Benchmarking Constraint Programming software: Insights from the Job Shop Scheduling Problem

Abstract

In this article, we benchmark three popular constraint programming (CP) solvers for the job shop scheduling problem (JSSP) under identical hardware and computation-time conditions, with the objective of minimizing the makespan. For evaluation purposes, we use 80 classic Taillard instances. Overall, CP Optimizer proved to be the most competitive solver, demonstrating optimality in 38 instances, despite a slightly higher average optimality gap (3.06%). OR-Tools followed closely, achieving a marginally smaller average gap of 2.86%, proving optimality in 23 instances, and consistently producing competitive lower bounds. In third place, Hexaly exhibited improved performance as the instance size increased.