Role division approach for firefly algorithm based on t-distribution perturbation and differential mutation

Aiming at the problems of premature convergence and insufficient diversity of multi-objective firefly algorithm, this paper proposes a role division approach for firefly algorithm based on t-distribution perturbation and differential mutation. The idea of role division in nature is integrated into the firefly algorithm, and different roles are assigned to fireflies with different performances by the role division index, and the best learning mode is assigned according to the different roles, so as to realize the diversified learning of the population. The t-distribution perturbation with different degrees of freedom parameters is used instead of the original random perturbation, which can dynamically adjust the development and exploration ability of the algorithm in different periods. To avoid the algorithm falling into local optimality due to individual convergence at a later stage, differential mutation of the global optimal solution is performed to reduce the probability of the algorithm falling into stagnation and to balance convergence and diversity of the population. MOFA-PD is compared with 5 classical and 12 recent multi-objective optimization algorithms on 18 test functions, and the experimental results show that MOFA-PD has better advantages in convergence and diversity.