DIDIM-CLIE method for dynamic parameter identification of flexible joint robots

This paper deals with a new CLIE (Closed Loop Input Error) method for dynamic identification of flexible joint robots. This is a straightforward extension of the DIDIM (Direct and Inverse Dynamic Identification Models) method from rigid robots to flexible joint robots. DIDIM is a fast closed loop input error method which minimizes the quadratic error between the actual motor force/torque (the control input) and the simulated one. This method which does need neither joint position, nor velocity nor acceleration is particularly attractive to identify flexible manipulators where data of flexible degree of freedom are not available at all. This paper shows how to easily choose good initial values based on mechanical design of joint transmission to apply DIDIM method to flexible joint manipulators. This is a dramatic improvement of our former three-step method which allows a one-step procedure with faster convergence thanks to the clever initialization. An experimental validation on a flexible joint robot and a comparison with the previous three-step procedure shows the effectiveness of the new method.