A Multi-population Schema Designed for Biased Random-Key Genetic Algorithms on Continuous Optimisation Problems

In Evolutionary Algorithms, population diversity is a determinant factor for the quality of the final solutions. Due to diverse problem characteristics, many techniques face difficulties and converge prematurely in local optima. The maintenance of diversity allows the algorithm to explore the search space and efficiently achieve better results. Parallel models are well-known techniques to maintain population diversity; however, design choices lead to different characteristics for the optimization process. For instance, the migration policy on the Island model can control how fast the algorithm converges. This work proposes a new migration policy designed for the Biased Random-Key Genetic Algorithm (BRKGA). Also, the proposal is compared with two traditional strategies and evaluates its performance in continuous search spaces. The results show that the proposal can improve the BRKGA optimization capability with suitable parameters.