Enhancing heuristics for order acceptance and scheduling using genetic programming

Abstract

Order acceptance and scheduling (OAS) in make-to-order manufacturing systems is a NP-hard problem for which finding optimal solutions for problem instances can be challenging. Because of this, several heuristic approaches have been proposed in the literature to find near-optimal solutions to OAS. Many previous heuristic approaches are very effective, but require careful design and developing new heuristics can be difficult. Genetic Programming (GP) has been used to generate reusable and efficient heuristics in OAS and shows promising results. However, in terms of solution quality, the evolved heuristics are still less competitive as compared to highly customised heuristics designed by human experts. To overcome these limitations, this paper proposes two new Particle Swarm Optimisation (PSO) approaches to OAS. Afterwards, GP evolved rules are combined with an existing Tabu Search (TS) heuristic and with the proposed PSO algorithms as hybrid approaches to OAS. The experimental results show that these PSO approaches are competitive with effective heuristics such as TS. In addition, TS heuristic greatly benefits from evolved rules, whereas PSO approaches do not benefit.