A Multi-objective Genetic Algorithm with Relative Distance: Method, Performance Measures and Constraint Handling

Abstract

A novel multi-objective evolutionary algorithm (MOEA), called multi-objective genetic algorithm with relative distance (MOGARD) is described. A novel relative distance parameter that ensures convergence to the Pareto optimal front and a nearest neighbour based method for maintaining diversity in the non-dominated set is used. Two novel performance measures are formulated to estimate the performance of the MOEAs. A penalty based constraint handling concept is introduced in MOGARD, for handling constraints. Experimental results demonstrate the superiority of MOGARD on several test problems, as compared to other recent and well known algorithms