Identifying mass, center of mass, and moment of inertia through natural oscillations for full-dynamics control of robot manipulators

• Main Author: Jamisola, Rodrigo S., Jr.
• Format: text
• Language: English
• Published: Animo Repository 2010
• Subjects: Oscillations; Robots > control; Optimization; Robots > Control systems; Mass; Other Engineering; Robotics
• Online Access: https://animorepository.dlsu.edu.ph/etd_doctoral/287; https://animorepository.dlsu.edu.ph/context/etd_doctoral/article/1286/viewcontent/CDTG004754_P__1_.pdf
• Institution: De La Salle University

Inclusion of the dynamics parameters information into the robot control is critical in achieving robustness in the robots performance. These dynamics parameters are mass, center of mass, and moment of inertia. Integrating the correct values of these parameters in the dynamics model helps achieve full-dynamics force and motion control of robot manipulators. Such type of control is necessary for a manipulator to behave robustly, especially when it is required to interact with an unstructured environment. However, the greatest challenge remains on how to identify these parameters accurately. The absence of a generic method to identify dynamics parameters of a multi-body system makes full-dynamics control hard to achieve. One successful full-dynamics control of a mobile manipulator used a simplified symbolic mathematical model (whose lumped inertias are independent) to identify its dynamics parameters. Despite its success, its major limitation is its requirement of a simplified symbolic mathematical model which can become very computationally expensive especially with higher degrees of freedom robots. In addition, only the lumped inertias, and not the individual inertias, were identified in that previous successful implementation. This work has improved on the mentioned previous work and has identified the individual dynamics parameters without the need of a symbolic dynamics model, thus, avoiding the computationally expensive process in deriving it. In addition, individual dynamics parameters were treated separately. Two experimental procedures are presented: mass and center of mass identification, and inertia identification. In both cases the experimental methods are optimization computations with well-defined objective functions. For the mass and center of mass identification experiment, the correct parameters are identified when the minimum natural frequency of oscillation, min2, is achieved. In the inertia identification experiment, the correct parameters are identified when the square of the natural frequency of oscillations become equal to the proportional gain, 2 = kp. The experimental methods are analyzed and theorems are presented that support the claims presented in this work. These theorems can be used as tools for verification to check the accuracy of a given manipulator dynamics model. In addition, the experimental procedure presented in this work is unique because, to the best of our knowledge, it is only in this work where known dynamics parameters are derived experimentally to verify the accuracy of the proposed experimental procedure. In addition to the dynamics identification procedures, optimization methods using metaheuristic computations are presented in this work, namely, probabilistic artificial neural network, simulated annealing, and modified genetic algorithm. These optimization computations are used in finding the minimum-norm-residual solution to linear systems of equations. By demonstrating a set of input parameters, the objective function and the expected results, solutions are computed for determined, overdetermined, and underdetermined linear systems. This work presented an overview and provide the basic understanding on implementing metaheuristic optimization techniques. In addition, this work has presented a genetic algorithm with a modified approach in terms of reproduction and mutation. Experimental results for randomly generated matrices with increasing matrix sizes are presented and analyzed.