Application of variable search space genetic algorithms to fine gain tuning of model-based robotic servo controller

摘要

In this paper, genetic algorithms with a variable search space function are proposed for fine gain tuning of a resolved acceleration controller which is one of model-based robotic servo controllers. Genetic algorithms proposed in this paper have a variable search space function which is activated if the optimal solution is not updated for a fixed number of generations. The function is terminated when the optimal solution is updated, or if the optimal solution is not updated within certain generations. This proposed method is evaluated through a trajectory following control problem in simulation. Simulations for sine curve trajectories are conducted using the dynamic model of the PUMA560 manipulator. The result shows the improvement of optimal solution and its convergence.