Full Text

Introduction

In this work, a robotic manipulator dynamic model is developed starting from the Gibbs-Appell equations. It is assumed that the robot is constituted by rigid elements, being the actuator dynamic behavior independently modeled. In order to model joint friction, several friction models have been considered.

The equations constituting the dynamic model have been rewritten linearly with respect to the model dynamic parameters, i.e. masses, first-order moments, components of the link inertia tensors and friction parameters, which considerably simplifies identification. These linear expressions have been written in matrix form. The calculation of the different terms, as well as its computational complexity is detailed. The matrix form expression of the system equations allows the application of numerical techniques for analyzing, reducing and solving the obtained equation system and is used both for identification and later to solve the direct and inverse dynamic problem.

By means of a parameter identification process, a precise estimation for the values of masses, locations of the gravity centers and components of the link inertia tensors, as well as for the joint friction terms, have been obtained. Therefore, experimental measurements have been used, obtained during robot motion for an optimized trajectory exciting the different terms in the equations modeling the dynamic behavior of the system.

Once the dynamic parameters of the robot dynamic model are obtained, starting from the matrix form of this dynamic model, the different terms in the system dynamic equation are determined, which allows to solve the direct and inverse dynamic problems.

Both the robot dynamic modeling and the identification methods have been validated by their experimental application in a PUMA 560 industrial robot. The proposed techniques have been validated by the development and execution of several control strategies based in the dynamic equation whose parameters have been identified. Therefore, the industrial robot has been provided with an open control unit based in an industrial computer. This control architecture allows, by means of C++ programs, the development of applications and real-time control algorithms.

Dynamic modeling and parameter identification in robotic manipulators

Motion equations modeling the dynamic behavior of a mechanical system can be obtained starting from different dynamics principles. The first works date back to 1965 when [24] Uicker (1965) formulated a method based in the Lagrange equations and...