L-SHADE with an adaptive penalty method of balancing the objective value and the constraint violation.

In the constraint-handling techniques, the penalty approaches (especially the adaptive penalty methods) are simple and flexible, and have been combined with various Evolutionary Algorithms so far. In this paper, we propose a new adaptive penalty method combined with L-SHADE as a method to optimize the 28 benchmark problems provided for the GECCO 2019 Competition on constrained single-objective numerical optimization effectively. The penalty factor is adjusted based on the trade-off information between the objective function value and the constraint violation that can be taken by individuals, the ranges of the objective function value and the constraint violation that can be taken by individuals and the proportion of feasible individuals in the current population. By doing this, the proposed method balances the objective function value and the constraint violation, and population is not converged in the only direction in which one improves and the other becomes worse. In addition, we use a few parameters that are easy to set up.