Self-learning optical system based on a genetic-algorithm driven spatial light modulator

We demonstrate the applicability of a genetic algorithm (GA) to control the focus of an adaptive optical system using a liquid crystal spatial light modulator. The optical setup and the algorithm applied are set to fitness type reinforcement for learning. The particular GA developed optimizes the phase shifts in 32 independent pixels, and is biased towards approximating continuous functions that suit the focusing problem. The learning process is demonstrated to work reliably even in the presence of experimental noise.