A new differential evolution for discontinuous optimization problems

This paper presents a stochastic method based on the differential evolution (DE) to address a wide range of discontinuous optimization problems such as scheduling and multi-item inventory control. DE is a novel evolutionary approach capable of handling non-differentiable, nonlinear and multi-modal objective functions. Previous studies have shown that DE is an efficient, effective and robust evolutionary algorithm, but it is not used to solve discontinuous problems. A novel solution encoding mechanism is used to handle discrete variables. In order to improve its search efficiency, a local search procedure is designed. Its performance is tested by some well known benchmark problems. Finally, it is used to solve a multi-item inventory problem which is a complicated mixed integer nonlinear optimization with complex constraints. Numerical results shown the useful of our method.