This work transfers knowledge in the reverse direction, using knowledge extracted from high-quality GP rules to guide an online LLM decision maker, and injects it through Feature Selection, Feature Hint, Rule Reference, and Rule Follow.
Abstract
In dynamic multi-mode project scheduling, activities have alternative execution modes and uncertain durations, while precedence relations and limited resources constrain their execution. Heuristic priority rules support fast online decisions, but their design requires substantial domain expertise. Genetic programming (GP) hyper-heuristics can automatically evolve such rules. Large language models (LLMs), meanwhile, provide a flexible interface for interpreting scheduling information and explaining decisions. However, zero-shot LLM decisions may lack domain knowledge, consume many tokens, and vary across repeated queries. GP-evolved rules therefore provide a potential source of scheduling knowledge for guiding LLM decisions. Unlike existing LLM--GP hybrids that use LLMs to support heuristic evolution, we transfer knowledge in the reverse direction, using knowledge extracted from high-quality GP rules to guide an online LLM decision maker. We extract knowledge from high-quality GP rules and inject it through Feature Selection, Feature Hint, Rule Reference, and Rule Follow. These mechanisms are evaluated in terms of scheduling performance, token consumption, decision stability, and the feature focus expressed in generated rationales. GP-derived guidance generally improves the unguided LLM, but its representation matters. Simplifying the decision context or supplying explicit decision logic is more effective than highlighting important features. Feature Selection offers the best token efficiency, whereas Rule Follow achieves strong performance at greater token cost. Guidance also improves decision stability and changes the features expressed in generated rationales.
Experimental results demonstrate that the proposed framework consistently outperforms classical ALNS baselines and several competitive metaheuristic methods in terms of solution quality and convergence speed.
T. M. Nguyen· Annual Conference on Genetic...· 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
Results show that simulation traces can guide targeted code-level policy improvement in complex simulation-based scheduling in dynamic production and automated guided vehicle (AGV) scheduling.
This study addresses the grey dynamic flexible job shop scheduling problem (GDFJSP), in which jobs with uncertain grey-number processing times arrive stochastically and must be dispatched in real time. It aims to develop a genetic programming algorithm that evolves interpretable heuristic dispatching rules while handling stochastic arrivals and iterative grey-time updating efficiently.
A memory-guided adaptive feature genetic programming (MGAFGP) algorithm is proposed with a dual-tree encoding for routing and sequencing decisions. The algorithm combines parallel simulation for concurrent fitness evaluation, an elite-memory-guided strategy with separate feature probability vectors for routing and sequencing trees, and a generation-dependent parent selection function. Its performance is evaluated across multiple scenarios defined by different objectives, utilization levels, and due-date tightness conditions.
MGAFGP reaches high-quality rules substantially faster than standard GP under the tested scenarios, showing corrected significant advantages during early evolution and reaching GP's full-budget mean performance with a substantially smaller iteration budget. No corrected full-budget comparison favours standard GP. The evolved rules outperform classical heuristic combinations after independent test re-evaluation, while feature-use patterns, symbolic expressions, and tree-complexity statistics show that the resulting dispatching logic remains inspectable.
The approach provides a computationally tractable way to discover interpretable dispatching rules for dynamic manufacturing environments with uncertain processing times and limited historical data. By reaching strong rules earlier, MGAFGP can reduce the simulation budget needed for rule evolution, support managerial inspection of scheduling logic and reduce reliance on expert-designed heuristics.
The study integrates generalized grey-number processing times into the DFJSP and develops a GP algorithm with separate feature probability adaptation for routing and sequencing. The elite-guided strategy with generation-dependent parent selection provides a mechanism for accelerating convergence in simulation-based GP under grey processing-time uncertainty.
Nanlei Chen, Bentao Su· Grey Systems Theory and Appl...· 0 citations
Space Station Operation Mission Planning (SSOMP) requires coordinated decisions across long-term activity allocation, mid-term logistics optimization, and short-term execution scheduling and is a key component of autonomous mission operations for high-precision space missions. Existing optimization methods have achieved substantial progress at individual planning levels, but their dependence on problem-specific models, limited support for semantic review of decision rationale, and computational cost restrict their adaptability to multi-level planning scenarios. This paper proposes a Large Language Model (LLM)-assisted framework for multi-level SSOMP. The framework combines Staged Fine-Tuning (Staged-FT), Reflective Constraint–Repair Prompting (RCRP), and LLM-Guided Evolutionary Variation (LGEV). Staged-FT uses a Cognitive-Load-Theory-informed curriculum with Low-Rank Adaptation to adapt general-purpose LLMs to SSOMP domain knowledge. RCRP couples a Deterministic Rule Engine with LLM-based semantic repair to improve hard constraint satisfaction. LGEV embeds the fine-tuned LLM into NSGA-III as a fitness-aware variation operator for multi-objective activity allocation. Three case studies are conducted on literature-derived benchmark scenarios of logistics optimization, emergency re-planning, and activity allocation with logistics design, corresponding to Flight Increment Planning, Short-Term Execution Planning, and Overall Operation Planning, respectively. Results show that Staged-FT produces solutions close to traditional algorithms, RCRP achieves full hard constraint satisfaction in the emergency re-planning and logistics planning cases, and LGEV reduces the convergence generations of NSGA-III while improving Pareto-front quality. The framework provides a constraint-aware approach with explicit reasoning traces that can support expert review of AI-assisted planning for autonomous space mission operations.
Large language model-based automated heuristic design (LLM-AHD) has shown strong potential in discovering effective heuristics for combinatorial optimization problems. However, existing methods primarily optimize a single heuristic, whereas practical optimization frameworks often rely on multiple interacting components. Directly extending single-heuristic methods is challenging because early component selection can overlook components with late potential, while independent evolution ignores inter-component dependencies. We propose MuEvo, an LLM-driven framework for evolving heuristic ensembles under ensemble-level feedback. MuEvo combines Dynamic Component Management, which uses short-budget probing and a reversible lifecycle to revise component priorities throughout the search, with LLM-Driven Co-Evolution, which coordinates component populations through Multi-Ensemble Evaluation, Cross-Component Information Sharing, Relation-Guided Pair Evolution, and Adaptive Budget Allocation. We evaluate MuEvo on selection hyper-heuristics and componentized ant colony optimization across four combinatorial optimization domains. Results show that MuEvo consistently improves human-designed frameworks and outperforms representative multi-component extensions of state-of-the-art LLM-AHD methods, demonstrating its effectiveness across both controller-mediated heuristic pools and functionally differentiated algorithmic components.
Haoze Lv, Ning Lu, Shengcai Liu et al.· 0 citations