COLSHADE for Real-World Single-Objective Constrained optimization Problems
Javier Gurrola-Ramos, Arturo Hernández-Aguirre, Oscar Dalmau
Abstract
In this paper we present the COLSHADE algorithm for real parameter constrained optimization problems. COLSHADE evolved from the basic L-SHADE algorithm by introducing significant features such as adaptive Lévy flights and dynamic tolerance (included in the constraint handling technique). Lévy flights mainly perform the exploration phase in algorithms such as the Firefly algorithm and Cuckoo search; in COLSHADE, however, the goal of the Lévy flights is to administer the selection pressure exerted over the population as to find the feasible region and keeping diversity. Thus, a new adaptive Lévy flight mutation operator is introduced here and called the levy/1/bin. In many problems the levy/1/bin excels during the exploration phase whilst the exploitation phase is performed by current-to-pbest mutation. However, the adaptive strategy propitiates the emergence of these two mutation approaches at different rates and times. The proposed method is tested on 57 constrained optimization functions of the benchmark provided for the CEC 2020 realworld single-objective constrained optimization competition.