SBL) based human intention predictor is also proposed to predict the future human desired position, thus the MRAC based compensation can work on-line. For the inner loop, the compensated force is used as the input of prescribed impedance, which gives compliance to the robot. Then, the Cartesian drive torque for a n-DOF robot manipulator has been analyzed and a Proportional + Derivative (PD) based controller with gravity compensation is adopted to generate the drive torque. So that the robot can track the output of prescribed impedance accurately. The effectiveness of the proposed framework is verified by simulation study.