A new dynamic multi-objective optimization evolutionary algorithm

Dynamic multi-objective optimization problem (DMOP) often involves incommensurable, competing and varying objectives with time (environment), and the number of their Pareto optimal solutions is usually infinite, thus how to find a sufficient number of uniformly distributed and representative Pareto optimal solutions at any en vironment for the decision maker is very important. In this paper, we divide the time period of DMOP into several equal subperiods. In each subperiod, the DMOP is approximated by a static multi-objective optimization problem (SMOP). Furthermore, the static rank variance and the static density variance of the population are defined, by using the static rank variance and the static density variance of the population as two objectives, each SMOP is further transformed into a bi-objective optimization problem. As a result, the original DMOP is approximately transformed into several static multi-objective optimization problems. Thereafter, an environment changing feedback operator which can automatically check out the environment variation is proposed and an improved non-uniform mutation operator with quantization is. designed. Based on these, a new dynamic multi-objective optimization evolutionary algorithm (denoted by DMEA) is proposed. The comparative study shows that DMEA is more effective and can find better solution set in environment-varying than the compared algorithms can in terms of convergence, diversity, and the distribution of the obtained Pareto optimal solutions.