The effective computed--torque control of robotic manipulators relies
on accurate estimations of the inertial parameters of the manipulator.
Identification techniques based upon manipulator force and motion
measurements are attractive for a number of reasons. However, the
complex structure of robot dynamics can make estimation of the
parameters difficult and the results of experiments hard to interpret.
In this project we examine the algebraic structure of the models used
in robot dynamics for purposes of inertial parameter identification.
Emphasis is placed on off--line techniques and the use of special test
motions. Issues of concern are joint--friction modelling, trajectory
optimization for identification, verification of the identified
parameters, and the use of linear and nonlinear error-in-variables
identification methods. Both manipulator joint torque and base
reaction models are being studied. Experiments involving the two
models are also undertaken, those for the base reaction model
employing a high resolution force-and-moment sensing platform.
R. Lucyshyn, J. Angeles