Impact
FACTORY-AI is designed to improve the way factories sense change, coordinate machines, and run AI at the edge. Its evaluation connects network performance to practical manufacturing decisions. The figures below are proposed validation targets, to be tested experimentally.
5 ms
Target wireless latency, with jitter at or below 0.5 ms, for responsive industrial coordination.
≥99%
Target anomaly-alert reliability, supporting timely and dependable maintenance decisions.
≥20%
Target reduction in energy per visual inference, helping make edge AI more efficient.
A 36-month path from algorithms to field experiments
Year 1 • Establish the foundations. Develop semantic tokens, the reinforcement-learning orchestrator, cross-thrust interfaces, digital twins, and initial DISCOVER demonstrations.
Year 2 • Integrate the intelligence. Connect federated learning, LLM-assisted orchestration, hardware-aware mapping, and runtime control; target an integrated PROTON demonstration.
Year 3 • Validate the complete system. Evaluate multiple service slices across both testbeds and prepare software, datasets, documentation, and reproducibility materials.
Build capability beyond the prototype
Education • Give students hands-on experience across sensing, private wireless networks, AI, and hardware systems.
Open research • Plan reproducible experimental workflows and shareable research outputs that help others evaluate and extend the work.
Industrial relevance • Connect academic advances to robotic coordination, predictive maintenance, and visual quality inspection, with pathways toward other connected infrastructure.