TABLE OF CONTENTS In the last few years, several control laws have been proposed based on the dynamic model, as we will see in Chapter 5. However, whether we deal with the implementation of these laws in a robot controller or with the simulation of dynamic equations, it is necessary to have a good knowledge of the numeric values of the dynamic parameters of the robot. This section will show how to exploit that the dynamic model is linear in these parameters in order to identify them. Hence, the principle returns to the solution of an over-determined linear system using least squares techniques.
We suppose that the geometric parameter values are known (see section 1.4). The dynamic parameters of link j correspond to the inertial parameters of the link and of the actuator's rotor j as well as to the friction parameters. Thus, we write them as a vector 
:
The dynamic parameters of the robot are represented by vector 
as follows:
n being the number of joints of the robot.
The identification of the dynamic parameters of robots has been the subject of plenty of research: [FER 84], [MAY 84], [AN 85], [ATK 86], [KAW 88], [KHO 85], [GAU 86], [OLS 86], [BOU 89], [RAU 90], [AUB 91], [PRU 94], [RES 96]. The proposed methods have numerous common points, such as:
use of a model linear in the unknown dynamic parameters (dynamic model, energetic model, power model or...
