Automatic Heuristic Generation with Scatter Programming to Solve the Hybrid Flow Shop Problem

Abstract

The hybrid flow shop (HFS) problem is solved in this study through composite dispatching rules (CDRs) generated automatically. CDR is a simple approach to solve the HFS problem rapidly and easily. Genetic programming (GP) is the most common methodology to generate CDRs to solve combinational optimization problems. The disruption in the main operations of GP, such as crossover and mutation, is high. Scatter programming is proposed in this paper to generate CDRs with improved scalability and flexibility for HFS problems. The proposed algorithm includes a novel local improvement method, one-point traversal shaking search, to accelerate convergence. Simulation studies demonstrate that the proposed algorithm outperforms genetic programming and scatter programming with shaking. The dispatching rules generated automatically can also be reused in similar instances and yield better results than existing dispatching rules. The rate of convergence of the proposed algorithm is higher than that of scatter programming with shaking.