MIMO identification with optimal experiment design for rigid robot manipulators

This paper proposes a practical parameters identification procedure for robot manipulators. It is based on the well-known maximum likelihood method adopting particular techniques to facilitate the estimation: condition number reduction methods for the input data matrix with optimal trajectory planning, and two different methods for variances estimation. Moreover, the identification problem is solved with reference to the multi input multi output coupled system. A closed loop identification is needed because the system is open loop unstable, and, moreover, because to correctly execute the identification procedure, the robot has to track an optimal reference input. Some solutions are also presented to overtake common identification problems, such as bias of the estimated parameters, outliers detection and elimination, and noise sensitivity of the estimation. The presented procedure was successfully tested on a COMAU SMART3-S2 industrial manipulator demonstrating its efficiency.