On the Dynamics of Evolutionary Multi-Objective Optimization


In evolutionary multi-objective optimisation (EMOO) using dynamic weighted aggregation (DWA), very interesting dynamic behaviours of the individuals have been observed [7] [8]. In this paper, the dynamics of the individuals on fitness space (FS) during multi-objective optimisation (MOO) using evolution strategies (ES) is studied by investigating the mapping of a normal distribution in the parameter space (PS) onto the FS. It is found that the movement of the individuals on the FS is strongly dependent on the characteristics of the projected distribution. Results on three test functions are given to show the good agreement of dynamics predicted from theoretical calculation with that observed in MOO using DWA.