PACMAN integrates real-time AI prediction and control across five DIII-D fusion experiments
PPPL and Princeton researchers published the PACMAN modular real-time AI control architecture after deployment on the DIII-D tokamak across five experimental control use cases, including reinforcement-learning heating control, plasma-event prediction and instability avoidance. The work is peer-reviewed in Nuclear Fusion; it remains a DIII-D demonstration rather than an independently reproduced multi-device result.