Scheduling by Genetic Local Search with Multi-Step Crossover

In this paper, multi-step crossover (MSX) and a local search method are unified as a single operator called MSXF. MSX and MSXF utilize a neighborhood structure and a distance measure in the search space. In MSXF, a solution, initially set to be one of the parents, is stochastically replaced by a relatively good solution in the neighborhood, where the replacement is biased toward the other parent. After a certain number of iterations of this process, the best solution from those generated is selected as an offspring. Using job-shop scheduling problem benchmarks, MSXF was evaluated in a GA framework as a high-level crossover working on the critical path of a schedule. Experiments showed promising performance for the proposed method.