Genetic Algorithm Based On A Variety Of Groups

Many applications are dynamoic optimization problems and multi-objective problems in practice. However, traditional mathematical methods can't deal with problems which are generally uncontinuous, derivative-free and so on, and each run can only find one solution. Therefore, evolutionary algorithms are studied in this paper to solve dynamoic optimization problems and multi-objective problems. These algorithms all pay attentions to diversity reservation of their populations. So the algorithms can trace the moving optima in dynamic environment or find the diverse Pareto solutions in multi-objective optimization problems. Multiple populations are employed in this paper to keep population diversity and these populations evolve in different ways. The main research topics and contributions are as follows:(1) Evolutionary algorithms are reviewed firstly and then genetic algorithmis detailed, including their origin, operators, evolutionary process and related theories. The progress of dynamoic optimization problems and multi-objective problems are analyzed specially.(2) A multi-population genetic algorithm (MPGA) is proposed for solving dynamoic optimization problems. The proposed algorithm employs two populations with different evolutionary process and exchanges their best individuals at checkpoint. MPGA can convergence fast and trace the moving optima in dynamic environment.(3) After studying the nature of multi-objective optimization and its research progress, a multi-objective genetic algorithm based on multi-population andε-dominance is proposed. Simulation experiments on several multi-objective optimization benchmarks demonstrate that the proposed algorithm can solve various multi-objective optimization problems and keep diversity of the Pareto solutions.