OPTIMAL EXCITING TRAJECTORIES FOR ROBOT DYNAMICS IDENTIFICATION

in this study we discuss different techniques to find an optimal traje ctory for the purpose of identification of an inertial parameters of r obot. Two dynamical models have been introduced: differential and inte gral; both of them can be used to estimate the robot dynamic parameter s in which they appear to be linear. Functions associated with the par ameters are highly nonlinear which in general involve joint positions, velocities and accelerations. A standard sequential least squares tec hnique has been used to estimate the unknown parameters. This method r equires an, information matrix which has to be inverted. For particula r joint signals this matrix is singular. In this paper we proposes to optimize the condition number of this matrix. This work illustrates th e difficulty of maintaining persistent excitation during the experimen tal identification of complex models.