Dynamic Modelling of Flexible Manipulator System Using Genetic Algorithm

Abstract

Flexible robotic manipulators pose various challenges in modelling, design, structural optimisation and control. This paper presents investigations into practical dynamic modelling of a flexible manipulator system using genetic algorithm (GA). Conventional genetic algorithms (GAs) often converge prematurely to a suboptimal region and fail to provide effective solutions due to lack of diversity in the population set as the algorithm proceeds. In order to improve and maintain diversity in the population set, a relatively new variant of GA, namely, fitness sharing based replacement genetic algorithm (FSR-GA¹) is employed where some individuals are replaced periodically based on a fitness sharing method. The algorithm is utilised to extract dynamic model of 1-DOF (degree of freedom) motion of a flexible manipulator system. A comparative assessment between FSR-GA and conventional GA is presented in the same application to highlight the novelty of the used GA. Results show that the FSR-GA significantly improves the searching capability of the optimisation process compared to conventional GA. Time domain and frequency domain results clearly reveal the potential of the proposed method in modelling flexible manipulator systems.