Determining the Dynamic Parameters of a Robot Using the Method of Instrument Variables

The identification of robot dynamic parameters is usually based on the use of the reverse model which is linear in relation to the dynamic parameters to be identified. In order to get an overdetermined system, this model is sampled while the robot is trained with fascinating movements. The optimum solution is obtained with linear least squares (LCL). The efficacy of this approach has been demonstrated through the experimental identification of numerous industrial prototypes and robots. It has been expanded to other systems including compactors, cars, engines and haptic interfaces. Though, this method requires accurate measurement at a fairly high rate of torque and position image. It also requires estimating speeds and accelerations by band-pass filtering positions. This results in a noisy observation matrix. Furthermore, the identification is done in a closed loop due to the unstable nature of the robots' dual integrator. The instrumental variable (IV) method addresses the problem of the noisy observational matrix and can be statistically optimal. This article highlights this technique and proposes an extension of this method which is applied to a flat robot of 2 degrees of freedom (DOF).