LLM-Driven Multi-Agent Control for Skill-Based Smart Manufacturing
LLM agents controlling a simulated six-module factory reach a 93% solve rate and diagnose silent faults.
The paper studies LLM agents for skill-based smart manufacturing, generating production sequences offline and handling runtime faults online. Each module is paired with an agent and an MCP server that exposes skills through OPC UA calls, while agents coordinate over MQTT and are grounded by live factory state. In a simulated six-module hexagonal factory, monolithic and peer-to-peer architectures both reach a 93% mean solve rate across nine challenges. An orchestrator uniquely resolves a silent conveyor-belt fault in all ten runs by rerouting plates, and every architecture shows emergent fault diagnosis without explicit failure logic.
- Each factory module has an LLM agent and an MCP server exposing OPC UA skills.
- Agents coordinate over MQTT using injected real-time factory state.
- Monolithic and peer-to-peer designs both reach a 93% mean solve rate.
- The orchestrator reroutes around a silent conveyor fault in all ten runs.
- Fault diagnosis emerges without any explicit failure-handling logic.
From the original article
1 min read at arxiv.orgFactories are shifting toward smaller lot sizes with high product customization, requiring frequent re-programming of flexible and reconfigurable automation systems.
By Kay Köhle · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.