Dynamic Stiffness Identification of Adjustable Stiffness Joint Based on Decoupling and Linearization

In order to take advantage of adjustable stiffness joint to adjust robot's dynamic feature, it is necessary to effectively identify and control the dynamic stiffness of the joint.Firstly, a simplified model is derived based on the structure features of robotic adjustable stiffness joint, and the assumption of stiffness output form is made.Then the torque related parameters in the model are decoupled, to eliminate the effect of adjusting parameter of joint stiffness on the torque, and thus the unified torque expression for stiffness identification is acquired.Linearization of the unified torque expression is then carried out by utilizing Tailor expansion, and Kalman filter is applied to optimizing the factors of the expansion.Based on this, the joint dynamic stiffness identification is achieved.It is proved the identification error is controlled within ± 2% by the dynamic stiffness online identification method in simulation.Based on the result of dynamic stiffness identification, feedforward based joint stiffness closed-loop control method is then studied.Simulation experiments show that the method is effective for robotic joint stiffness closed-loop control.