Automatic Optical Fiber Alignment System Using Genetic Algorithms

We propose and demonstrate an automatic optical fiber alignment system using genetic algorithms. Connecting optical fibers is difficult because the connecting edges should be aligned with sub-micronmeter resolution. It, therefore, takes long time even for a human expert. Although automatic fiber alignment systems are being developed, they cannot be used practically if the degrees of freedom of fiber edges are large. To overcome this difficulty, we have developed an automatic fiber alignment system using genetic algorithms, which incorporate a special local learning method. In experiments, fiber alignment of five degrees of freedom can be completed within a few minutes, whereas it would take a human expert about half an hour.