Single and Multi-Objective Dynamic Optimization: Two Tales from an Evolutionary Perspective

Many real-world optimization problems involve objectives, constraints, and parameters which constantly change with time. However, to avoid complications, such problems are usually treated as static optimization problems demanding the knowledge of the pattern of change a priori. If the problem is optimized in its totality for the entire duration of application, the procedure can be computationally expensive, involving a large number of variables. Despite some studies on the use of evolutionary algorithms in solving single-objective dynamic optimization problems, there has been a lukewarm interest in solving dynamic multi-objective optimization problems. In this paper, we discuss two different approaches to dynamic optimization for single as well as multi-objective problems. Both methods are discussed and their working principles are illustrated by applying them to different practical optimization problems. The off-line optimization approach in arriving at a knowledge base which can then be used for on-line applications is applicable when the change in the problem is significant. On the other hand, an off-line approach to arrive at an minimal time window for treating the problem in a static manner is more appropriate for problems having a slow change. Further approaches and applications of these two techniques remain as important future work in making on-line optimization task a reality in the coming years.