Comparative Analysis of Metaheuristic Algorithms for Standard Dynamic Multiobjective Optimization Problems

Dynamic Multiobjective Optimization Problems (DMOPs) are those involving multiple conflicting objectives to be simultaneously optimized subject to a certain number of constraints. The main characteristic of DMOPs is that the objectives and/or constraints change over time. Due to the dynamic component, DMOPs are generally more complex and difficult to solve than static multiobjective optimization problems. This is why an important number of researchers have proposed several benchmarks containing DMOPs of different types. In this chapter, we conduct a comparative analysis of two prominent metaheuristic algorithms for DMOPs: DNSGA–II and DSPEA–II. We use the well–known FDA test suite with five instances for continuous search spaces. The instances in the FDA benchmark consider three types of changes. In the first type, the Pareto Optimal Set (POS) changes while the Pareto Optimal Front (POF) do not change. In the second type, both the POS and the POF change. Finally, in the third type, the POS does not change but the POF changes. The computational experiments revealed that DNSGA–II excels in Type 1 problems characterized by convex Pareto Optimal Fronts (POFs). In contrast, DSPEA–II demonstrated its suitability for solving problems falling into both Type 1 and Type 2 categories, where nonconvex POFs are prevalent. Nevertheless, despite the aforementioned findigs, the empirical evidence does not support the superiority of one over the other in general.