Usefulness of lethal chromosomes in genetic algorithms solving the constrained optimization problems

Yalong Zhang¹, Hisakazu Ogura², Xuan Ma³

¹College of Electrical and Information Engineering, Quzhou University, Quzhou 324000, China

²Graduate School of Engineering, University of Fukui, Fukui 910-8507, Japan

³Faculty of Automation and Information Engineering, Xi'an University of Technology, Xi'an 710048, China

The infeasible solutions are often generated in population as evolutionary computation solving the combinatorial optimization prob­lems. The number of infeasible solutions impacts the performance of the evolutionary computation searching the optimal solution,in the worst case the algorithm ceases to run. In genetic algorithms, encoding of infeasible solutions is referred to as lethal chromosomes. In this study, we discover a propertyof lethal chromosomes that: although lethal chromosomes carry out the infeasible solutions in genetic algorithms, their statisticalproperty implies an underlying similarity with the exact solution of the optimization problems. Hereby we propose an operation using statisticalproperty of lethal chromosomes to handle with the lethal chromosomes themselves. Simulation experiments on a large number of test cases demonstrated that it can improves obviously the performance of genetic algo­rithms to use the statisticalproperty of lethal chromosomes.