A modified genetic algorithm for single machine scheduling

In this paper we propose a modified genetic algorithm for the single machine scheduling problem with ready times. This algorithm improves the simple genetic algorithm by introducing two new steps: (1) a filtering step to filter out the worst solutions in each generation and fill in their positions with the best solutions of previous generations; and (2) a selective cultivation step to cultivate the most promising individual when no improvement is made for certain generations. Improvement is also made on the crossover operation for the problem. Computational experiments are carried out, comparing the performance of the proposed algorithm, the simple genetic algorithm and special purpose heuristics. The contribution of each modification measure to the performance improvement is also analyzed.