Pacman AI framework for controlling fusion systems safely makes key decisions (www.pppl.gov)

Researchers at Princeton Plasma Physics Laboratory and Princeton University have unveiled a groundbreaking AI framework named PACMAN (Prediction And Control using MAchiNe learning), successfully tested in five real-world fusion experiments. Designed to control tokamak systems, PACMAN allows AI to make rapid decisions autonomously, responding in approximately 20 milliseconds—much faster than human operators. This speed is crucial for managing the unpredictable nature of plasma, ensuring safety and stability in fusion reactions, which are fundamental to unlocking limitless clean energy. The PACMAN framework functions as an integrated control loop, incorporating various machine learning models that communicate and adapt in real-time. It not only anticipates plasma instabilities but also adjusts heating systems and other parameters dynamically, enhancing overall efficiency. Notably, it can predict instability events around 200 milliseconds in advance, enabling proactive corrections rather than reactive measures. The modular design allows easy integration of new models, paving the way for broader applications across different tokamak systems, thereby laying a foundation for the future of fusion energy research and its scalable implementation in various experimental setups.