A Large Decision Model for Autonomous and Interactive Manufacturing Management
A machine learning-based manufacturing decision system that generates real-time production decisions directly from learned distributions, enabling faster response to disruptions.
Manufacturing operations managers and digital transformation teams struggle to coordinate complex, highly autonomous production systems fast enough to respond to disruptions and changing demand. Current Manufacturing Execution Systems (MES) rely on rule-based or iterative optimization approaches that become slow and difficult to scale as system complexity grows. This technology introduces a novel class of manufacturing decision models that generate system-level decisions directly from learned decision distributions, rather than repeatedly solving optimization problems. Compared with conventional MES scheduling and optimization, the approach has demonstrated orders-of-magnitude faster decision generation while maintaining strong alignment with human intent in simulation studies. The technology has been validated in simulation using a realistic digital twin of a reconfigurable manufacturing system. If adopted, it could enable real-time, human-in-the-loop orchestration of large-scale manufacturing systems with significantly improved responsiveness and flexibility.
Technology Validation
The approach was evaluated in simulation using a digital twin of a multi-stage manufacturing system subject to disruptions. Results showed rapid generation of feasible system-level decisions that closely matched specified performance targets without iterative optimization.
Advantages
-Substantially faster decision generation compared to traditional optimization-based MES tools
-Scales to complex, high-dimensional manufacturing systems with many autonomous resources
-Supports natural human-in-the-loop interaction rather than fixed rule programming
-Compatible with existing MES and digital twin infrastructures
-Reduces reliance on handcrafted models and manual tuning
Applications
-Real-time production planning and rescheduling in discrete manufacturing
-Rapid system reconfiguration after machine or labor disruptions
-Human-in-the-loop decision support for advanced automation and robotics
-Digital twin-enabled operational decision support platforms
Next-generation manufacturing execution and orchestration software
TRL: 3
Intellectual Property:
Provisional-Patent, 2026-03-03, United States