Investigating Predator-Prey Algorithms for Multi-Objective Optimization

This paper describes the GA model using a new selection method inspired by predator-prey interactions. In this model, prey, which represents the decision space vector, will be placed on the vertices of a two-dimensional lattice. Predator, which deals with objective functions, will also be placed on the same lattice randomly. Basic algorithm proposed by Professor Hans-Paul Schwefel and reported in Laumanns et al. (1998) and the modifications on it are discussed here. Thereafter, we propose a number of modifications to the basic model and apply the final algorithm to standard two and three-objective optimization problems. The predator and prey approach is also used to find preferred Pareto-optimal solutions (preys) corresponding to user-supplied reference points (treated as predators). This study should encourage further use of predator-prey approaches to multi-objective optimization.