A Constraints Scattered Memetic Algorithm for Constrained Optimization Problem

In this paper, a constraints scattered memetic algorithm (CSMA), which integrates a novel constraints scattered genetic algorithm (CSGA) and the traditional interior point method, is proposed for solving constrained optimization problems. In CSGA, a constraint scattering operation, a sub-population crossover method and a new population performance evaluation mechanism are employed. The complete constraints of a problem are divided into several sub-populations and all these subpopulations are crossed after their respective evolution process. In this way, the difficulty of obtaining feasible individuals in many strong constrained conditions is well overcome. And according to the newly defined population performance index, individuals with larger population diversity are chosen for further local search. These new mechanisms are combined in CSGA and interior point method is further employed as a local search operator for exploitation. Experiments and comparisons over a set of standard test functions show that population with better performance can be generated during the iteration of CSGA and the proposed CSMA has a better solution precision at less computation cost than most of the other algorithms reported in literature.