Coordinating a team of robots in aircraft skin fabrication requires allocating and sequencing tightly coupled subtasks under spatio-temporal constraints, while the fleet must react to runtime disturbances such as robot failures and urgent task arrivals. Mixed-Integer Linear Programming (MILP) yields provably optimal coordination, but a disturbance often introduces new constraint logic rather than a mere parameter change, leaving the existing formulation structurally inadequate and requiring expert-led reformulation to accommodate the new logic. This letter proposes a Large Language Model (LLM)-based framework that automates the path from natural-language scheduling requirements to MILP formulation, executable solver code, and event-driven rescheduling, so that the optimization model can be restructured online without expert intervention. Two compact LLMs are specialized for complementary roles: a modeling LLM, empowered via knowledge augmentation, supervised fine-tuning on industrial constraint descriptions, and direct preference optimization on self-generated negatives; and a code LLM, trained via sandbox-validated knowledge distillation. On ten industrial constraint classes, the 8B modeling LLM reaches 100% constraint-level accuracy, exceeding cloud-scale general-purpose LLMs equipped with retrieval over the same knowledge base, while the 8B code LLM attains 86% under fully on-premises deployment. When a disturbance occurs, an event-triggered mechanism regenerates only the affected constraints. A case study on multi-robot scheduling for aircraft skin fabrication shows that the framework matches the schedule quality of baselines while shifting adaptation cost from offline expert labour to a bounded online inference budget.
Zhendong Chen, Mingming Peng, Hao Zhang et al.· IEEE Robotics and Automation...· 0 citations
The dynamic job shop scheduling problem (DJSSP) is critical for optimizing production efficiency in intelligent manufacturing systems under dynamic constraints. Traditional approaches, including heuristic dispatching rules (HDRs) and evolutionary hyper-heuristics, often struggle to generalize across dynamic and unseen scheduling scenarios. The advent of large language models (LLMs) offers new opportunities for scheduling optimization, yet their deployment in real-world manufacturing is constrained by data privacy concerns that limit the use of online LLM services and by the restricted inference capability of local small-scale models. To address these challenges, this paper proposes a novel feedback-driven self-evolution framework employing dual-stage knowledge distillation. The framework utilizes a feedback-driven refinement mechanism to distill expert knowledge from advanced LLMs, which is then used to fine-tune the Qwen2.5-Coder-7B model into ModouGPT, a specialized model for generating HDRs. Experimental results indicate that our approach performs better than traditional methods and five state-of-the-art LLM-based evolutionary frameworks, particularly in dynamic and unseen scheduling scenarios. ModouGPT demonstrates competitive performance with online LLMs in complex dynamic environments, while ensuring data privacy. This study provides manufacturing enterprises, particularly small and medium-sized enterprises, with a deployable scheduling system that balances intelligence, privacy, and computational efficiency in dynamic job shop environments. Note to Practitioners—DJSSP involves constant disruptions such as machine breakdowns and dynamic order arrivals. Deploying artificial intelligence (AI)-based scheduling in real manufacturing environments remains difficult: cloud-based LLM services risk exposing sensitive production data, while large-scale local models demand prohibitive hardware investments, placing both options out of reach for most enterprises. This paper presents a practical solution. The proposed framework distills scheduling knowledge from a powerful online teacher model into a compact 7B local model, ModouGPT, deployable on a single consumer-grade graphics processing unit (GPU). All scheduling decisions are made entirely on-premises, with no data leaving the factory floor. Practitioners do not need domain expertise to design HDRs. The system automatically generates HDRs tailored to the specific production environment, covering job processing characteristics and dynamic disturbances. An offline HDR library can be built in advance and updated periodically without interrupting production, and new HDRs can be synthesized in under one minute during live operations. This makes the system immediately actionable for engineers seeking to upgrade scheduling capabilities without relying on external LLM services or specialized algorithm designers.
Jin Huang, Zhengqi Shi, Qihao Liu et al.· IEEE Transactions on Automat...· 0 citations