Enhancing Genetic Algorithms using a Dynamic Mutation Value Approach: An Application to the Control of Flexible Robot Systems

This paper presents an investigation into a new optimization technique based on genetic algorithm (GA). A dynamically-changed mutation value approach is introduced to increase the diversity in the search space and avoid premature convergence caused by simple genetic algorithm (SGA). The enhanced genetic algorithm (EGA) is used to tune the feedback gains of a PD controller which controls both the position and vibration of a single-link flexible arm. The dynamic model of the system is derived using Hamilton’s principle and modeled using the finite element method (FEM). A multi-objective function is defined and altered to reach a range of specified system responses and therefore it is shown to be able to satisfy different objectives. Adaptive Genetic Algorithm (AGA) and Cloud Model Based Adaptive Genetic Algorithm (CAGA) techniques are used to challenge the proposed technique. Results obtained show that EGA creates significant improvement in the speed of convergence compared to other techniques. Moreover, the obtained solutions are of higher average fitness values. EGA succeeded to consistently reach a global solution for an objective function that needs rigorous search mechanism which encourages for further application to various control problems, complex mathematical functions and real time applications.