Componentwise link identification of direct-drive robots

We present a novel joint-torque based identification method of link dynamic parameters for direct-drive rigid manipulators with revolute/translational joints. The key of the proposed method resides in the construction of the base parameter set. This set is extracted from the equations of motion of the manipulator by initially separating joint displacements instead of parameters, as done in classical methods. This was achieved by dynamically separating the equations of motion of the manipulator. We separate the gravitational, Coriolis-centripetal, and inertia dynamic components of the joint torques/forces. Three associated subsets of base dynamic parameters are used in such a way that the original identification problem is replaced with three smaller problems. These simpler problems are then solved successively. This increases the accuracy of estimation, reduces the number of parameters to be identified, reduces the number of identification experiments, and creates additional means to estimate the accuracy of identification. Finally the proposed methodology facilitates the solution of persistent excitation problem