The next step for smart factories isn’t collecting more sensor data — it’s building machines that can reason about what they sense and act on it without waiting for the cloud.
A recent project from Edge Impulse shows how far this idea has come. Engineers built a conveyor prototype that runs entirely at the edge: a vibration sensor feeds an on-device anomaly detection model that learns what healthy operation looks like, so it needs no library of labeled failure examples. When vibration drifts outside normal patterns, the board doesn’t just raise an alarm — three local AI agents take over. One triages the fault, a second opens a maintenance window and coordinates with the enterprise, and a third re-plans production around the downtime. Key performance indicators like OEE update in real time through the whole cycle.
Everything runs on the new Arduino VENTUNO Q, with a Unified Namespace (UNS) built on MQTT serving as the plant-wide data backbone — an architecture where every sensor, agent, and service reads from and writes to one shared, event-driven model of the factory. Agent-to-agent communication handles task delegation, keeping the flow of state and the flow of work cleanly separated.
What stands out is the closed loop: vibration anomaly detection flows into maintenance scheduling and then production re-planning, all at the machine itself, with no cloud round-trip required. For maintenance and reliability teams, this is a practical blueprint for moving beyond fixed threshold alarms toward genuinely autonomous asset care.
Source: Marc Pous via Edge Impulse blog
Read the full article here: https://edgeimpulse.com/blog/from-vibration-to-repair-an-edge-agentic-ai-conveyor-with-edge-impulse-uns-and-arduino-ventuno-q/
Why it matters to Intelligent ShopFloor readers: This architecture mirrors exactly where shopfloor AI is heading — unified data models, edge intelligence, and agents that turn condition monitoring into decisions. The project’s separation of state (UNS) from tasks (agent-to-agent messaging) is a design pattern worth adopting in our own deployments.