Federated production planning architecture
Started a research project on architecture for production planning support in large, distributed networks of organisations. Demand is described in natural language; product, process, machine, material, and spare-capacity data live in many independent plants. The system under study should: turn a need description into a formal specification; check units, contradictions, and completeness; find products, substitutes, and manufacturing processes; ask plants for local feasibility confirmation; assign tasks via mathematical optimisation; reserve resources and replan when system state changes; keep an audit trail for every decision. Key assumption: limited role for the language model. An LLM may propose a reading of the text, but is not the source of truth about plant capability. It does not generate solver code, approve feasibility, or make reservations. Its output must pass schema-valid formal checks. Not a claim to plan an entire economy or guarantee optimal allocation. Goal: study concrete IT infrastructure that could support production coordination in industrial consortia, the public sector, crisis logistics, cooperative networks, or socialist planning systems. Research question: can a federated architecture combining controlled natural-language processing, hybrid search, local feasibility assessment, and mathematical optimisation improve production resource assignment under distributed, heterogeneous data?