A Controllable-Stiffness Tensegrity Robot Joint for Robust Compliant Manipulation

With concerns about safety in human–robot interactions, there is a growing demand for inherent compliance in robot manipulation, especially in healthcare applications, where remote tendon-driven mechanisms have drawn increasing attention, as they can reduce the overall robot joint size and weight by decoupling the motor from the rotary joint via tendon transmission. However, tendon preloading or any external load on the robot links would still be the predominant cause of excessive friction at the joint, deteriorating overall efficiency of the remote mechanical transmission. Our recent work proposed a tensegrity structure as a tendon-driven parallel mechanism for robot joint actuation, in which rotary/sliding friction at the joint can be totally avoided. The stiffness of such tensional integrity structure could be controlled by just tightening the tendons in parallel. Herein, we intend to integrate both tendon force/tension sensors and primitive axial stiffness modulators along the tendons, to close the stiffness control loop by a feedback model that estimates statics-equilibrium stiffness. A stiffness controller is also proposed, which can be operated in hybrid feedback modes involving a model-based stiffness estimator and a data-driven compensator. The proposed control framework is validated in particular for robot-assisted ultrasound scanning. We demonstrate that, even using simple or primitive stiffness modulators integrated along the tensegrity tendons, the robot joint stiffness can be controlled steadily under synthesized dynamic disturbances. The proposed data-driven stiffness compensator could compensate for uncertainty in modeling the complex statics equilibrium of our tensegrity structure, ensuring high-fidelity stiffness control.