Job Shop Scheduling Benchmark: Environments and Instances for Learning and Non-learning Methods

TL;DR

This work presents a unified, open-source benchmarking platform for job shop scheduling problems (JSP, FSP, FJSP, AJSP) and variants with sequence-dependent setups and online arrivals. It provides a modular environment with five components (configurations, data, scheduling environment, solution methods, visualization) and supports a spectrum of methods from exact solvers (CP-SAT, MILP) to heuristics and learning-based approaches (L2D, FJSP-DRL, DANIEL). The paper details problem formulations via disjunctive graphs, a flexible JSON-based API for custom instances, and extensive experiments on public benchmarks, highlighting CP-SAT’s strong performance and the competitive potential of learning-based and heuristic methods. The resource aims to standardize evaluations, enable fair comparisons, and spur the development of next-generation scheduling algorithms with broad applicability in industry and research.

Abstract

Job shop scheduling problems address the routing and sequencing of tasks in a job shop setting. Despite significant interest from operations research and machine learning communities over the years, a comprehensive platform for testing and comparing solution methods has been notably lacking. To fill this gap, we introduce a unified implementation of job shop scheduling problems and their solution methods, addressing the long-standing need for a standardized benchmarking platform in this domain. Our platform supports classic Job Shop (JSP), Flow Shop (FSP), Flexible Job Shop (FJSP), and Assembly Job Shop (AJSP), as well as variants featuring Sequence-Dependent Setup Times (SDST), variants with online arrivals of jobs, and combinations of these problems (e.g., FJSP-SDST and FAJSP). The platfrom provides a wide range of scheduling solution methods, from heuristics, metaheuristics, and exact optimization to deep reinforcement learning. The implementation is available as an open-source GitHub repository, serving as a collaborative hub for researchers, practitioners, and those new to the field. Beyond enabling direct comparisons with existing methods on widely studied benchmark problems, this resource serves as a robust starting point for addressing constrained and complex problem variants. By establishing a comprehensive and unified foundation, this platform is designed to consolidate existing knowledge and to inspire the development of next-generation algorithms in job shop scheduling research.

Figures (7)

• Figure 1: Disjunctive Graph Representation of a Job Shop Scheduling Problem (JSP). Left Panel: This illustrates a 3 (jobs) $\times$ 3 (machines) JSP instance. Black arrows indicate conjunctive arcs, enforcing precedence constraints among operations within the same job. Dotted lines represent disjunctive arcs, whose directions need to be determined. Disjunctive arcs (or operation nodes) sharing the same color correspond to operations requiring the same machine. Right Panel: Represents a feasible solution (best viewed in color).
• Figure 2: Disjunctive Graph Representation of a Flow Shop Scheduling Problem (FSP). This illustrates a 3 (jobs) $\times$ 3 (machines) FSP instance. All jobs follow the same fixed machine order, with same-colored operation nodes representing operations to be scheduled on the same machine.
• Figure 3: Disjunctive Graph Representation of a Flexible Job Shop Scheduling Problem (FJSP). This illustrates a 3 (jobs) $\times$ 3 (machines) FJSP instance. Operations, represented by gray nodes, can be assigned to different machines, allowing alternative routing options. Disjunctive arcs reflect both sequencing and machine assignment decisions, increasing scheduling complexity.
• Figure 4: Disjunctive Graph Representation of an Assembly Job Shop Scheduling Problem (AJSP). This illustrates a 4 (jobs) $\times$ 3 (machines) AJSP instance. The first three jobs follow standard processing constraints, while the fourth job depends on their completion, enforcing an additional assembly precedence structure. Black arrows indicate the conjunctive arcs, representing both within-job precedence constraints and final assembly dependencies.
• Figure 5: Disjunctive Graph Representation of a scheduling problem with Sequence-Dependent Setup Times (SDST). This illustrates a 3 (jobs) $\times$ 3 (machines) scheduling instance where disjunctive arcs not only determine operation sequencing on shared machines but also encode sequence-dependent setup times.