A knee point-driven multi-objective artificial flora optimization algorithm

In recent days, swarm intelligent (SI) optimization algorithms have been proved to be a powerful framework for finding tradeoff solutions of multi-objective optimization problems (MOPs). Many researchers have proposed various SI optimization algorithms. Multi-objective artificial flora (MOAF) optimization algorithm is a recently proposed algorithm for solving MOPs. However, problems of decreased population diversity and uniformity of solutions distribution in the late evolutionary period is existed in the algorithm. Hence, this paper proposes a knee point-driven MOAF (kpMOAF) optimization algorithm to address the vulnerability of MOAF optimization algorithm. Knee points of the non-dominant solutions are taken by the proposed algorithm as criterion to guide the population evolution. Researchers have proved that select knee points equals to select a large hypervolume. Therefore, using it as criterion is an effective way to enhance the population convergence rate and maintain the diversity of solutions. In addition, adaptive neighborhood control method is introduced in the evolution process to improve the algorithm development capability. Simulation results on 10 benchmark functions demonstrate the competitiveness of kpMOAF optimization algorithm.