Optimizing Adjustable Parameters of Servo Controller by Using UniNeuro-HUDGA for Laser-Auto-Focus-Based Tracking System

This paper aims to minimize the tracking error of a laser auto-focus system that in developing treatment, due to uncertainty setting and modeling of its control system. The error is derived from the imperfect response to the standardized object reference. Optimizing procedure is obtained via multi-variable parameters by using a UniNeuro-hybrid uniform design genetic algorithm (HUDGA). In general, the parameter setting of a servo-controller is determined by some complex analysis or the trial-and-error of an expert person; when the controlled model is distinctly undefined, the process requires considerable time. The UniNeuro-HUDGA requires only 40 experiments to be conducted in the uniform design (UD) of building the metamodel via a neural network (UniNeuro), which is used as the fitness function in the optimization procedure by combining a genetic algorithm with UD. UD is then embedded in the HUDGA for initializing and enriching the solution set, whereas chromosomes used in crossover and mutations generated by UD chromosomes are individually conveyed using a selection procedure combined with the Euclidean distance; then, the optimized setting has investigated by the equipment. This paper concludes that the proposed algorithm optimizes the adjustable parameters of a servo-controller and outperforms the trial-and-error of an expert person.