Multimodal and constrained optimization in artificial immune system

Scope and Method of Study. This paper proposes two algorithms in artificial immune system. Firstly, a multimodal artificial immune system is proposed that uses vaccines that are extracted from the decision space. The vaccine enhances diversity of individuals in the antibody population and promotes the exploration power of the algorithm. Furthermore, the algorithm provides good exploration of decision space and a population that decreases finally to the number of available peaks. The proposed algorithm reduces computation time by avoiding re-exploration of already explored regions.;Secondly, this paper proposes a new constraint handling technique for artificial immune system by employing a directed mutation process to infeasible individuals. Through this design, the decision space is divided into sub regions to extract vaccines as proposed in the previous work. In feasible regions the objective functions are used to sort antibodies and in the infeasible region the extent of constraint violation as well as the objective value is used to sort the antibodies. Better individuals are then used to direct the search process towards feasible regions. The proposed method is simple to implement and does not need any parameter tuning.;Findings and Conclusions. The performance of each proposed algorithm is tested on several benchmark problems and both algorithms have shown competitive and in some cases better results when compared to similar work by other researchers.