Parameters Identification of an Experimental Vision-based Target Tracker Robot Using Genetic Algorithm

Abstract

In this paper, the uncertain dynamic parameters of an experimental target tracker robot are identified through the application of genetic algorithm. The considered serial robot is a two-degree-of-freedom dynamic system with two revolute joints in which damping coefficients and inertia terms are uncertain. First, dynamic equations governing the robot system are extracted and then, simulated numerically. Next, an open-loop experiment with finite duration step inputs is implemented on the experimental setup to collect practical output data. Accordingly, a desired objective function is defined as the sum of discrepancy between the experimental and simulated output data. Subsequently, a genetic algorithm is employed to explore the best damping coefficients and inertia terms of the simulation scheme so as to minimize the presented cost function and taking into account the same input data for both simulation and experiment. Finally, the simulated output data based on the identified robot parameters reveal an acceptable agreement with the measured outputs through which validity of the identification scheme is affirmed.