An Overview of Multi-population Methods for Dynamic Environments

Dynamic optimization problems (DOPs) can be found almost everywhere, from ship navigation at sea (Michalewicz et al. 2007) to aerospace design (Mack et al. 2007). In general terms, all aspects of science and engineering include the optimization of a set of complex problems, in which the objectives of the optimization, some restrictions, or other elements may vary over time. Since exact algorithms are impractical in dynamic environments, stochastic optimization techniques have gained much popularity. Among them, evolutionary computation (EC) techniques have attracted a great deal of attention due to their potential for solving complex optimization problems. Nevertheless, EC methods should undergo certain adjustments to work well when applying on DOPs. Diversity loss is by far the most severe challenge to EC methods in DOPs. This issue appears due to the tendency of the individuals to converge to a single optimum. As a result, when the global optimum is shifted away, the number of function evaluations (FEs) required for a partially converged population to relocate the optimum is quite harmful to the performance. In this chapter, we provide an overview of the multi-population methods for dynamic environments.