This paper presents an identification method for robotic manipulators.
It demonstrates how a dynamic model can be constructed with the help of the
modified Newton–Euler formula. To model the friction of the joints, static
friction modelling is used, in which the friction behaviour depends only on
the actual velocity of the given joint. With these techniques, the model can
be converted into a linear-in-parameters form, which can make the
identification process easier. Two estimators are introduced to solve the
identification problem, the least-squares and the weighted least-squares
estimators, and the determination of the independently identifiable
parameter vector to make the regression matrix maximal column rank is
presented. The Frobenius norm is used as the condition of the regression
matrix to optimise the excitation trajectories, and the form of the
trajectories has been selected from the finite Fourier series. The method is
tested in a simulated environment to achieve a three-degrees-of-freedom
manipulator.