Dynamic multi-objective optimization evolutionary algorithm

A new evolutionary algorithm for Dynamic multiobjective optimization is proposed in this paper First, the time period is divided into several random subperiods. In each subperiod, the problem is approximated by a static multiobjective optimization problem. Thus, the dynamic multiobjective optimization problem is approximately transformed into several static multiobjective problems. Second, for each static multiobjective optimization problem, the expected rank variance and the expected density variance of the population are firstly defined. By using the expected rank variance and the expected density variance of the population, the dynamic multiobjective optimization problem is transformed into a bi-objective optimization problem. Third, a new evolutionary algorithm is proposed based on a new self-check operator which can automatically check out the time variation. At last, the simulation is made and the results demonstrate the effectiveness of the proposed algorithm.