A Real-Coded Predator-Prey Genetic Algorithm for Multiobjective Optimization


This paper proposes a real-coded predator-prey CA for multiobjective optimization (RCPPGA). The model takes its inspiration from the spatial predator-prey dynamics observed in nature. RCPPCA differs itself from previous similar work by placing a specific emphasis on introducing a dynamic spatial structure to the predator-prey population. RCPPGA allows dynamic changes of the prey population size depending on available space and employs a BLX-alpha crossover operator that encourages a more self-adaptive search. Experiments using two different fitness assignment methods have been carried out, and the results are compared with previous related work. Although RCPPGA does not employ elitism explicitly (such as using an external archive), it has been demonstrated that given a sufficiently large lattice size, RCPPGA can consistently produce and maintain a diverse distribution of nondominated optimal solutions along the Pareto-optimal front even after many generations.