FACTORY-AI

AI-native networks for smart manufacturing

 

Sensing. Communication. Computing. One intelligent control loop.

FACTORY-AI brings perception, wireless network orchestration, and heterogeneous computing together to help factories respond faster, operate more reliably, and use energy more efficiently. The project connects AI-driven sensing with programmable O-RAN networks and hardware-aware edge intelligence.

Arizona State University & Clemson University

Built around real factory decisions

Robotic coordination — Connect machinery and autonomous guided vehicles with responsive, dependable wireless control.

Predictive maintenance — Turn sensing and learning into timely anomaly alerts that help reduce unplanned downtime.

Visual quality inspection — Move AI inference closer to production while managing latency, energy, and compute resources.

Our ambition: a faster, more resilient factory

The proposed validation targets include 5 ms wireless latency with jitter at or below 0.5 ms, at least 99% anomaly-alert reliability, and at least 20% lower energy per visual inference. These are research targets to be evaluated through experiments.

One Connected Architecture

One connected architecture

From actionable sensing to programmable networks and efficient computing, FACTORY-AI closes the loop with real execution feedback.

Original architecture figure from the proposal.

Three thrusts

Three thrusts. One reflex loop.

Semantic perception, O-RAN edge intelligence, and heterogeneous compute infrastructure continuously adapt together. Hardware-aware feedback helps the system respond to changing factory conditions.

Original research-thrust figure from the proposal.