LLM-Assisted Automatic Dispatching Rule Design for Dynamic Flexible Assembly Flow Shop Scheduling

TL;DR

This work tackles online scheduling for dynamic FAFSPs with dual kitting constraints and multi-product delivery by introducing LLM4DRD, an LLM-assisted dynamic rule design framework. It couples an algorithm-design LLM (LLM-A) with a schedule-evaluation LLM (LLM-S) in a hybrid, elite-initialized evolutionary loop, featuring a heterogeneous graph MDP representation and a co-evolution of design knowledge and rules. The approach uses elite-knowledge initialization, a hybrid evaluation that combines objective tardiness with expert feedback, and prompts that realize crossover/improvement operators, enabling continuous adaptation to dynamic features. Empirical results on real and synthetic instances show LLM4DRD reduces average tardiness by up to 3.17–12.39% against strong baselines and up to 11.10% against the second-best competitor across 24 scenarios, demonstrating robust generalization and practical potential for intelligent, online dispatching in complex multi-stage manufacturing systems.

Abstract

Dynamic multi-product delivery environments demand rapid coordination of part completion and product-level kitting within hybrid processing and assembly systems to satisfy strict hierarchical supply constraints. The flexible assembly flow shop scheduling problem formally defines dependencies for multi-stage kitting, yet dynamic variants make designing integrated scheduling rules under multi-level time coupling highly challenging. Existing automated heuristic design methods, particularly genetic programming constrained to fixed terminal symbol sets, struggle to capture and leverage dynamic uncertainties and hierarchical dependency information under transient decision states. This study develops an LLM-assisted Dynamic Rule Design framework (LLM4DRD) that automatically evolves integrated online scheduling rules adapted to scheduling features. Firstly, multi-stage processing and assembly supply decisions are transformed into feasible directed edge orderings based on heterogeneous graph. Then, an elite knowledge guided initialization embeds advanced design expertise into initial rules to enhance initial quality. Additionally, a dual-expert mechanism is introduced in which LLM-A evolutionary code to generate candidate rules and LLM-S conducts scheduling evaluation, while dynamic feature-fitting rule evolution combined with hybrid evaluation enables continuous improvement and extracts adaptive rules with strong generalization capability. A series of experiments are conducted to validate the effectiveness of the method. The average tardiness of LLM4DRD is 3.17-12.39% higher than state-of-the-art methods in 20 practical instances used for training and testing, respectively. In 24 scenarios with different resource configurations, order loads, and disturbance levels totaling 480 instances, it achieves 11.10% higher performance than the second best competitor, exhibiting excellent robustness.

Figures (9)

• Figure 1: Dynamic PDR automated design for FAFSP based on LLM4DRD framework.
• Figure 2: Automated design and evolution process for dynamic PDR.
• Figure 3: Design strategies and prompt engineering.
• Figure 4: Offline design and online application process of LLM4DRD.
• Figure 5: The box plot of tardiness performance for ablation experiment.