Adaptive scheduling method based on large language model dynamically adjusting reward weight
By using a large language model and multi-agent collaborative decision-making, the reward weights of the flexible workshop are dynamically adjusted, solving the problem that static reward functions cannot adapt to dynamic production scenarios. This achieves adaptive scheduling with multi-objective optimization, improving the production efficiency and flexibility of the flexible workshop.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SOUTHWEAT UNIV OF SCI & TECH
- Filing Date
- 2026-06-02
- Publication Date
- 2026-07-17
AI Technical Summary
Existing flexible job shop scheduling methods cannot adapt to dynamically changing production scenarios due to the inability of static reward functions, leading to multi-objective optimization conflicts. Furthermore, manual optimization is costly and makes it difficult to achieve adaptive scheduling.
An adaptive scheduling method that dynamically adjusts reward weights using a large language model is adopted. By using disjunctive graph modeling, graph neural networks, and hierarchical reward function design, combined with multi-agent collaborative decision-making, the dynamic adaptive adjustment of the reward function is achieved, thereby optimizing multi-objective conflicts.
It achieves dynamic optimal trade-offs among multiple objectives in dynamic production scenarios, reduces manual tuning costs, improves the adaptability and convergence speed of the scheduling model, and adapts to multi-variety, small-batch production modes.
Smart Images
Figure CN122414719A_ABST