A common way to identify the inertial parameters of robots is to use a linear model as function of a minimal set of base parameters and standard least squares techniques. In experimental applications, noise on position and torque measurements, friction modeling error and bad excitation restrict dramatically the identification. This paper presents a methodology to overcome these difficulties. The proposed identification method is based on experiments which are designed by means of physical interpretation and spectrum analysis of the robot dynamic model in order to reduce sensitivity to noise. These experiments are planned in order to ensure optimal condition number of the observation matrix. The proposed algorithms have been integrated in a software package called Robot Identification Software Tool (RIST). The successful application of this new method to a 3 degrees of freedom robot proves the efficiency of the algorithms.