Research On Mode Conversion Genetic Algorithm In Mode Division Multiplexing

Recently, the transmission capacity of single-mode fiber has been greatly improved thanks to various multiplexing technology such as wavelength division multiplexing, polarization multiplexing, frequency division multiplexing, etc., approaching the non-linear Shannon limit. However, with a variety of bandwidth-consuming services are emerging, the growth rate of demand for bandwidth is much larger than the growth rate of the single-mode fiber bandwidth. It’s high time that new multiplexing technology which can further improve the transmission capacity of optical fiber transmission system should be studied. Mode division multiplexing which has become a research hotspot, utilizes multi-mode fiber to transmit information in different modes. It greatly improves the transmission capacity of the fiber. In this paper, we will focus on the mode conversion which is the key technology of mode division multiplexing, and apply the global optimization algorithms-genetic algorithm into it.The main contents of this paper can be divided into the following three parts:The first part is improvement of the traditional genetic algorithm. We adopt the matrix coding which is more conducive to reflect the inherent structure of issue. What’s more, we introduce the simulated annealing algorithm to the matrix-encoded genetic algorithm on the mutation, not only remedied the shortcoming of the genetic algorithm, but also improved the convergence rate of the genetic algorithm.The second part, we applied the improved genetic algorithm to the mode conversion of the mode division multiplexing system, and simulated the mutual conversion between low-level modes. In addition, we compared it with the other global optimization algorithm-simulated annealing algorithm.The third part, taking mode conversion characteristics and the supporting hardware into consideration, we tried to improve the mode conversion efficiency of genetic algorithms from the following two aspects: on one hand, we introduced the multi-threaded so that the evolution of genetic algorithms can be processed concurrently, as a result, the conversion efficiency was improved greatly; on the other hand, GPU computing was adopted to further shorten the time-consuming conversion.