Variable Force Tracking Control via Online Learning

With the expanding range of manipulator applications, the ability to interact with the environment in a compliant and safe manner becomes increasingly important. In this paper, a force tracking impedance controller is proposed to perform variable force tracking in unknown environments. Specifically, an RBFNN-based adaptation method is employed to explore the contact dynamics and modify the reference trajectory online. The inner loop impedance controller generates compliant interactions and unifies task space motion and force tracking in the target model. Meanwhile, the force manipulability is optimized using null space control, which is motivated by human arm configuration during force exertion. Numerical studies are conducted to validate the proposed control scheme and results show high performance in force tracking and reduction of maximum joint torque.