Adaptive archive-based multifactorial evolutionary algorithm for constrained multitasking optimization

Abstract

Evolutionary multitasking has attracted much attention in the field of evolutionary computing. Most of the existing multitasking evolutionary algorithms aim at solving unconstrained multitasking optimization problems. The study on constrained multitasking optimization problems is scarce. However, in practical applications, lots of optimization problems contain constraints. In this paper, an adaptive archive-based multifactorial evolutionary algorithm is proposed to solve constrained multitasking optimization problems. First, an archiving strategy is proposed to store infeasible solutions with better objective function values. With this strategy, useful information on infeasible solutions can be exploited to accelerate the convergence rate. Second, the random mating probability is adjusted through an adaptive strategy to facilitate positive knowledge transfer. Finally, a new mutation strategy is proposed to promote convergence by mutating some random individuals and replacing the individuals with the largest constraint violation. By comparing existing constrained multitasking evolutionary algorithms and some constrained single-task evolutionary algorithms, the results reveal the effectiveness of the proposed algorithm in solving constrained multitasking optimization problems.