Please use this identifier to cite or link to this item: http://hdl.handle.net/123456789/6539
Title: GENETIC FUZZY CONTROL OF PUMA 560 ROBOT MANIPULATOR
Authors: A, Mona Subramaniam
Keywords: ELECTRONICS AND COMPUTER ENGINEERING
GENETIC FUZZY CONTROL
PUMA 560 ROBOT MANIPULATOR
ROBOT MANIPULATOR
Issue Date: 2010
Abstract: Designing of control systems for a robot manipulator has many practical and theoretical challenges due to the complexities of the robot dynamics involved while achieving high precision- high velocity trajectory tracking in varying load conditions. Conventional robot control methods require highly accurate mathematical modeling, analysis and synthesis. Fuzzy Logic Controllers are a class of non-linear controllers that make use of human expert knowledge. Genetic algorithm is a class of evolutionary algorithms that can be applied to any problem which can be formulated as function optimization problem and it provides a way of optimizing fuzzy controller design. PUMA 560 is a six Degree of Freedom robot arm that has to be controlled. A Fuzzy PD+I controller is used for the control of this robot arm. In this dissertation Genetic Algorithm is used to optimize, approximate and minimize the Fuzzy Logic Controllers. An algorithm is proposed which has a faster convergence when compared to Genetic Algorithm. Simulations and computations are carried out using MATLAB version 7.0.1.24704(R14). Tuning of fuzzy parameters using genetic algorithm is carried out on reference Fuzzy PD+I controller. The parameters tuned are Rules, Rule-weights and Membership functions. As a part of the initial study, the parameters mentioned are tuned independently and it is seen that there is improvement in the performance. With this preliminary study, progressive tuning of parameters in three stages is carried out. At the end of second stage, tuning the rules in the rule base and then assigning weights to individual rule antecedent, considerable effect on the performance of the system is observed. Proceeding further with the third stage, membership function tuning is performed. There is an improvement in the performance from the second stage. In this study it is also observed that progressive tuning performs better than simultaneous tuning of parameters. One of the objectives in machine learning is to learn any system from data. Approximation of system from input-output data presents a way to learn a system. With the input and output data points collected from the reference Fuzzy PD controllers of individual joints, bi-variate polynomial approximation function and Weighted-rule Fuzzy approximation function is generated through interpolation process using genetic algorithm. The results of above two approximation functions are compared. The results obtained with that of a weighted Fuzzy approximation function is found to be superior with respect to the other. For any given input trajectory, all the rules in the rulebase are not fired and those fired may not be the best. This suggests that the rulebase can be minimized with only the best rule entries. The rule base so obtained is an optimally minimum rulebase which is achieved by tuning the rulebase in such a way that the number of rules is minimized and the rules in them are optimized simultaneously. A small disturbance is given to the robot arm and the trajectory tracking is evaluated. One of the main drawbacks of Genetic algorithm is its slow convergence rate. A simple stochastic optimization algorithm is proposed which has a good convergence rate. This algorithm is used for tuning the gains of the Fuzzy PD+I controller. The convergence is compared with that of Genetic algorithm and results of the proposed algorithm are quite encouraging. All the results are presented in Chapter 6. iv
URI: http://hdl.handle.net/123456789/6539
Other Identifiers: M.Tech
Appears in Collections:MASTERS' DISSERTATIONS (E & C)

Files in This Item:
File Description SizeFormat 
ECED G20444.pdf14.27 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.