Pakistan Research Repository

ADVANCE CONTINUOUS GENETIC ALGORITHMS AND THEIR APPLICATIONS IN THE MOTION PLANNING OF ROBOT MANIPULATORS AND IN THE NUMERICAL SOLUTION OF BOUNDARY VALUE PROBLEMS

Hammour, Za’er Salem Abo (2002) ADVANCE CONTINUOUS GENETIC ALGORITHMS AND THEIR APPLICATIONS IN THE MOTION PLANNING OF ROBOT MANIPULATORS AND IN THE NUMERICAL SOLUTION OF BOUNDARY VALUE PROBLEMS. PhD thesis, Quaid-i-Azam University, Islamabad.

[img]HTML
34Kb

Abstract

In past, Gutowski proposed a special type of real-coded genetic algorithms, which is suited for continuous optimization problems with continuous and/or smooth solution curves. The algorithm uses smooth operators throughout the evolution process and results in smooth solution curves. However, Gutowski's algorithm is restricted to problems involving single solution curve and suffers from slow convergence rates. In this work, an advanced continuous genetic algorithm (ACGA) has been developed based on Gutowski's algorithm by incorporating new initialization functions and a novel performance enhancement scheme. ACGA is designed to deal with single as well as multiple solution curves. It can also handle problems with bounded variables. ACGA has been applied for the solution of two problems that are of great importance in the engineering field in order to demonstrate its efficiency. The problems include the Cartesian path generation of robot manipulators and the second- order, two-point boundary-value problem. The novel application of the algorithm to these problems possesses several advantages. First, it guarantees the smoothness of the solution curves. Second" the results obtained using ACGA are found in excellent agreement with the analytical solutions. Third, it can be applied to linear and nonlinear problems without any modification in the algorithm. Fourth, when applied on parallel computers, real time implementation is possible. A detailed convergence analysis of the proposed algorithm has been also carried out in order to find its optimum working environment. The analysis indicates that the algorithm works best with rank-based selection scheme, relatively high crossover and mutation probabilities, generational replacement and moderate population sizes. Regarding the performance enhancement scheme, it has been found that the deterministic mutation center and the knowledgeable dispersion factor result in drastic improvements in the convergence speed. A comparative study of ACGA and conventional GA has been performed. It has been found that the solution curves obtained using the conventional GA are of highly oscillatory nature. In addition to that, results have shown that the convergence speed of ACGA is much superior to that of conventional GA. ACGA has also shown much less se!1sitivity to problem-related parameters than conventional GA.

Item Type:Thesis (PhD)
Uncontrolled Keywords:Robot, Genetic Algorithms, Cartesian Path Generation, Advanced continuous genetic, Algorithm, Algorithm, Gutowsks Glgorithm, Convergence analysis
Subjects:Engineering & Technology (e) > Engineering(e1) > Systems engineering (e1.29)
ID Code:571
Deposited By:Mr. Muhammad Asif
Deposited On:19 Sep 2006
Last Modified:04 Oct 2007 21:01

Repository Staff Only: item control page